《Artefact :2023零售业数据洞察报告(英文版)(44页).pdf》由会员分享,可在线阅读,更多相关《Artefact :2023零售业数据洞察报告(英文版)(44页).pdf(44页珍藏版)》请在三个皮匠报告上搜索。
1、DT FOR RETILWE UNLEASH THE FULL VALUE OF DATA THROUGH DEMOCRATIZATIONDATA ACCELERATION PROGRAMS|AI SOLUTIONS|DATA MARKETINGArtefact is a global leader in consulting services,specialized in data transformation and data&digital marketing,from strategy to the deployment of AI solutions.We are offering
2、a unique combination of innovation(Art)and data science(Fact).18COUNTRIES+1000CLIENTS+1300EMPLOYEESLos AngelesMexicoColombiaNew YorkThe NetherlandsGermanyUKSwitzerlandLebanonFrance UAESpainSingaporeMoroccoChinaSouth KoreaBrazilMalaysiaSaudi ArabiaTABLE OF CONTENTSData for Retail4 Introduction by Vin
3、cent Luciani CEO&co-founder of Artefact7 Becoming data-driven(and reaping immediate benefits)has never been easier in the retail sector9 CARREFOUR GROUP How Data&AI can accelerate sustainable business transformation 12 Data-driven marketing:the rise of the customer data platform15 CONFORAMA AI-enabl
4、ed personalization boosts Conforama CRM campaign revenues19 Retail Media:An indispensable asset for brands21 UNILEVER How does Artefact support UNILEVER on Retail Media use cases to increase its sales?31 Data monetization opportunities for retailers:Retail Media within the CPG/Retailer data ecosyste
5、m34 CARREFOUR LINKS With Artefact are helping brands to improve media targeting thanks to Carrefours data36 MATTEL RelevanC Advertising How Artefact helped boost Mattels online sales on Cdiscount(RelevanC Advertising)retail media platform?38 CARREFOUR GROUP How to reduce food waste in the Bakery-Pas
6、try department?41 Demand forecasting:Using machine learning to predict retail sales43 Sales forecasting in retail:what we learned from the M5 competitionDATA FOR RETAIL34Interview with Vincent Luciani about Data&AI market perspectivesThe strategic importance of data for companies is no longer in que
7、stion.Aware of this reality,Artefact helps companies to capitalize on this performance,growth and improvement lever.How would you analyze the latest trends in business data transformation?The shift to a data-driven business model where decisions are made based on what we know to be true rather than
8、on our intuition is at the heart of the wave of digital transformation that has been sweeping through all sectors in recent years.Data analysis helps us respond with more certainty in the face of uncertainty.When wars and pandemics disrupt the order of things and lead to massive inflation,using adva
9、nced data analytics to make better decisions becomes more critical for businesses.AI also needs data to learn,which frequently means dealing with sensitive personal data.For this reason,it is imperative for organizations to deploy responsible and trustworthy AI.It is key to ensuring that the values
10、of inclusiveness and diversity are respected.Data has an essential role to play in creating a more ethical and just world.Another major emerging trend,linked to the acceleration of global warming,is the use of data as a tool to help reduce ecological impact.Data and algorithms can measure the carbon
11、 footprint of activities and analyze how to develop products,services and infrastructures in more energy-efficient ways,by identifying sources of waste and inefficiency.Artefact identified two drivers:first,to structurally reduce the carbon emissions of logistics and digital infrastructures,and seco
12、nd,to engage consumers to participate in this ecological transformation by displaying the carbon content of their shopping baskets and recommending actions to compensate for their emissions.Business data maturity has advanced rapidly over the past decade.How has Artefact evolved as a leader in data
13、consulting services?Companies have implemented data governance policies,which are a prerequisite for any transformation,but there are still sectors that lag far behind in terms of their data processing,with a real potential for efficiency,such as healthcare or“heavy”industries.This is especially tru
14、e in comparison to the consumer and retail sectors,which have begun their data revolution and which we know very well,such as LOral,Danone,Unilever,Samsung,etc.We started to transform marketing departments by making them more profitable and relevant in their multi-channel media investments with pion
15、eering targeting,measurement and personalisation solutions.For the past few years,we have also been deploying acceleration programs in all business areas(Sales,Supply Chain,Operations,Call Centers,HR&Finance,etc.).We create value wherever there is data,and with our clients,we improve their processes
16、 and produce customized business applications.Can you give us a concrete example that shows,from a precise business and operational objective,how Artefact designs AI solutions that improve business uses?Data is the key to understanding customers,developing better products 5INTERVIEWand services,and
17、streamlining internal operations to reduce costs and waste.For example,weve been working with the Orange telecommunications group for over six years,and among the many use cases for leveraging the companys automation and AI potential,we deployed a solution with their teams to optimize their technici
18、ans interventions on the fiber network.The solution is based on visual recognition technology that helps operators improve the quality of their installations or repairs.This application,available on a tablet,is currently used by more than 10,000 Orange technicians throughout the country a resounding
19、 success!This case perfectly illustrates Artefacts firm belief that to achieve true data maturity,companies have no choice but to make data accessible to everyone:not only to experts,but also to operational staff in the field.This will lead to new forms of augmented work,where applications and their
20、 interfaces put intelligent information in everyones hands to work more efficiently and with more autonomy.It is key to ensuring that the values of inclusiveness and diversity are respected.Data has an essential role to play in creating a more ethical and just world.”“6Democratizing data and making
21、it accessible to all is key to accelerating business and creating value.”“Listening to you,its clear that data should no longer be a subject reserved for experts only.How does Artefact see the vision of data democratization materializing?The companies that will endure are those that successfully fos
22、ter a data culture with access to knowledge and data for all.Weve undertaken several initiatives in this area that are highly strategic to Artefacts positioning as a major player in data democratization,in order to fully exploit its potential for positive transformation.The first was the launch of t
23、he Artefact School of Data two years ago,a key pillar in our strategy of providing clients with training adapted to the constantly evolving skills of the data industry.Were also developing“la carte”e-learning platforms for clients to quickly share knowledge of data and AI with all of their employees
24、.And at the beginning of this year,we launched our Artefact Research Center under the leadership of Emmanuel Malherbe(class of 2008),to bridge the gap between fundamental research and its democratization for business applications,in collaboration with clients who provide access to their data and use
25、 cases.To achieve this,we have partnered with several renowned university laboratories,including CMAP and MEI at Polytechnique,to host PhD students at Artefact,who work on data model improvement and organization studies.They will publish scientific articles and participate in international conferenc
26、es to share their knowledge.These programs are just the first steps were taking to democratize data and help our clients transform faster and better.After a year of robust growth in 2022,what are your forecasts for 2023?After+50%organic growth in 2022,our objective is to maintain the momentum in 202
27、3 with a sustained recruitment policy in France and in our 16 subsidiaries in Europe,MENA,Asia,North and South America.Weve just deployed our Artefact Africa entity from Morocco and will soon open an office in Korea.We will also accelerate our development in LATAM and the United States.Artefacts exp
28、ansion also involves an ambitious M&A policy that will continue in 2023.In 2021,two acquisitions in particular enabled us to expand Artefacts portfolio of clients and services:the acquisition of Startup Inside,a pioneer in open innovation&intrapreneurship strategy consulting and international Data a
29、nd AI conference organizer,and,more recently,the merger with the Arca Blanca group,a leader in data consulting in the United Kingdom.We are quite optimistic about the future because even though the economy is currently strained,companies need to better understand the shifting environment and find ra
30、pid solutions for adaptation and progress through data.Becoming data-driven(and reaping immediate benefits)has never been easier in the retail sectorData technologies are now at the head of the aisle in the great supermarket of retail opportunities.Serve yourself:take advantage of these unbeatable o
31、ffers today!Particularly turbulent market conditions in 2022 disrupted the retail sectors pricing and supply chain policies.While price optimization has always been a daily concern for this industry,the 5.2%increase in consumer prices over the whole of 2022(compared to 1.6%in 2021)and its accelerati
32、on in the first quarter of 2023 have intensified the need to react.More frequent price adjustment decisions must be made every day to absorb cost increases or react to competitor repositioning,and be immediately deployed in store networks.To make matters worse,a shadow still hangs over supply chains
33、,inherited from pandemic events.Many retailers are no longer sure they can offer all products due to lack of stock.To address both of these challenges,players who are able to rely heavily on their data and process it instantly and at scale through machine learning stand out.Of course,you cant go fro
34、m data-rich to data-driven overnight,but contrary to common fears,the maturity of the technologies is such that the first building blocks,which can rapidly bear fruit,can be deployed in mere weeks.Machine learning to optimize pricesInflation has focused retailer attention on the daily challenge of p
35、rice optimization:how to identify products that are most sensitive to price variations(elasticity)?Which products shape the price image of a brand(known value items)?How to react in real time to competitor repositioning?Since promotions represent a growing share of sales(more than ever,consumers are
