How does Majid Al Futtaim measure the efficiency of your analytics?
Guillaume Thfoin. It’s on a case-by-case basis. When we talk about analytics, is shouldn’t be just about money. It’s really how you transform this organization, how you change people’s mindset. We’re going to deliver a suite of tools to give people access to the data, analytics and use cases.
On the use cases, you have financial KPIs because they are supposed to trigger money, but the rest is how much data we have, how clean it is, how much of the population we cover – not how many dashboards we build, but how many dashboards are consumed and finally, how many complex analyses are being asked to the analytics team.
Joe Abi Akl. The simple logic is that data is required the same way that, when you open a store, you put a door. No one measures the ROI on the door or the lighting, because they have to be there; you cannot do business without them.
Now, we use data in different ways: some you would be able to measure in a qualitative way, others in a more quantitative way. If you think of data as an internal currency to optimize your operations, it is actually much easier to measure because you know you have optimized or automated a process and you know the ROI on that. On assortment for example, we’ve been seeing a 2.3% uplift in sales over the past five months.
If you use data as a currency for stakeholder relationship building – offering mall tenants value-added services and insights that they can leverage to do better business, on top of the physical space – it makes you way more interesting, but can you measure this? No. What you’re eventually going to see is an uplift on the revenues.
If you use data as a currency for internal stakeholder management – providing someone in operations or marketing with all the analytics and the insights that they need to do their job better – it’s very difficult to pinpoint exactly what’s the impact, but it’s logically and intuitively there.
And if you use data as a currency for better decision-making – you want to start a new business, go to a new market, launch a new product – all of these decisions can be driven by analytics.
A simple example is the parking issue. According to our studies a year and a half ago, one of Mall of the Emirates’s biggest pain points was that people couldn’t find parking spots. Everyone thought we should go invest a couple of million in expanding that parking. The team went, did a simple analysis using the sensors, and found that 20% of the spots were available at peak time. People were not finding them because the way we managed our flow wasn’t optimal. So, the team changed the flow and now Mall of the Emirates doesn’t have a parking issue. It’s a simple process. In effect, how can you say that it has an impact? It’s not the revenue; it’s cost optimization and customer experience optimization.
How are you preparing for the introduction of 5G?
Joe Abi Akl. 5G is going to help us to collect more data from an IoT perspective. Then, it becomes a question of how much more storage we’ll need and starting to think of use cases to leverage this. In our data strategy, we’re thinking of the next things we need to start collecting data from, even from a strategic direction. We don’t collect data for our business needs only. If we want to be a leader in this, we need to collect data that we don’t even know why we are collecting it. Because once it’s there, you can figure out ways of using it.
How are you using new technologies like face recognition?
Guillaume Thfoin. We are already using camera vision – not face recognition – to do gender, ethnicity and age recognition in shopping malls. For example, today, when a cinema is about to launch a movie or plans for a movie, they don’t know who is going to watch it. The app might say that, for a kid’s movie like Frozen, the customer is a 40-year old male because, usually, the dad pays for the family. But they don’t exactly know who sees the movies and they plan using bad data. Now, today, we are working with Fox and with our brand VOX Cinemas, using cameras to see the audience composition of each movie. That’s how we use those technologies, which are quite cutting-edge, to predict how we can better target. But there’s no face storage anywhere. It’s a process on the fly.
What about AI and machine learning?
Guillaume Thfoin. When you have 1.3 million loyal customers at Carrefour – and 200 nationalities in the UAE – making rules cannot apply. You can only scale through machine learning to do that.
We use AI on things like assortment in Carrefour. Typically, product listing and de-listing was done leaving it to the people in the store to decide what’s working, what’s not working, what to put where and how to do it. Now, all of this is, to an extent, driven by the analytics work we’ve done and the product we’ve provided to them. Assortment is no longer based on the gut feeling of the store manager, but on what the tool is saying is happening in that store versus this other store.
Promotion work also used to be based either on human decisions or on high-level metrics. Now all the promotional work is done in a dynamic way, taking customer-level data to understand exactly how to do it, who to do it with, in which store, in which area, etc.
The last example is the Carrefour leaflet. This 20-page leaflet is the most-read publication in the UAE, and it is done with machine learning every 10 days. We looked at all the transaction history, taking into account the cycle of purchase history, and developed a tool that identifies what items to put there, in which page, in which order. Now that we do it to programmatically versus last year, we’ve seen an 11.5% uplift on every single campaign on average.
There’s no more guesswork.
This article was first published in communicateonline.me