Every day during this 13th edition, journalist Geneviève Petit will be offering an immersive summary of everything you don't want to miss at the event. Keynotes, parties, networking... One to One Retail E-commerce as if you were there!

On Wednesday morning, the day begins with new buying behaviours:

  • Autone workshop

 

For the first time, a study into purchasing behaviour carried out by Opinion Way for Autone shows that unavailability is at the top of the list of irritants for consumers, along with lack of price information. 39% of the French people questioned were confronted with unavailability in the fashion sector, 26% in the household goods sector and 11% in the luxury goods sector. 64% of consumers are not informed of the unavailability of products in shop. And 25% of retailers inform them that the product is available in another shop. Consumers can wait an average of 19 days for a luxury product to become available, compared with 12 days in the fashion sector.

This is a serious matter: the frequent unavailability of products makes consumers unfaithful. Vanessa Gignoux, Director of Commerce and Omnichannel at Gemo has 400 physical shops and an e-commerce channel that accounts for 10% of the total. 50% of online customers come to the shop. The retailer has adopted Proximis, to avoid disappointing customers, by creating a dedicated stock and free in-store delivery, with click and collect in an hour. The retailer has also developed its free product reservation service, after the customer has chosen the nearest shop: "We have one reservation for every two sales. The item waits in the cabin". A true luxury experience... Gémo has a very wide range of products. "It's difficult to keep a deep stock of all the products. Shops that don't have any stock can use the online stock. The latest innovation is that in around fifteen shops, a QR code provides consumers with product information. Finally, the brand will be tackling CSR: "Having 6 parcels for every 10 products ordered is not acceptable, either to the customer or to the planet".

Arnaud Bazin, VP of Roberto Cavalli, has 15 shops worldwide and e-commerce accounts for 11% of business. "Today we have complete visibility of our stocks. GB we pay too many customs fees to bring in goods. The company has all the problems of a start-up. We're looking for bestsellers. Autone helps us to quantify this. A change in mentality has been necessary: "It's a cultural change. Before, it was done by telephone, between shops, exchanging products, but there was no follow-up, we didn't know if the sale had been made".

Yann Rivoallan, President of the women's ready-to-wear federation, illustrates the changes in the sector with a brand that both fascinates and pushes back its sector: Shein. Brands need to work on desirability, cost and technology. Technology makes it possible to provide service at every level of the chain. On 14 March, the French Parliament will vote on a bill targeting fast fashion and Shein in particular. "AI is everywhere in this company. It allows us to create 8,000 products a day, to have production up and running in three days. Two-week delivery should be just a passing thing, if we remember Amazon, which ended up chartering its own planes".

Adil Bouhdadi, Autone's CEO, created this stock optimisation platform three years ago, a perfect time to market: "Optimising stock management was a nice to have 5 years ago. Now we realise that it's about optimising cash. The CFOs of retailers have become the new CEOs". The platform makes it possible to reduce stock shortages (on average, 15% of products are out of stock), avoid over-stocking, optimise transport costs and better forecast the replenishment of best-sellers. Autone also makes it possible to assess the number of customers who buy only once, and to understand the customer journey.

The system can inject external data, such as the weather, inflation, traffic data or social movements. Ultimately, the system will enable augmented sales staff to anticipate when customers are coming, what they need and what substitute products to offer them if necessary. Adil Bouhdadi is not going to stop at stocks: "We want to tackle all the friction between the production of a product and the customer. We've started with stocks, and we're going to continue with the other archaic bricks, such as warehouse logistics".

 

  • Upgrading skills, the key challenge for e-tailers in 2024

 

In this workshop led by Flore Fauconnier, who together with Patrick Robin launched Eduteka, a training programme for retail professionals, Thierry Lernon, But's Director of Marketing, Customer Experience, Data and E-Commerce, dispelled the legend: "The practice of retail players is very far removed from Shein, which puts data everywhere.

It's not easy for traditional retailers to exchange objective data in Excel and get away from the urban legend. Google Analytics is very granular, so we don't have a good historical comparison. From there, it's difficult to discuss things objectively. Tools like Tablo or Google Analytics give access to extreme granularity and are difficult to get people to talk about". At the same time, e-commerce professionals, who are becoming increasingly specialised, need to be aware of other professions: "Anyone doing UX on the payment tunnel needs to understand the user, but also the way in which banks pass the buck in the event of fraud, or the position of credit in the way they talk to customers".

As the Visa keynote made clear, the ability to work together is very important. "E-commerce has to understand shops. We need to move people from one office to another". For Yann Rivoallan, artificial intelligence can bridge the gap between e-commerce and the shop.

The President of the French Federation of Women's Ready-to-Wear draws a comparison between the LLM and the artistic director of a luxury brand. "An artist has a strong sensibility; he or she accumulates data and information all day long and puts it together in the form of a moodboard for the business. With an LLM, you scrape the data and come up with something that makes sense. But AI is scary and creates two worlds: those who want to go there and the others, who represent 30% of people: "there are still a lot of barriers. It's the same thing with AI gen. It's a third narcissistic flaw for humanity, after Galileo and Freud. The big fear is that AI will replace us". And indeed, the jobs that will be replaced are those earning €80,000 a year and less. "Young people have settled the issue: 100% of students use AI. How do you acclimatise your teams to AI?