36、 looking for bargains that fit a tight budget),how can they be optimized without impacting profitability?All the answers are in the data!Customer data and proprietary sales history can be combined with external data such as competitor price and promotion records,seasonality,calendar events,and,yes,e
37、ven the weather(!)to make the right decisions.Retailer pricers and category management teams usually do all this.Unfortunately,they often need to work in Excel but have limited time to devote to it,due to their other responsibilities.So,whats new?Why,its the availability today of proven off-the-shel
38、f algorithms that automate the simplest decision-making Jrme PetitManaging Partner Retail&eCommerceARTEFACTDATA FOR RETAIL7processes.Its the seamless and massive availability of third-party data in the market.Its the ease of use of technology stacks that allow millions of transactions to be processe
39、d in a few milliseconds.Today,it only takes three months to build a data platform that combines all transaction data,promotions,stock,product hierarchy,store hierarchy,customer data,etc.And a technology partner can manage the infrastructure,resource deployment and network dimensions in the cloud thr
40、ough its managed services.Today,like Monsieur Jourdain,if you know Excel and PowerPoint,youre a data analyst without realizing it:in a matter of days,you can take control of data in BigQuery(Google),Synapse(Microsoft),or Snowflake and build interactive dashboards in Looker,Power BI or Tableau.and be
41、tter manage inventoryIn recent years,the health and geopolitical context has also challenged supply chains.Today,the supplier service rate varies from week to week and delivery times can be very uncertain.Distribution channels have also grown highly complex:not only do stores need to be stocked,but
42、home deliveries,click-and-collect,and partnerships need to be served as well.Once again,data science comes to the rescue of retailers by allowing them to better control inventory management.By leveraging machine learning,retailers can now analyze receipts in real time to immediately detect out-of-st
43、ock items,calculate the spread of uncertainties across all links in the chain to better size buffer stocks,or improve stock allocation under an infinite number of constraints(to optimize costs,shorten delivery times,or reduce carbon footprint).Democratizing data use throughout the company:data is ab
44、out peopleIn a business where margins are so tight that operational excellence is a necessity,the notion of a data-driven company is far from new.Whats changing today is the ease of access and use of technological platforms.If technology is no longer a barrier,the challenge is still to make these so
45、lutions available to the widest possible audience.To democratize their use,simple solutions need to be deployed on a massive scale,employee training programs need to be multiplied,whether on-demand or more intensive,and events(e.g.,hackathons)need to be organized to engage the managers who are drivi
46、ng the transformation.Data is about people.This is Artefacts slogan,and rightly so.Directly monetizable data But perhaps not immediately for everyoneThe icing on the cake is that data itself is a goldmine,thanks to retail media and data sharing.As digital signals become more difficult to capture,the
47、 billions of transactions and customer interactions that retailers generate have become a critical strategic advantage for them.This data,which provides in-depth understanding of consumer expectations,has great potential for monetization.But it represents a profound,existential transformation of the
48、 retailer business model:moving from a self-financed model(with negative working capital)but with very narrow margins,to a model where initial investments are substantial but margins are high.A Copernican revolution,perhaps not the easiest to undertake for all players.In the great supermarket of ret
49、ail value creation opportunities,data technologies are now at the top of the shelf,in self-service.Retailers,why wait to share these unbeatable offers with your partners?8How Artefact is helping Carrefour achieve carbon neutrality for its e-commerce activity?CARREFOUR GROUPHow Data&AI can accelerate
50、 sustainable business transformationCASE STUDYCarrefour Group is the leading European retailer and the worlds second largest retailer,and is present in more than 30 countries.Carrefours international profile raises a number of ecological challenges and a desire to offer its customers,regardless of t
51、heir level of awareness,quality food and services accessible to all.The Groups aim is to become the world leader in food system transformation for all by committing to four major objectives,including achieving carbon neutrality by 2030 for its e-commerce activities.To achieve carbon neutrality,three
52、 main levers of action have been identified:Measuring the ecological impact of a delivery in order to manage the strategy;Reducing the carbon emissions of its logistic and digital infrastructures;Engaging customers to become participants in ecological transformation.This aim also provides a triple o
53、pportunity for the Groups e-commerce activity:reducing its operating costs,improving its NPS score(customer satisfaction indicator)and anticipating legal changes.To seize these opportunities and take concrete action on each of these levers,Carrefour must be able to measure all greenhouse gas emissio
54、ns by drawing on real data that compiles all data storage,transport and logistics activities,from first click to final delivery,whether to the home or by store pickup.CHALLENGES9DATA FOR RETAILDefining a reliable,actionable,transparent carbon measure for Carrefour and its customers“The challenge we
55、gave Artefact was to calculate the CO2 emission of an online order.How much CO2 will a customer produce if their order is delivered or if its picked up at the store?”Bertrand Swiderski Chief Sustainability Officer CARREFOURThe first step for the Artefact and Carrefour teams was to agree on the scope
56、 of action for measuring this carbon footprint.They decided to limit themselves to measuring greenhouse gas emissions generated by e-commerce orders in 2021SOLUTIONThe second step was to collect activity data in order to convert it into carbon emissions.As this data wasnt already present and documen
57、ted in Carrefours data platform,the business teams(logistics,warehouses,e-commerce)had to be brought together to obtain it.This step proved to be crucial to the operations success,as it allowed all stakeholders to become ambassadors for the groups“carbon neutrality 2030”objective.Carrefours strategy
58、 for measuring its carbon footprint was based on a systemic,unifying,long-term,iterative approach.The strategy was successful thanks to the participation of over 30 employees and the involvement of Carrefour customers via their“Engaged Consumers Clubs”.“Today,we recognize that consumers are becoming
59、 experts on these topics.They want to understand how things are done and want to challenge companies.Thanks to them,the project has matured.”Lonard Cahon Consulting Manager-ARTEFACTEncouraged by these initial results,Carrefour will continue its commitment by publishing the carbon footprint of each o
60、f its orders on its e-commerce site in the near future.“Soon,customers will clearly see the number of kilograms per CO2 on their orders,thanks to the insights gained from our carbon assessment.”Manuel Chatain E-commerce CSR Manager-CARREFOUR 10RESULTSOpening a wider field of possibilities and ecolog
61、ical alternativesBy analyzing its carbon footprint and implementing this first measure,Carrefour now has a way to pilot its e-commerce carbon emissions reduction strategy.The group can now encourage its customers to review their consumption patterns in order to be more responsible,encourage its serv
62、ice providers to reduce their emissions,and also promote internal awareness by proposing several possible levers of action:Act on the choice of delivery slots in order to optimize truck loading,routes and schedules;Increase the number of clean vehicles(biogas,electric or hydrogen)by 2030;Reduce the
63、amount of packaging used.And to ensure the sustainability of this measurement and its easy use by the teams,Artefact teams worked on three elements:A dashboard to run trajectory simulations by combining forecasted activity data;Training to teach how to use and modify the dashboard;Detailed documenta
64、tion to enable employees and clients to understand and reproduce the measurement process from start to finish.Using data as a key lever to help businesses achieve their environmental objectivesCarrefours e-commerce,supply chain and logistics platform teams worked together to meet this challenge,supp
65、orted by the collaboration and expertise of the Artefact,Aktio and Google teams.The project is part of Artefacts“Data for Sustainability”solutions,which aim to create a positive impact on the environment through data by accelerating the ecological transformation of businesses.“We expected a very cle
66、ar vision of what each basket order would emit.Our request was complex,but Artefact responded to it with great dynamism and agility.”Bertrand Swiderski Chief Sustainability Officer-CARREFOUR“At Artefact,we believe data will play a major role in helping companies achieve their carbon neutrality goals
67、.”Vincent Blaclard Partner-ARTEFACT11Data-driven marketing:the rise of the Customer Data PlatformEverything seems to justify the current explosion of the Customer Data Platform(CDP)market.CDPs main advantage over older generation Data Management Platforms(DMPs)is that they ea-sily integrate identifi
68、able first-party data(email,phone number)and arent dependent on using third-party cookies or browsing data to refine customer and prospect knowledge.CDPs are a true asset in a world that is becoming increasingly cookie-and ad ID-free.At a time when the pandemic is forcing brands to digitise at break
69、neck speed,and when the transforma-tion of the technical and regulatory environment surrounding advertising trackers is forcing data marketers to revise their approaches,CDPs are here to optimise the customer experience.Florian ThiebautManaging Partner-Data-Marketing ARTEFACTA game-changing technica
70、l and legal environmentFollowing Safaris lead in 2016,the worlds three main browsers eliminated(or will eliminate)the use of third-party cookies.On the mobile/tablet devices side,Apples iOS 14 now requires explicit consent for any mobile ID collection.As for regulation,GDPR laws in Europe have given
71、 consumers more control over their personal data,requiring them to give explicit consent for the use of cookies.This regulation represents a major shift in the world of data-driven marketing,as it has reduced the number of cookies placed on European devices by 30%.This global trend restricting the u
72、se of IDs and advertising cookies sharply impacts the targeting capabilities of advertisers,who are often dependent on third party data.The vast majority 12of them use or have used retargeting and old generation DMPs that rely heavily on segments fed by third party data.Along with targeting,measurem
73、ent must also be transformed.With more stringent consent collection requirements,its more difficult to collect the consumer IDs needed to track impressions,clicks or views,and reconstruct complete customer journeys.Four pillars for a sustainable data strategyTo maintain the same performance and diff
74、erentiate themselves from the competition,advertisers must design a sustainable data strategy and exploit their customer and prospect data to its full potential.This requires focus on four actions:The CDP:The first step is to establish a CDP environment based on a suite of tools that is both complia
75、nt and sustainable.This will enable data to be collected,stored,processed,visualised and activated,whatever the source.From this foundation,the focus must be on first-party data.Data governance:Brands need to rethink data governance and processes to enable secure and compliant end-to-end data collec
76、tion.Audience segmentation:This data,centralised for a unified view of the consumer,can then be used to create new audience segments and define new metrics for measuring campaign results.Second-party partnerships:In addition,its becoming increasingly strategic to form so-called“second party”partners
77、hips with other partner companies to exploit first-party data and create win-win situations.This data completes a database that is incomplete at certain points in the consumer journey.Examples might be an agreement between an FMCG brand and a retailer,a mobile phone manufacturer with a telco or a ho
78、tel chain with an airline.Three types of data to activate via a suite of toolsFirst-and second-party data are key to meeting the challenges of the post-cookie world.But what are they and what tools can be used to manage them?PII or Personally Identifiable Information is essentially CRM(customer rela
79、tionship management)data.It can precisely identify an individual and is often an email address or a phone number for example.Once anonymised,it can be used via the APIs of media partners(e.g.,Google Customer Match,Facebook Custom Audience/conversion API,DATA FOR RETAIL13Amazon,WeChat,etc.)to build a
80、udience segments,perform audience extensions,and reconstruct paths to measure the influence of digital campaigns on offline sales,etc.Non-PII data can be browsing data that cannot lead directly to the identification of an individual.It can be used to build more granular segments via analytics and au
81、dience creation solutions for measuring precision marketing actions without relying on third party dataData that is purely media-related,such as campaign impressions,video views and click rates,is more voluminous and less granular than the other two types of data.It is more difficult to use but ther
82、e is a robust market of tools capable of treating it in a secure and compliant manner,such as Google Ads Data Hub,Facebook Advanced Analytics and Amazon Marketing Cloud.These different data flows are injected into an ecosystem of interconnected tools,which are useful for a range of tasks from data c
83、ollection to performance measurement of the actions carried out and can be activated on all channels,whether media,direct marketing or site personalisation.This entire ecosystem,the result of all the connections built between the different tools already used by the company(also known as“full-stack”s
84、olutions),is what is called the CDP.When it comes to the adoption of this way of working,the numbers dont lie.Fundraising for CDP providers is soaring,the tech giants are all positioned in the sector,and the number of users is exploding.In fact,according to the Customer Data Institute,the market inc
85、reased 30%from$1 billion in 2019 to$1.3 billion last year.Estimates see this figure reaching$1.55 billion in 2021 as conditions are even more favourable for the adoption of CDPs.As the data-driven world continues to evolve at a rapid pace,there seems little doubt in the business value of the CDP.Now
86、 is the time for organisations to consider deploying this future-facing technology.14Conforama is the second largest home furnishings retailer in France and is present in seven countries,with 300 stores,including 200 in France.The company sells furniture and decorative items in kit form and posted s
87、ales of 1.7 billion euros in 2022.As a gateway brand,Conforamas goal is to“Make what people want most accessible at the best price.”Its an ambition backed by a transformation plan to deliver an omnichannel experience through data and AI.An initial audit and data marketing vision with Artefact identi
88、fied and prioritized 12 use cases and 25 technical and organizational enablers.The first use case was to integrate a personalized product recommendation into the companys weekly emails.Several challenges needed to be addressed through this use case:How to understand the needs of three million custom
89、ers and recommend the most relevant products from a catalog with 42,000 references?How to propose only products currently in stock,on promotion,and not already suggested to customers?How to easily operate and maintain the technical solution?CHALLENGESCONFORAMA AI-enabled personalization boosts Confo
90、rama CRM campaign revenuesCASE STUDY 15Saving consumers time,improving business productivitySOLUTIONBy using machine learning algorithms to analyze user data,such as preferences,purchase history and online behavior,artificial intelligence-based product recommendation suggests products relevant to co
91、nsumers in a personalized way.This allows companies to better understand their customers needs and recommend products that match their interests,resulting in increased sales and customer retention.One of the main benefits of this solution is that it saves customers time.Rather than scrolling through
92、 countless product pages to find what theyre looking for,customers can quickly access a selection of recommended products that specifically meet their needs.AI-based product recommendation can enhance the online shopping experience and encourage customers to return for more purchases.A strategic adv
93、antage,given that 72%of consumers only interact with marketing messages that are personalized and tailored to their interests.In addition,AI-based product recommendation can boost business productivity:machine learning algorithms can analyze large amounts of data in real time,allowing companies to c
94、ontinuously monitor customer trends and buying behavior.This can help organizations better understand customer desires and quickly adapt their product offerings accordingly.It can also enable companies to optimize their inventory by offering products that are more likely to sell,which can lower cost
95、s and maximize profits.Lastly,AI-based product recommendation can offer significant business benefits.By suggesting relevant and personalized products to customers,companies can improve their conversion rate,increase sales and strengthen their brand image.From a market perspective,AI-based product r
96、ecommendation has been shown to deliver+2.5%incremental growth.“Time savings,yes,but above all a business benefit for our CRM teams.Because thanks to this personalization,customers click more and therefore buy more.Weve gained 15%of the click rate following the personalization of these emails,which
97、represents several million in incremental sales.”Mlodie Charles,Marketing Director CONFORAMA16A first use case focused on email campaign personalizationPrior to this project,all Conforama customers received emails featuring the same eight products selected each week by the marketing teams.This was a
98、 labor-intensive task,as it required identifying the eight products most likely to interest three million customers,each of whom had unique interests.All this time spent analyzing data could have been spent on more strategic activities,such as creating editorial content for those emails.Today,an ema
99、il is sent to every Conforama customer each Tuesday containing eight product recommendations.But these recommendations are personalized according to purchase history,and filtered exclusively for products that are on sale,are available in stores,and that havent been featured in previous activations.T
100、he implemented AI solution includes 4 main data processing steps:Collection of transaction histories,customer and product references,then data preparation;Building the“Collaborative Filtering”model to calculate customer appetite for the product catalog;Product filtering based on available inventory,
101、commercial news(sales,promotions,etc.),past activations and purchases;Product data enrichment(photos,prices,descriptions,etc.)for activation.This solution is based on 16 data tables,25 transformation and modeling steps,and 40 automated quality tests.Dozens of iterations of the model made it possible
102、 to choose the most efficient approach based on transaction history.Thanks to this solution,Conforama now generates several million recommendations each week in 45 minutes at a cost of 50 euros per week.In other words,if you count development and operation costs,as well as incremental sales,the proj
103、ect break-even point is reached in one week,with an automated and reliable solution.17DATA FOR RETAILRESULTSA smooth transition to AI:lessons from Conforamas success storyFor many players,there are three challenges linked to their level of maturity:LEVEL 1Personalizing a currently rule-based touchpo
104、int using an AI algorithmic approach;LEVEL 2Extending AI-based personalized recommendation across the entire customer journey(similar products/complementary products/suggestion based on purchase history);LEVEL 3Optimizing the orchestration of recommendations across channels to ensure an omnichannel
105、experience.Level 1 is often the most difficult,as it requires laying the foundations for four separate dimensions:target vision,user experience and priorities;data sources;technological tools;project team and work method.The Conforama example offers valuable lessons about these four dimensions:Selec