Thierry Lernon has trained all the teams on two occasions but also the Executive Board, and offered a GPT 4 chat licence to its teams. In the end, around fifteen of the 100 people in his department use it. In terms of usage, 100% of the product sheets are written by an AI set up with a partner. Well, almost 100%: "But when it comes to bedding, the AI says anything about mattresses and bases". Another use case is the completion and verification of the 400,000 product data sheets in the market place. How do we tackle the ecological transition? For Yann Rivoallan, we need to be open-minded and understand that the single price is an accident of history, invented by Emile Boucicaut at Le Bon Marché. On Vinted or BonCoin, the same product can have a different price. The same logic applies to Shein. But was much more discreet than its CSR actions: "We manufacture 80% in Europe and 40% in France, but we didn't make it known". The brand was also the first to bring together its Chinese suppliers to discuss the subject of responsible production: "Most of But's CO2 production comes from the manufacture of products, as we learned from our extensive assessment.

We have reduced energy consumption in shop by 40%, but that's only a small part of it. We'll solve the challenge once we've added a production engineer to our buyers". The mindset is fundamental to survival: "successful brands have a permanent test & learn approach, like Bash, run by Pierre Arnaud Grenade, who is passionate about innovation and the evolution of society". We also need to rethink our rhythms: "Everything changes in the space of a year. Start-ups evaluate their teams every 6 months, or even every 3 months. Let's do the same. It's much more fun to do it that way", concludes Thierry Lernon.

Yann Rivoallan also suggests taking inspiration from the IT run for all projects: "You can launch a project over 6 weeks, take a break for a week and then start again, with Notion for monitoring. In this way, we can make pragmatic progress as a team". Finally, you also need to know how to take your time, "it's important to be able to delay a release by 3 months if you realise along the way that you've missed something. But you also need to step back and take your time, even when you're in the red".

 

 

  • Round Table: Is ecology a rich man's problem?

 

Philippe Berlan, CEO of La Redoute, discussed the new European regulation, the CSRD, which provides an accounting framework for CSR. "It's a question of identifying in the strategy what makes the impact. And distinguishing between what is important for the company and what has an impact on society. The company chooses which indicators to publish. And this publication is certified by a third party". Coercion will come from the comparison of publications, by NGOs or customers. La Redoute is taking action on product manufacturing: "Two-thirds of our turnover is made up of products that we design. We will choose between switching from cotton to linen, or from geographical zone A to zone B, depending on the supply of fossil or renewable energy". The One Ragtime investment fund is committed to having a quota of impact companies. Today, 10% of investments are geared towards impact. "We've invested in Pick me, which lets you leave the last metre of delivery to your neighbour, reconciling the end of the month with the end of the world. We've also invested in Stokelp, which sells surplus agricultural produce to manufacturers. And Captain Cause, which helps companies target their marketing budgets at major causes. FDJ has created an organic diversity lottery.... Or Hive, which uses the computing power and storage capacity of computers to benefit the community.

For Philippe Berlan: "The ecological transition is a bit like the digital transformation a few years ago. Only more profound. It raises the question of how to transform ourselves in depth". Finally, Stéphanie Lhospital calls on the public not to forget water. This is the cause of Mayan Labs, which measures water footprints. And in everyday life, if we care about water, we eat locally and not industrial food....

 

  • Brevo Keynote: Ensure a positive ROI from your AI investments

 

Mickaël Arias, Chief Engagement Officer at Brevo. This is Brevo's first One to one Monaco. Brevo was born 11 years ago with a lot of small customers. Today we have 500,000. We are rapidly coming up against AI and data for anti-fraud purposes. Mickael Arial is in charge of the corporate business, which has 2,500 customers. But it's the former CTO who gives his vision of AI, which is far from just Chat GPT.

1 - AI is expensive. MIT has carried out a study to model the costs of AI: technical cost, cost of dv, training, re-training and operation. Also missing is the human cost of training teams. When AI replaces a human being, in 77% of cases it costs more than salary for an equivalent result. The study focuses on jobs (automated inventory, automated quality control). Why are we talking so much about it? In French, there were 30,000 articles published per month. 200 in 2013. Gen AI is based on deep learning, which appeared 20 years ago and is based on neural networks invented 40 years ago. Yan Le Cun, who is in charge of the AI branch at Facebook, says that Chat GPT has contributed nothing. It had already been available on platforms for 5 years. But Open AI has made open AI accessible to everyone. It's based on 3 factors that exploded in 2023:

  1. Access to data
  2. Scientific advances in GPT
  3. Computing power thanks to edge computing and the creation of new chips

 

This computing power costs Chat GPT $700K per day to run. Cloud providers are very expensive, so are we dependent on them? You have to go there to keep up to date, things are going to change very quickly, computing power is going to cost less. The impact will be colossal, so don't hesitate. Gen AI is making a lot of noise, with use cases, such as at Cdiscount, which has shown some concrete examples. But that's just one part of AI. Deep learning, machine learning, artificial intelligence, GPT chat: these are probability stats, an algo that predicts the next most likely word.