106、t a first use case and functionalities that can be quickly implemented and measured to put the organization on the road to success.For example,this initial victory means Conforama can now plan the deployment of product recommendations in stores or the improvement of their algorithm thanks to browsin
107、g data.Ensure the data is reliable.Good data modeling relies first and foremost on good quality data.For Conforama,exploratory analyses were performed on more than 50 tables to select data sources in areas such as customer knowledge,product repositories and transactions.Use technologies that allow t
108、eams to deploy a technical solution quickly and collaboratively.Conforama selected the most appropriate tools for this type of workflow:DBT,BigQuery ML and Vertex AI for their performance,modularity and portability.Build a dedicated team capable of dealing with all potential problems,and adopt a tes
109、t and learn approach.To do this,a multidisciplinary IT/Conforama business team was formed,and a 2-week sprint approach was adopted.18Retail Media:An indispensable asset for brandsWhile Retail Media represented only 9%of digital media investments for brands in 2019,it will soar to 43%of these investm
110、ents in 2023 and is expected to double in value by 2024 to reach 100bn.Vincent Cailliot,Director of Data Consulting and Sidney Zeder,Senior Consulting Manager Data Marketing,both of Artefact,explore the opportunities of retail media for Consumer Packaged Goods(CPG)brands.Vincent CailliotDirector Dat
111、a Consulting ARTEFACTSidney ZederSenior Consulting ManagerARTEFACTRetail Media investment driven by regulatory and tech developmentsThe increased importance of Retail Media in the digital strategies of brands can be explained by the evolution of the availability of consumer data,which is at the hear
112、t of any media personalization strategy.In the past,third-party cookies were mainly used to drive this strategy.Regulatory developments to better protect consumer data,including the GDPR in Europe,have led to new technological constraints,making cookies obsolete:This process began as early as 2016,w
113、hen Safari removed cookies from its platform.It continues today with Google Chrome,the most popular browser,announcing the removal of third-party cookies from its platform next year.As a result,brands need to find new data sources to build their digital activation strategy.One solution is for brands
114、 to better leverage their first-party(1P)data assets,i.e.,their proprietary data,by collecting more and better data from their customers.But to create digital activation strategies with long-term value,proprietary data is no longer sufficient:it needs to be enriched with data especially transactiona
115、l data from retail partners.Facilitated data sharing between brands and retailers allows brands to deploy ever more relevant digital marketing strategies with a high level of autonomy.This increased collaboration and data sharing between retailers and brands is possible thanks to“data clean rooms”su
116、ch as Amazon Marketing Cloud,LiveRamp or Decentriq,which allow DATA FOR RETAIL19the sharing of personally identifiable consumer data at the individual level in an anonymous way.A rapidly evolving ecosystemThe ecosystem of technology partners around Retail Media is highly fragmented and constantly ev
117、olving,with partners that are more or less specialized depending on major Retail Media activities:first-party,second-party or third-party cookie data collection tools,data processing and audience creation,activation or analysis,etc.The challenge for brands will be to identify which combination of te
118、chnology partners will best meet their needs,depending on their current technical ecosystem and their own business challenges.Retailers are also an essential part of this ecosystem,providing access to transactional data to build their Retail Media strategy.While the majority of retailers in the US h
119、ave launched Retail Media offerings,most retailers in Europe are still in the experimentation and use case-testing phases;few have yet industrialized use cases with brands.Valuable use cases beyond Retail MediaLeRetail Media allows brands to address marketing use cases from consumer insight generati
120、on to digital campaign activation and marketing performance measurement.The availability of transactional data(previously unavailable to CPG B2B2C brands)at the“individual”level enables the construction of insights and activation plans that are all the more impactful.The same data can be used to mea
121、sure their effect on sales and ROI,enabling effective optimization of activation plans.Retail Media is just the next step towards more collaboration between retailers and brands.In a long-term partnership perspective,collaboration and data sharing can enable the implementation of more advanced categ
122、ory management and supply chain use cases,such as the analysis of the long-term value of existing promotions or the prediction of in-store product demand and thusoptimize supply chain operations.Which Go-To-Market strategy to launch?For retailers,its important to define a new offer to monetize their
123、 data.This can range from monetizing their owned media inventory(website),to sharing data“as a service”in a clean room,to offering services(campaign management or reporting as proposed by Amazon for example).These new offers can be marketed internally or via partners.The internal or external develop
124、ment strategy will determine the associated costs,in terms of salaried resources(commercial and technical profiles to be recruited)and technical resources(clean room tools,technical infrastructure to be set up).For brands,the approach will be more traditional,from defining their business needs by id
125、entifying and prioritizing use cases,to setting up the partnership with their key business partners.Likewise,the implementation of pilots with a retailer to test the value of use cases can be carried out by a third-party partner.20UNILEVER How does Artefact support Unilever on Retail Media use cases
126、 to increase its sales?Thanks to Retail Media,Unilever identifies new growth opportunities and increases sales of its priority product categories.The global Unilever group has a portfolio of 400 brands that are anchored in the daily lives of its 5 billion consumers.Present in France for 125 years,Un
127、ilever is the leader in several market categories including ice cream,condiments and deodorants.The consumer packaged goods(CPG)sector has the particularity of being intermediated.Since the distribution of their products is carried out by different retailers,there is little direct relationship betwe
128、en CPG brands and their clients.Yet,consumer knowledge is key to optimize media and promotional strategies,product assortment in the territory or to identify new growth levers.It is from this challenge that the so-called Retail Media offers were born,i.e.2nd party data partnerships.A retailers data
129、is made available to a brand to enrich its own data assets in a win/win approach.This movement was initiated in 2012/14 by pure players such as Amazon and Alibaba,and traditional players such as Casino,Walmart and Carrefour have gradually followed.Accompanied by Artefact since 2019 on various data i
130、ssues,the Unilever Group seized this opportunity to identify new growth levers,develop a common consumer and product approach with a leading retailer and increase sales of certain priority product categories.This is the case,in particular,for the Magnum brands mini ice cream bars range.CASE STUDYCHA
131、LLENGES21DATA FOR RETAILRetail Media in a win/win partnership with Trade Marketing Advertising expenditure(media and traditional marketing)in the consumer goods sector amounts to nearly 680 billion dollars worldwide and 16 billion in France.The investments allocated to Trade Marketing are just as ma
132、ssive.While the advertising levers have been highly digitized over the last twenty years,Trade Marketing is still very underdeveloped.However,driven by e-commerce,it represents half of the budgets of CPG advertisers(600 billion dollars worldwide,16 billion in France).Trade marketing is strategic for
133、 these brands because it consists in carrying out actions in partnership with mass retailers to better meet consumers expectations:prospectus,merchandising,point of sale animations but also sales forecast and supply chain optimization.Thus,the digitalization of trade marketing represents a real grow
134、th opportunity for advertisers and brands.Retail Media is based on this concept of partnership in the promotion of brands products.The sharing of 2nd party data favors client knowledge and allows a better segmentation and therefore a better targeting of advertisements or promotions.This targeting ca
135、n be activated in both proprietary brand environments and in external audience extension environments.Retail Media infuses Unilevers media,promotional and supply chain strategyUnilever,together with Artefact,has identified 6 strategic axes to exploit the full potential of Retail Media:Media activati
136、on Measuring Customer Lifetime Value Coupon customization Optimization of store assortments Short-term sales forecast Supply of shelves Since the CPG group initiated these retail partnerships,each of these axes has been worked on through concrete use cases.For this,Unilever has benefited from Artefa
137、cts end-to-end support.This concerns data strategy,the launch of pilots,the construction of new data products,but also the provision of human resources(data analysts,data scientists and data engineers),as well as the training of Unilever employees.“Retail Media is a win/win strategy for brands and r
138、etailers.Retailers data allows us to enrich the shoppers knowledge and accurately measure our activities on all channels,throughout the transformation tunnel.For their part,retailers find a new source of revenue and differentiation from their competitors.In addition,it is a way to better satisfy the
139、ir clients with more personalized offers and a better anticipation of stock shortages.”Sarah Baqa Head of Performance Marketing-Unilever 2222An iterative approachThe Retail Media project was envisaged through an iterative approach in 4 steps:Identify and select the right distribution partners based
140、on their pre-existing relationship with Unilever and their technology infrastructure.Laying the foundations of the partnership,i.e.the use cases that can be implemented as well as the necessary legal and technical requirements.The modus operandi is also built to facilitate collaboration between all
141、stakeholders(agencies,internal and external teams of Unilever and the brand).Implement pilots for each use case.They allow for testing and optimization of devices on key Unilever brands(Magnum and Sun)before scaling up.Deploy devices on a greater number of brands,use cases,markets and partner brands
142、.The vision of this data partnership is therefore based on a virtuous circle:the measurement of actions carried out makes it possible to collect insights which in turn feed the next activations.”Our data strategy is really ambitious,so we wanted to be accompanied by a team of experts.We chose Artefa
143、ct,which already has experience with similar products and works in a very advanced way on data and Retail Media solutions.In addition,they have the technical capabilities and resources to take use cases to scale.This is the key to this long-term partnership.”Sarah BaqaSOLUTION23DATA FOR RETAIL23”Ret
144、ail Media allows us to activate audiences and precisely measure the link between digital actions(advertising,promotions)and sales.The data from the retailers loyalty cards also has the advantage of tracking consumers over time(Customer Lifetime Value)to see if they have increased their purchase freq
145、uency,if they have increased the value of their basket or if they have switched brands,etc.”Florian Thiebaut,Managing Partner Data Marketing-ArtefactA use case conducted on Magnums mini bars segmentFor example,retailers data was activated in the media to meet a challenge for the Magnum brand:to conf
146、irm its leadership during the summer period,which is key for this product line,and to acquire new clients in the Confectionery segment(mini bars).To achieve this,the first step was to use shopper data to identify the different consumer segments.We showed them the same banners during a defined period
147、.This allowed us to measure the increase in on and offline sales,but also to track the buyers recruited to retain them post-campaign.Through this first wave,Unilever was able to identify the best performing and most profitable audiences.The ultimate goal is to use these results at scale to target th
148、ese audiences with more personalized ad creative and promotional offers.This pilot campaign significantly increased the brands revenue.Unilevers data and Retail Media strategy will continue in 2022 alongside Artefact.In addition to the two Precision Media levers,the exploitation of the 4 other ident
149、ified work axes is also planned.RESULTS24DATA FOR E-COMMERCE&RETAILData monetization opportunities for retailers:Retail Media within the CPG/Retailer data ecosystemRetail media has been around for quite a while,but thanks to the evolution of new uses of consumer data,its potential is gaining attenti
150、on.Sidney Zeder,Senior Manager and Gatan Blan,Senior Data Consultant&Product Owner,both of Artefact,explore the opportunities of data monetization for retailers.Gatan BlanSenior Data Consultant&Product OwnerARTEFACTSidney ZederSenior Consulting ManagerARTEFACTRetail media has been on the rise on dig
151、ital platforms for the last six years,most notably on Amazon.The Covid-19 crisis accelerated this trend for traditional retailers.Retail media,in simple terms,is the means for retailers to sell media inventory on their e-commerce platforms.Because the Covid-19 pandemic fueled the shift to digital wa
152、ys of buying,such as e-commerce or click-and-collect,even for grocery shopping,retailers had no choice but to go with the flow.In fact,between 2019 and 2020,CPG e-commerce penetration increased by five points,from 10 to 15 percent.For retailers,the downside is that margins are lower in e-commerce th
153、an in brick-and-mortar.The upside is that by selling online,they collect a lot of consumer data that can be monetized or used to create new services.In a Goldman Sachs study,82%of CPG companies surveyed said they were already investing in at least one retail media platform.This represents approximat
154、ely 17%of digital budgets already allocated to retail media.Media investment has effectively shifted down the marketing funnel.Although many brands search investments still flow into the“Google family,”were seeing brands diversify their digital spend into e-commerce platforms to capitalize on the“se
155、arch destination”status they hold.When youre on Amazon as a consumer,youre very close to the“moment of truth”:youre in a purchasing mindset.Therefore,when youre on Amazon as a brand or product,the closer you can get to that funnel,the better.Goldman Sachs expects this trend to translate into a 6-8 p
156、ercent increase in total CPG e-commerce sales through retail media over the next four years.DATA FOR RETAIL25Retailer data monetization opportunities with CPG brandsThis close-to-the-funnel media investment trend has created opportunities for retailers around three types of data monetization with CP
157、G brands:1.Inventory monetization:traditional retail media consisting of selling media inventory on proprietary assets.This can be offline inventory retailers have long monetized their customers to offer coupons or specific promotions to brands in their stores but also their online inventory on thei
158、r own platforms,such as their e-commerce website,where brands can display banner ads,emails or even shopping mobile applications to deliver personalized promotions to their customers.2.Data monetization:retailers are monetizing existing consumer data across CPG brands to support their customer centr
159、icity.1P data shared by retailers originates from their loyalty program.The cardholder data they share can be socio-demographic(e.g.the age of their consumers),transactional(e.g.what did they buy),behavioral(e.g.what did they look at),loyalty data(e.g.did they buy again),etc.This data is shared“as-a
160、-service”in a data clean room where brands can access the retailers data in a secure environment to carry out specific use cases defined by the two partners.Carrefour,for example,has created a consumer intelligence service called Carrefour Links,based on the LiveRamp clean room,where partners can ac
161、cess their cardholder data.This is a self-service platform that allows users to perform basic activities such as reconciling retailer and brand databases on individual customers to build a more complete view of the consumer and thus improve their experience.It also provides analytics and measurement
162、 capabilities that Carrefour can bill to its partners.Access to this data can unlock three types of use cases for brands:Marketing:the data shared by retailers allows brands to gain insights about their consumers,activate them with media,or measure marketing performance through transactional data.Fo
163、r example,an ice cream brand partnered with a retailer to build advanced audiences for a digital marketing campaign.Using the retailers transactional data,the brand was able to build and activate two audiences:the brands current ice cream buyers and ice cream buyers from competitor brands.As a resul
164、t,the brand was able to increase the uplift of its campaigns by targeting the two relevant audiences with adapted messaging.Trade:the data shared by retailers allows brands to perform revenue growth management use cases by better optimizing promotions or assortment It also unlocks store optimization
165、 use cases through enhanced in-store experience or sales force optimization.For example,one brand worked with a retailer to analyze the short-and long-term impact of promotions on incremental margin.This enabled them to identify certain types of promotions that were margin-destroying for both the br
166、and and the retailer,as opposed to those that generated a positive long-term business impact.Operations:the data shared by retailers allows brands to optimize 26their supply chain through demand forecasting and demand management use cases.It can also fuel sustainability as well as production and inn
167、ovation use cases.Brands usually start implementing marketing use cases to deliver short-term value with a simple set-up,while long-term partnerships can then address very valuable use cases for Trade and operations that benefit both partners.Across all three categories,data monetization unlocks bet
168、ter measurement capabilities:optimized media performances,customer lifetime value calculation,dynamic budget allocation and global ROI optimization.3.Service monetization:along with inventory and data monetization,retailers can unlock additional revenue by providing different levels of services to b
169、rands.In the most developed additional service offering,retailers can propose managed services to brands,based on shared SLA and KPIs.In a long-term partnership approach,retailers optimize their revenue potential through best-in-class services for brands.Amazon offers its key clients advanced servic
170、es such as category management studies and dashboards or MMM(Media Mix Modeling)studies to help them improve their strategy in relation to this retailer.Des cas dusages valeur,au del du retail mediaData monetization:maturity in the Retail Media marketArtefact has benchmarked over 20 retailers in the
171、 US Retail Media market and analyzed their maturity with respect to Retail Media based on their value proposition and the comprehensiveness of their capabilities.We found that most retailers in the US have launched a Retail Media offering but are still in a nascent stage,mainly offering inventory to
172、 brands,while more mature retailers are focusing on data monetization or even service monetization for the best-in-class retailers.Data monetization offerings require setting up technical capabilities,such as a technical infrastructure to collect,store and process the first-party data to be shared,a
173、s well as a clean room to share data with brands,or even partnerships with DSPs to enable brands to directly activate audiences created in clean rooms.But the business opportunity is worth it:selling data as-a-service delivers margins often in excess of 80%,compared to inventory monetization where m
174、argins typically only reach 40%,as inventory assets are limited and therefore less scalable.DATA FOR RETAIL27CARREFOUR LINKSWith Artefact are helping brands to improve media targeting thanks to Carrefours dataCASE STUDYRetail ecosystem is currently going through a revolutionThe old rules of retail h
175、ave been deeply impacted by two new factors:digital assets(digital media)and shopper data,that helps leverage knowledge of customer behavior and granular activation of digital assets.Retail media mixes digital assets and shopper data.For many years,retail media has been helping brands and manufactur
176、ers become more efficient in four different ways:customer insights:a deeper understanding of your customers thanks to data from identified members of customer loyalty programs precise targeting:moving from mass marketing towards 1:1 marketing activation:omnichannel activations that can be synced thr