3 examples of what we can do:

  1. Optimising the times at which emails are sent
  2. Product recommendation based on collaborative filtering. We are trying to reproduce this reasoning with neural networks. 1/18 which is better than matrix calculation, but at a higher cost.
  3. Product information management, PIM. Determine if two products per week. POC on GPT chat. ChatGPT allows us to move very quickly. At Brevo, we've categorised tickets using AI, to allocate them to the agent best able to respond.

 

ChatGPT creates categories that don't exist, high cost and data protection. Training cost for each category added, the model must be re-trained.

Chat GPT Embedded Vector can represent text in mathematical form and perform vector operations. Chat GPT describes the entire image and can ask a very specific question. But when it comes to use cases, you often need something specialised. And there are hundreds of open source algo's to do this. Hugging Face benchmarks all these algos.Chat GPT allows you to do the POC but you can implement your version, independently of Gafam at a lower cost with a better result.

Good and bad use cases with Alltricks. "We wanted to include recommended products in the newsletters. But newsletters are well designed, so it's difficult to find additional conversion". The head of Alltricks came back with the idea of automating the creation of the NL, to relieve the CRM team. 2 no data, no AI.

Another trend is the need to manipulate data, whatever the technology. It needs to be trained on the customer knowledge base, as this is data that needs to be made available. You have to teach the algo what data interests me and what data doesn't (label the data).
Chihuahua or muffin?
Dog or mop?

Putting together the data set is the longest and most difficult part of the project. It's hard to find the data, because sometimes it doesn't exist because you don't create it. We have a lot of account creation; when we send the email we have to avoid spamming and phishing. At the beginning we had manual validation. When we had several million people testing. We validated but didn't de-validate. We had a number of issues about context. People who fall through the cracks, people who read Soc Gen or phishing, can't tell them apart. There's also filtering. When we gave out messages from X, we forgot to remove our answers".

The big trend in recent years has been the development of a wide range of tools to do this. Having a CDP makes it possible to work quickly on data sets, which improves time to market. To reduce the cost and environmental impact of generative AI, "it will probably take 5 years. In the meantime, we need to prepare the data and use algos. We go for it if it's at the heart of our business, which doesn't exist, and we implement it. We used to do POCs with ChatGPT-type technologies. You have to look at the budget and the environmental impact. There are too many companies that are not ready when it comes to their data.

 

  • Composable commerce in the age of AI: how can you transform your brand?

 

With Michael Scholz, VP Product and Consumer Commerce tools and Nicolas Pastorino Group Chief Product and Digital Officer Interflora

What is composable commerce?

Decouple the experience by adding social and mobile, it's more flexible when you have a decomposed experience. Decouple the front and back end, search, my personalisation, etc. For the last 5 years, there have only been two truths. Change is what doesn't change. The future is unpredictable. Nicolas Pastorino fell into composable commerce "by accident". "In 2021, we created a group and a platform for all our entities. We had to make sure there was no big bang. We had 8 different countries, 8 different platforms and 8 different cultures. The first layer had to prove its value, and composability was essential for rapid transformation. This gives us greater adaptability and enables us to make changes without breaking everything. An example: when you offer a bouquet, you want to have a beautiful card. We were able to measure the traction of the market, which was validated. So we deployed a quality printer and deployed. This shows how this tool is transforming the way we operate. Our product managers are now focused on the problems to be solved, not on a list of tasks.

“We went from “time to market” to “time to money”. Decisions are based on data. The solution generates savings that are reinvested. Composable commerce improves data quality. By nature if you plug fewer tools, the data is cleaner. As Jim Clark of Netscape said, “If we have data, let’s decide with it. If we don’t, let’s take my opinion.” You have to test, prove the value. And beware, if you feed GPT Chat confidential information, it will be public. We have a premium GPT Chat account. A use case? Gifts.com, personalized gifts, we had signals we hadn’t seen. Gifts for my father, who is 60 years old and a fan of fishing. We were not positioned on that. You can create product detailed product that would meet those demands. Advanced prompts for texts and images were created 2,000 to 15,000 references in 15 days and it worked. We didn’t have a data scientist. We had to test chat GPT to find the use case and scale it. Which keyword to discard rather than which keyword to choose”.

Finally, Interflora integrated in photoshop a prompt to create images, «it is very used by our designers». Michael Scholz, Commerce Tools, is looking for the best applications «Today, Gen AI is used to create content and texts, it should be in the CMS». Interflora has improved fraud management across Gen AI and optimized the way products are presented in PRP with AI.
AI will also revolutionize search: "Search pages will change a lot this year. How will we exist in this new context? Will SEO disappear? Where will SEA be?” With Marbling, you say what app you want to build, and it comes back with a multi-layer app. "At some point, it’s going to work. Code generation should probably be supported, coders will need to become prompt.

”A few recipes to get started? “You have to have the right to fail, and act step by step. It is not the privilege of the CTO, the executive must get involved because it produces very profound changes in the organization.”

 

Geneviève Petit