177、ough several touchpoints enhanced by shopper data measuring:measure of the return on ad spend and impact on salesCHALLENGES28Carrefours vision on their Retail Media systemCarrefour provides manufacturers with cutting-edge solutions to conduct retail media campaigns and to develop their sales and dis
178、play presence both in store and on their ecommerce website.According to Nicolas Trol,Carrefour Media has the unique opportunity to be able to leverage assets from the whole entire division,such as 5000 physical stores,a digital platform that reaches 10 millions customers every month,and the key elem
179、ent in their omnichannel offering,their unique data.In 2021,Carrefour took the decision to develop a new full stack platform on carrefour.fr,with Criteo as partner,to offer advertisers new activation solutions and better measuring and targeting.Artefact as a trusted partner to maximise the value of
180、Carrefours dataThomas Faure reviews the four pillars of the partnership between Carrefour Links and Artefact:Business consultancy:Manufacturers want to understand where growth comes from.“Carrefours data helps brands identify business opportunities for them to accelerate”Profiling and customer analy
181、sis:“Who is my customer?”One of the main challenges for manufacturers is to define who their customer is.Carrefours data helps brands represent the personas of their brand buyers Strategy,operation and activation:helping brands create media activation strategies(budget,KPIs)Advanced data use cases:p
182、rojects that take into account the whole retail value proposition“When we talk about data at Carrefour,we are talking about data from our customer loyalty program,which represents 14 million cardholders,7 million customers registered online and offline,and 5 million digital-only customers“Nicolas Tr
183、ol,Chief Revenue Officer-Carrefour Links“Retail media is an ecosystem that needs a certain number of mandatories on the retailers and manufacturers sides“Thomas Faure,Senior Consulting Manager E-Retail Lead-ArtefactRESULTSSOLUTIONDATA FOR RETAIL2930How Artefact helped boost Mattels online sales on C
184、discount(RelevanC Advertising)retail media platform?Increase presence and visibility for the brands of the Mattel Group on e-retailers media ecosystems by leveraging retail media.E-retail is a strategic distribution channel for the Mattel Group as 32%of transactions in the gaming category were condu
185、cted online during the 1st semester of 2020.Retail media is used in an objective of performance enabled by leveraging shopper data.Retail media is especially helpful for the conversion part of the sales funnel.Retail media lives in different types of environments such as:in-platform,products that ma
186、ximise the presence of products directly on the eCommerce website such as sponsored brands,sponsored products,merchandising banners off-platform,solutions that generate additional traffic redirected to the product pages on the eCommerce website such as search engine ads,social ads,display&video adsC
187、ASE STUDYCHALLENGESDATA FOR RETAIL31High touch handling of Mattels retail media investments in partnership with relevanC Advertising.Ad campaigns six times more effective!Mattel achieved extremely positive results through retail media activations in 2020:600%increase on return on ad spend for media
188、investments using retail media data(compared to traditional campaigns)from 4 up to 10 euros earned when spending one euro on display and search ad campaigns“relevanC Advertising has the most qualitative data that you can find on the market.The daily optimization by Artefact really moved the needle,”
189、mentions Maiana Darmendrail Digital Manager&E-Retail Manager,Mattel FranceBrands are able to find success when mixing business metrics with media expertise,i.e.sync retail signals(stock level,promotion level)with media KPIs to better optimize advertising campaigns.In 2020,Cdiscount cemented its posi
190、tion as the number one French e-retailer,generating 25 million unique visitors and 10 million customers(adding a million new buyers during the year).Cdiscounts vision of its retail media ecosystem covers the entire sales funnel and encompasses different steps(brand awareness,consideration,traffic,ac
191、quisition,and insights)In February 2020,relevanC introduced a self-serve platform,relevanC Advertising Platform,built entirely in-house,that offers retail media solutions through search and display on Cdiscount.“relevanC Advertising is a best-in-class tool to operate and manage retail media campaign
192、s,”says Maana Darmendrail Digital Manager&E-Retail Manager,Mattel France“CDiscount possesses very mature retail media solutions in search,display and video,”states Thomas Faure E-Retail Lead,ArtefactArtefacts made the decision to perform an all-year round online campaign aimed towards buyer intent.A
193、rtefact defined custom audiences out of precise shopper data items(purchase history,buyer intent,search history,browsing history,income level.).The key to increase performance of media investments was to conduct continuous optimizations on both retail and media KPIs.“We have the intimate conviction
194、that continuous optimization brings performance improvements,”explains Cdric Chamoux Directeur Retail Media,relevanC Advertising Optimizations were made on several factors:channels(onsite,offsite,search,display),segments(audience,keywords),creatives,formats.Optimizations were based on several metric
195、s:media metrics,business metrics,retail metrics(stock level,promotions,organic positioning)Optimizations happened on different solutions:search advertising,daily improvements on keyword selection,display advertising,bi-weekly improvements on impact measurementRESULTSSOLUTIONIn addition to leveraging
196、 the e-retailer ecosystem,Artefact kept looking for ways to innovate and test new solutions to build more expertise and scale projects.In that way,Artefact conducted off-platform campaigns through Shopping for Partners solutions(Google for Retail)that were activated to support milestones,such as pro
197、duct launches.How to reduce food waste in the Bakery-Pastry department?CHALLENGESCARREFOUR GROUP How to reduce food waste in the bakery-pastry department?CASE STUDYPredicting demand as accurately as possible is one of the foundations of the retail business.However,this challenge is becoming more com
198、plex as consumer habits evolve.From now on,it is necessary to take into account the combined use of different sales channels(from e-commerce,to local stores,to supermarkets)and the increasing demand for social and environmental responsibility.At the top of the list of CSR issues linked to mass retai
199、ling:food waste.In France,bread is the third most wasted product after fruits and vegetables.Indeed,bread,pastries and cakes are fresh products that have a very limited shelf life.But,they are also items that can generate a lot of frustration among customers when they are out of stock at the end of
200、the day.This is why Carrefour Group,together with Artefact,sought to use machine learning and data science to optimize the prediction of sales in the bakery-pastry department of its supermarkets.This project meets a double objective:produce enough to meet the demand,while reducing“scrappage”,i.e.the
201、 volume of unsold products.This use cases success relies,in particular,on the teams involving both the“jobs”concerned(department managers)and the various technical profiles of Artefact and Carrefour Group.32A new algorithm to prevent“scrappage”in the bakery-pastry departmentIn the retail world,every
202、 day is different.Sales are highly dependent on the context:holidays,weather,current promotions,merchandising highlights,etc.To take into account all the variables that have implications on demand,we need to be able to analyze the petabytes of data generated by the billions of transactions carried o
203、ut every year at Carrefour,to which we need to add external data that influences consumption.These calculations are only possible with Artificial Intelligence.The Carrefour Group and Artefact teams thus started with data from the sales receipts generated by more than 200 supermarkets in France.Every
204、 day,this data is collected,cleaned and enriched with external sources such as calendar data,for example to build a sales history over several years.This represents thousands of configurations for a single day,depending on the assortment,product prices,promotions,etc.This data is used to train super
205、vised machine learning models,built on the basis of decision trees,which determine the relationships between the target variable(future sales)for each product and the explanatory variables(promotion,cannibalization,etc.).Close collaboration with the“field”teamsFrom an organizational point of view,th
206、e project was led by multidisciplinary teams.Two teams on the Artefact and Carrefour sides combined technical and business profiles.The operational skills of Carrefours retail professionals played a crucial role.They were able to explain their business,their needs,and bring their vision,in order to
207、guarantee the success and adoption of the solution“in real life”.SOLUTIONFor example,the algorithms recommendations were first tested in pilot stores,only on pastries,to get feedback from the field teams.Their comments were used to improve the models,before the solution was deployed across all baker
208、y and pastry products in the supermarkets.Easy-to-access and interoperable toolsAll of Carrefours data-driven use cases are fed by a centralized platform in the cloud,which makes the data accessible,formatted and documented.The results of the processing are then fed back into Carrefours information
209、systems,shared by a wide range of employees.Several teams,made up of data scientists,data engineers,but also data translators(pivotal profiles,acting as a link between business and data),are likely to consult,transform and process them for specific uses.DATA FOR RETAIL33RESULTSConcrete results on th
210、e number of discarded productsand on the brand imageA few months after the implementation of this new prediction model,the results are very positive.In fact,over the last five months of 2021,approximately 100 tons of pastries were saved.At the same time,sales have increased due to fewer shortages at
211、 the end of the day.Finally,the Net Promoter Score,a performance indicator closely monitored by Carrefour,has evolved very positively.Carrefour Group is multiplying AI use cases to improve customer experienceFor the group,future experiments follow the same model:responding to business needs,working
212、jointly with operational teams,to feed the customer experience.This acceleration of Carrefours digital transformation was made possible by the creation of complete and expert data teams within the company,and the deployment of data platforms in all countries where the group operates.The volume and w
213、ealth of data collected by Carrefour provides a unique opportunity to explore the major challenges facing the retail sector:omnichannel,e-commerce,anticipation of consumer habits,etc.Carrefour recently unveiled other examples of how data can be used to improve the customer experience:five-minute sho
214、pping on Carrefour.fr,the implementation of personalized assortments for local stores,and the personalization of promotions.In fact,over the last five months of 2021,approximately 100 tons of pastries were saved.At the same time,sales have increased due to fewer shortages at the end of the day.34Dem
215、and forecasting:Using Machine Learning to predict sales in retail.All industries aim to manufacture just the right amount of products at the right time.But,for retailers,this issue is even more important as they also need to manage their stocks efficiently.Too many items in stock is bad.Too few item
216、s in stock is also bad.And to predict sales as close to possible,retailers used to only rely on the previous years past sales.This method is useful only to a certain point and suffers many biases.Thankfully,Machine Learning has now evolved to be able to provide very accurate predictive models using
217、different signals based on how they influence purchases.Retailers main challenge is to go beyond past sales to predict future sales accuratelyManaging orders and inventory is the one of the strongest competitive advantages that can help retailers achieve success.And it is a real challenge to master
218、as it involves processing a huge number of SKUs-some of them that are even perishable-ordered daily.We estimate that bad inventory management,whether its out of order items or excess stock,cost US retailers close to two billion dollars per year.For decades,retailers have been relying on the analysis
219、 of their past sales for their Enterprise Resource Planning(ERP)that helps them reduce their investment and exploitation costs.However,these methods are heavily biased and are not that useful when trying to predict accurate sales.Dozens of signals to take into account when assessing stock levelsThe
220、reason why predicting sales appears to be so complex and difficult is because,in a given period,many factors can affect purchase:weather,purchasing trends,regulation,product launches,global pandemic,buying behaviors.And the main issue with these types of predictions based on past recordings is that
221、they dont factor individual incidents,and they make monthly sales appear as if they were perfectly distributed when they were probably not.In fact,an out of order item might have caused a slowing of sales of a particular product or a particular category,but it wont show in the monthly reports.Even w
222、orse,bad numbers are viewed as a mark of buyers disinterest,when they could be the opposite as consumers over-purchase an item and cause it to sell out.It is also important to note that a missing product in store doesnt Jrme PetitManaging Partner Retail&eCommerceARTEFACTDATA FOR RETAIL35necessarily
223、mean that the product is out of stock.Big box retailers struggle to restock their shelves in real time so a product that becomes instantly popular might disappear from the shelves and thus perform worse than expected,when in fact,it is available in inventory.Retailers are in need of technology that
224、can help them step into a new paradigm that could seamlessly align offer and demand.Using Machine Learning to help employees in storesMachine Learning is the solution to these challenges.Predictive models are now able to forecast sales up to four weeks by using a number of signals that can affect sa
225、les,such as the season,the current trends or the price levels.The models are based on three indicators,the Day,the Product,and the Store.It is important that they dont rely on a single indicator to be the most accurate possible.For the sake of the argument,lets imagine that you analyse the season si
226、gnal.Your data is going to be biased because predictions based on dates are not 100%helpful because a certain date can fall on a weekday in a given year and fall on a weekend in the following year,and sales vary enormously between the weekday and the weekend.Also,this date could end up on a holiday,
227、or a calendar event(such as Christmas,Thanksgiving,Easter)or a sports event.All factors that might increase the consumption of some products.As another example,lets have a look at the price levels signal.Promotions at store level can severely affect the sales of a product from a given category or ev
228、en raise the attractiveness of a store as a whole.Retailers need to be able to predict the impact of their promotion strategy.Thats why it is important to take into account many different signals and indicators to accurately forecast sales through Machine Learning and advanced Artificial Intelligenc
229、e models.Concrete use cases of Machine Learning on inventory managementThe technology is there.Now for retailers to use it effectively and make accurate predictions,they need to be able to collect and analyse a huge amount of data.The problem lies in the fact that they have different sources of data
230、 and it can become complex to try to process multiple Excel and PDF files that contain previous reports and media plans.Retailers need to adopt big data tools that can process this information into a clean and lisible digest,that is necessary to be able to create predictive models that can prevent i
231、nventory issues.The vast majority of retailers are sitting on years of old sales data.However,this data can turn out to be inaccurate because of the effect of past promotions or events(heat wave,sports event,major local event).To get rid of this bias,predictive models are able to combine past sales
232、numbers with those of similar stores.Artefact has enjoyed success with this approach and conducted a successful experimentation in China on the 020 platform that raised by 20%the level of accuracy in sales predictions.The other big challenge is to be able to prevent items from being unavailable in s
233、helves when they are available in inventory.It is almost impossible for store managers to ask their employees to monitor shelves in real time and restock them immediately.Technology solutions using surveillance cameras and weight sensors are a huge investment.However,available information such as re
234、al-time sales,the characteristics of a SKU and the organization of a particular shelf can be useful when predicting out of order.Models are able to analyze the usual flow of sales of an item,meaning the time between two sales of the same product in a given store.In case of a statistical anomaly,a hu
235、man is notified for an intervention in the store to analyze the situation and solve it.There are many ways in which Machine Learning can help traditional retailers.Predictive analysis is only one of them.Retailers can gain a lot from relying on advanced technology to increase their store revenue by
236、better managing their inventory.But processing vast amounts of data can also help them on the supply chain front,or on the merchandising side.Well-devised tools can become a real asset to managers by taking care of complex and time-consuming tasks and provide them with accurate reports in a short am
237、ount of time and letting them shift their priorities to improve customer experience.36DATA FOR RETAIL37Sales forecasting in retail:what we learned from the M5 competition.In this article,Data Scientist Maxime Lutel sums up his learnings from the M5 sales forecasting competition,which consisted in pr
238、edicting future sales in several Walmart stores.He will walk you through our solution and discuss what machine learning model worked the best for this task.Maxime LutelData ScientistARTEFACTUsing machine learning to solve retailers business challengesAccurate sales forecasting is critical for retail
239、 companies to produce the required quantity at the right time.But even if avoiding waste and shortage is one of their main concerns,retailers still have a lot of room for improvement.At least,thats what people working at Walmart think,as they launched an open data science challenge in March 2020 the
240、 M5 competition to see how they could enhance their forecasting models.The competition aimed at predicting future sales at the product level,based on historical data.More than 5000 teams of data lovers and forecasting experts have discussed for months about the methods,features and models that would
241、 work best to address this well-known machine learning problem.These debates highlighted some recurring issues encountered in almost all forecasting projects.And even more importantly,they brought out a wide variety of approaches to tackle them.This article aspires to summarize some key insights tha
242、t emerged from the challenge.At Artefact,we believe in learning by doing,so we decided to take a shot and code our own solution to illustrate it.Now lets go through the whole forecasting pipeline and stop along the way to understand what worked and what failed.Problem statement:Hierarchical times se
243、ries forecastingThe dataset contains 5-year historical sales,from 2011 to 2016,for various products and stores.Some additional information is provided,such as sell prices and calendar events.Data is hierarchically organized:stores are divided into 3 states,and products are grouped by categories and
244、sub-categories.(cf Image 1)Our task is to predict sales for all products in each store,on the days right after the available dataset.It means that 30 490 forecasts have to be made for each day in the prediction horizonThis hierarchy will guide our modeling choices,because interactions within product
245、 categories or stores contain very useful information for prediction purposes.Indeed,items in the same stores and categories might have similar sales evolution,or on the contrary they could cannibalize each other.Therefore,we are going to describe each sample by features that capture these interacti
246、ons,Image 138and prioritize machine learning based approaches over traditional forecasting ones,to consider this information when training the model.Two main challenges:intermittent values and an extended prediction horizonAt this stage,you might think that it is a really common forecasting problem.
247、Youre right and thats why it is interesting:it can relate to a wide range of other projects,even if each industry has its own characteristics.However,this challenge has 2 important specificities that will make the task more difficult than expected.The first one is that the time series we are working
248、 with have a lot of intermittent values,i.e.long periods of consecutive days with no sales,as illustrated on the plot below.This could be due to stock-outs or limited shelves area in stores.In any case,this complicates the task,since the error will skyrocket if sales are predicted at a regular level
249、 while the product is out of shelves.(cf image 2)The second one comes from the task itself,and more precisely from the size of the prediction horizon.Competitors are required to generate forecasts not only for the next week,but for a 4-week period.Would you rather rely on the weather forecast for th
250、e next day or for 1 month from now?The same goes for sales forecasting:an extended prediction horizon makes the problem more complex as uncertainty increases with time.Feature engineering Modeling sales driving factorsNow that we have understood the task at hand,we can start to compute features mode
251、ling all phenomenons that might affect sales evolution.The objective here is to describe each triplet Day x Product x Store by a set of indicators that capture the effects of factors such as seasonality,trends or pricing.SEASONALITYRather than using the sales date directly as a predictor,it is usual
252、ly more relevant to decompose it into several features to characterize seasonality:year,month,week number,day of the week The latter is particularly insightful because the problem has a strong weekly periodicity:sales volumes are bigger on the weekends,when people spend more time in supermarkets.Cal
253、endar events such as holidays or NBA finals also have a strong seasonal impact.One feature has been created for each event,with the following values:Negative values for the 15 days before the event(-15 to-1)0 on the D-day Positive values for the 15 days following the event(1 to 15)No value on period
254、s more than 15 days away from the eventThe idea is to model the seasonal impact not only on the D-day,but also before and after.For example,a product that will be offered a lot as a Christmas present will experience a sales peak on the days before and a drop right after.TRENDSRecent trends also prov
255、ide useful information on future sales and are modeled thanks to lag features.A lag is the value of the target variable shifted by a certain period.For any specific item in a given store,the 1-week lag value would be the sales made one week ago for this particular item and store.Different shift valu
256、es can be considered,and the average of several lags is computed as well,to get more robust predictors.Lags can also be calculated on aggregated sales to capture more global trends,for example at the store level or at the product category level.PRICINGA products price can change from one store to an
257、other,and even from one week to another within the same store.These variations strongly influence sales and should therefore be described by some features.Rather than absolute prices,relative price differences between relevant products are more likely to explain sales evolutions.Thats why the follow
258、ing predictors have been computed:Relative difference between the current price of an item and its historical average price,to highlight promotional offers impact.Price relative difference with the same item sold in other stores,to understand whether or not the store has an attractive price.Price re
259、lative difference with Image 2DATA FOR RETAIL39other items sold in the same store and same product category,to capture some cannibalization effects.Categorical variables encodingCategorical variables such as the state,the store,the product name or its category also hold a significant predictive powe
260、r.This information has to be encoded into features to help the model leveraging the dataset hierarchy.One-hot encoding is not an option here because some of these categorical variables have a very high cardinality(3049 distinct products).Instead,we have used an ordered target encoding,which means th
261、at each observation is encoded by the average sales of past observations having the same categorical value.The dataset is ordered by time for this task to avoid data leakage.All categorical variables and some of their combinations have been encoded with this method.This results in very informative f
262、eatures,the best one being the encoding of product and store combination.If you wish to experiment other encoders,you can find a wide range of methods here.Tweedie loss to handle intermittent valuesDifferent possible strategies can be used to deal with the intermittent values issue.Some participants
263、 decided to create 2 separate models:one to predict whether or not the product will be available on a specific day,and a second one to forecast sales.Like many others,we have chosen another option,which is to rely on an objective function adapted to the problem:the tweedie loss.Without going into th
264、e mathematical details,lets try to understand why this loss function is appropriate for our problem,by comparing sales distribution in the training data and the tweedie distribution(cf image 3).They look quite similar and both have values concentrated around 0.Setting the tweedie loss as an objectiv
265、e function will basically force the model to maximize the likelihood of that distribution and thus predict the right amount of 0s.Besides,this loss function comes with a parameter whose values are ranging from 1 to 2 that can be tuned to fit the distribution of the problem at hand(cf image 4)Based o
266、n our dataset distribution,we can expect the optimal value to be between 1 and 1.5,but to be more precise we will tune that parameter later with cross-validation.This objective function is also available for other gradient boosting models such as XGBoost or CatBoost,so its definitely worth trying if
267、 youre dealing with intermittent values.How to forecast 28 days in advance?:Making the most out of lag featuresAs explained above,lag features are sales shifted by a given period of time.Thus,their values depend on where you stand in the forecasting horizon.The sales made on a particular day D can b
268、e considered as a 1-day lag if youre predicting one day ahead,or as a 28-day lag if youre predicting 28 days ahead.The following diagram illustrates this point(cf image 5)This concept is important to understand what features will be available at prediction time.Here,we are on day D and we would like
269、 to forecast sales for the next 28 days.If we want to use the same model and thus the same features to make predictions for the whole forecasting horizon,we can only use lags that are available to predict all days between D+1 and D+28.This means that if we use the 1-day lag feature to train the mode
270、l,that variable will also have to be filled for predictions at D+2,D+3,and D+28,whereas it refers to dates in the future.Still,lags are probably the features with the biggest predictive power,so Image 3Image 440its important to find a way to make the most out of this information.We have considered 3
271、 options to get around this problem,lets see how they performed.OPTION 1:ONE MODEL FOR ALL WEEKSThe first option is the most obvious one.It consists in using the same model to make predictions for all weeks in the forecasting horizon.As we just explained,it comes with a huge constraint:only features
272、 available for predicting at D+28 can be used.Therefore,we have to get rid of all the information given by the 27 most recent lags.It is a shame as the most recent lags are also the most informative ones,so we have considered another option.OPTION 2:WEEKLY MODELSThis alternative consists in training
273、 a different LightGBM model for each week.On the diagram above,every model is learning from the most recent possible lags with respect to the constraint imposed by its prediction horizon.Following the same logic as the previous option,it means that each model can leverage all lags except those that
274、are newer than the farthest day to predict.This method allows us to better capitalize on lag information for the first 3 weeks and thus improved our MORE PRECISELY:Model 1 makes forecasts for days 17,relying on all lags except the 6 most recent ones.Model 2 makes forecasts for days 814,relying on al
275、l lags except the 13 most recent ones.Image 5 Model 3 makes forecasts for days 1521,relying on all lags except the 20 most recent ones.Model 4 makes forecasts for days 2228,relying on all lags except the 27 most recent ones just like in option 1.DATA FOR RETAIL41solutions forecast accuracy.It was wo
276、rth it because it was a Kaggle competition,but for an industrialized project,questions of complexity,maintenance and interpretability should also come into consideration.Indeed,this option could be computationally expensive and if we are aiming at a rollout on a whole country scale,it would require
277、maintaining hundreds of models in live.In that case,it would be necessary to evaluate if the performance increment is large enough to justify this more complex implementation.OPTION 3:RECURSIVE MODELINGThe last option also uses weekly models,but this time in a recursive way.Recursive modeling means
278、that predictions generated for a given week will be used as lag features for the following weeks.This happens sequentially:we first make forecasts for the first week by using all lags except the 6 most recent ones.Then,we predict week 2 by using our previous predictions as 1-week lags,instead of exc
279、luding more lags like in option 2.By repeating the same process,we always get recent lags available,even for weeks 3 and 4,which allows us to leverage this information to train the models.This method is worth testing,but keep in mind that it is quite unstable as errors spread from week to week.If th
280、e first week model makes important errors,these errors will be taken as the truth by the next model,which will then inevitably be poorly performing,and so on.Thats why we decided to stick with option 2,that seems to be more reliable.Ensuring model robustness with an appropriate cross-validation:Why
281、cross-validation is critical for time seriesIn any machine learning project,adopting an appropriate cross-validation strategy is critical to simulate correctly out-of-sample accuracy,select hyper-parameters thoroughly and avoid over-fitting.When it comes to forecasting,this has to be done carefully
282、because there is a temporal dependency between observations that must be preserved.In other words,we want to prevent the model from looking into the future when we train it.The validation period during which the model is tested also has a greater importance when dealing with time series.Model perfor
283、mance and the optimal set of hyper-parameters can vary a lot depending on the period over which the model is trained and tested.Therefore,our objective is to find which parameters are the most likely to maximize performance not over a random period,but over the period that we want to forecast,i.e.th
284、e next 4 weeks.Adapting the validation process to the problem at handTo achieve that goal,we have selected 5 validation sets that were relevant to the prediction period.The diagram below shows how they are distributed over time.For each cross-validation fold,the model is trained with various combina
285、tions of parameters on the training set and evaluated on the validation set using the root mean squared error.(cf Image 9)Folds 1,2 and 3 aim at identifying parameters that would have maximized performance over recent periods,basically over the last 3 months.The problem is that these 3 months might
286、have different specificities than the upcoming period that we are willing to forecast.For example,lets imagine that stores launched a huge promotional season over the last few months,and that it just stopped today.These promotions would probably impact the models behavior,but it would be risky to re
287、ly only on these recent periods to tune it because this is not representative of what is going to happen next.To mitigate this risk,we have also included folds 4 and 5,which correspond to the forecasting period respectively shifted by 1 and 2 years.These periods are likely to be similar because the
288、problem has a strong yearly seasonality,which is often true in retail.In case we had a different periodicity,we could choose any cross-validation strategy that has more business sense.In the end,we have selected the hyper-parameters 42combination with the lowest error over the 5 folds to train the f
289、inal model.ResultsThe different techniques mentioned above allowed us to reach a 0.59 weighted RMSSE the metric used on Kaggle which is equivalent to a weighted forecast accuracy of 82.8%.The chart below sums up the incremental performance generated by each step.These figures are indicative:the Imag
290、e 9incremental accuracy also depends on the order in which each step is implemented.Key takeawaysWe have learned a lot from this challenge thanks to participants shared insights and we hope it gave you food for thoughts as well.Here are our key takeaways:Work on a small but representative subset of
291、data to iterate quickly.Be super careful about data leakage in the feature engineering process:make sure that all the features you compute will be available at prediction time.Select a model architecture that allows you to leverage lags as much as possible,but also keep in mind complexity considerat
292、ions if youre willing to go to production.Set-up a cross-validation strategy adapted to your business problem to evaluate correctly your experiments performance.WE OFFER END-TO-END DATA&AI SERVICESTRAVEL&TOURISM FMCG SPORTS&ENTERTAINMENT BANKING&INSURANCE RETAIL LUXURY&COSMETICS HEALTHCARE eCOMMERCE TELECOMMUNICATIONS MANUFACTURING&UTILITIES REAL ESTATE PUBLIC&GOVERNMENTCONTACT HEADQUARTERS19,rue Richer75009 Paris France