AI in Retail: Beyond The Hype and Into Reality for Store Operations.
The term "AI" is often thrown around, but what does it really mean?
To be the simple solution to a
big business problem
Our CEO and Founder Julian was recently invited back on to Chris Walton’s ‘Ask An Expert’ series.
Chris is the CEO and co-founder of Omni Talk, one of the leading blogs and podcasts in the retail industry. He is a regular Senior Contributor for Forbes. Previously Chris also worked for Target, where he was the Vice President of their Store of the Future project.
You can check out the full audio here, but we’ve also adapted the conversation to read as a blog (because it was that good).
Everyone’s talking about agility right now. Different retailers define it in different ways, but it effectively comes down to one need: how do I simplify work for my stores?
Simplifying can take on lots of forms: How do I take work off my store teams? How do I prioritize where they spend their time? How do I streamline store ops processes? How do I think about engaging my store colleagues? How do ensure the business rapidly responds to issues and learns from them?
Everyone is trying to focus on this nimbleness, and tech is only a small part of that. Most of it is about operational changes and, in some cases, people changes too.
For me, the term “agile retail” boils down to simplicity, speed, and engagement.
Pretty much every retailer I’m speaking to at the moment has the same technology dream – a single unified store app, that brings together the different tasks and data-driven alerts that each store colleague should be doing each day and intelligently prioritizes them. Tech that says, “Hey Chris, just go do these five things today.” And then let’s burn all the reports we don’t need and switch off all our other apps.
In the past four or five years, lots of people have done a lot of work to set up tools for handling compliance tasks. And they’re great, but it’s only part of the solution. You’ve still got all that data and all those reports that you’re throwing at people. Sales reports, shrink reports, damages reports. All kinds. So everyone’s now saying: how can you bring it all together? How can you, on one platform, tell me that I’ve run out of bananas in aisle seven and congratulate Bill because it’s his birthday today?
There are several very good vendors out there who handle tasks well and a number of big retailers have built their own too. The next bit – the data-driven alerts – is harder. You’re having to solve bringing together data from e.g. 14 different systems or 14 different use cases: sales, waste, CSAT, shrink etc. and turning it into something so simple that someone new to retail can go and take action that same day.
Prioritizing tasks is pretty easy – it just takes someone to look at the list and prioritize it. But prioritizing tasks and data alerts in a way that’s driven by deep-level, real-time insight derived from disparate management information, in a way that is simple enough that anyone can understand it, that’s where art and technology come in.
It’s hard. Speak to any large retailer and they’ll also tell you that: “Yes we’ve had a project to work on that, but we’ve been working on it for five years and gotten nowhere.”
Personalizing data is really hard to do.
So why is it so hard to do these data-driven alerts? It’s hard because it’s personalized.
In a typical thousand-store chain on any given day or any given week, our Quorso data shows only four stores will get the same data alerts. If you think about it, every store sells quite a big range – maybe 5,000 SKUs or even 50,000. That’s a lot of different areas that can go wrong. So the probability that any two stores simultaneously go wrong in the exact same area is very low.
Secondly, there’s no single cocktail recipe for coming up with the best alerts. Here’s an example. When we get started working with new customers, we typically find in the first two months of working with them that you can drive a lot of value by finding structural problems that otherwise get overlooked.
What do I mean by a structural problem? Let’s say your sales of breaded chicken are always $1,000 lower than they should be. You take a look and realize the planogram means certain SKUs are hidden behind a pillar and never sell well. This sort of thing would normally pass you by totally, but the data alerts you to a big problem you’ve been totally overlooking.
Once you’ve dealt with these big structural problems, the next stage is to start monitoring for one-off sudden drops. Why are my banana sales suddenly falling off? Why has my shrink suddenly spiked? Why are so many mangoes going to waste this month? Again, identifying these drop-offs requires highly personalized, real-time analysis of store performance data.
But personalizing data is one of the most valuable things a retailer can do – if they get it right. As I’ve said, these issues at store level vary hugely between stores – any given issue will only apply to about four of your stores. But that’s not how retail operates – it assumes that an issue applies to every single one of your stores. Retailer operators build a prototype, then rinse and repeat, thinking that if it works in one location it’s going to work in other locations too. So direction from headquarters has always been top-down, commanding the stores to do what they think will drive the most sales performance and productivity.
What we’re saying though is you can actually do it the other way up – stores can take their direction from real-time issues that are being flagged in store-level data and fix them instantly. Then share learnings back to the center so the full network can benefit. That’s a pretty phenomenal epiphany.
Most retailers have got everyone to run a decent store. The next curve they can answer is, “how do I get every store to be as good as it can be?”
Using data and science to inform retailers and store ops staff what they need to do day-to-day is one thing. Engaging them to do it – and enjoy it – every day is a whole other challenge.
We’re learning so much from our user data on this piece. One example is we’ve found we shouldn’t ask people to do too many things at once. In a list of 10 data alerts, the completion rate of alerts 6-10 is between 30-40% lower than the completion rate on alerts 1-5. So we realized we should be giving people five things to do, waiting till they’ve completed them all, then giving another five. Don’t give them 50!
Another interesting one is that people get really upset by being repeatedly bugged about the same problems. We’ve seen that the improvement rate is almost twice as high if they’re working on something new, compared to a problem they’ve had to work on before. If they’re told over and over to look at cookie sales their eyes glaze over, but tell them to look at cookies, then water, then beer, that’s much more fun.
Think of it like a giant set of switches. Effectively what we do for the retailers we work with, who are running maybe a $10 – $50 billion sales line, is move those switches to optimize the daily behaviors of colleagues in many thousands of stores.
Whether it’s Quorso or not, I think you need to look at technology that can actually sift through all the things you’re asking your store managers to do and identify the five things that each of them should be focusing on any given day. You’ve got to find a way to cut back on all their current workload volume.
We’ve been working with a big US retailer to bring all their damages data into Quorso. They had been sending damages data to 10,000 stores for years. What we found was that it’s actually really difficult for stores to reduce damages because about 95% occur when a cooler breaks, and when that happens there’s not much they can do about it. So the central ops team learned that they should instead buy better coolers instead of overwhelming stores with unactionable damages data.
You could see that as a failure – we spent time bringing damages data in and then they just went and switched it off. They now don’t share any damages data with any of their stores. I regard that as actually a huge win – we showed them where they were wasting huge amounts of people’s time – and that’s exactly the kind of simplification retailers need to be making right now.
A lot of retail execs are trying really hard to motivate and engage their teams at the moment. People are really stepping up their leadership and realizing the humanity they must show.
One of the things we’re seeing is that you’ve got to give people meaning. They may complete a list of things you give them to do because they’re a good soldier, but you’re basically asking them to be a compliant robot.
However, if you show them three things and say, “Can you go and figure out what’s going on here and I’ll tell you if the action you take fixes it or not?” And then, if they do fix it, you say, “You fixed that issue with the peanut butter, and you sold an extra $500 the next day as a result. Congrats!”
That’s so much more engaging – it’s kind of gamifying it. It’s showing them how they’re making a difference at work. What I find really exciting is when you see people who’ve been in retail, in some cases for 20 or 30 years, and others just assume they will be technology resistant, but we see a light come on in their eyes when they go through this sort of experience.
A good example of this is a big European home improvement store using Quorso. Someone came up with a clever idea to move the lawnmowers to a different place in the store. Some were even put outside the store. Sales increased six-fold, and that person was amazed at what a great merchant they were and they were recognized by their District Manager as a result.
In addition to creating meaningful work, another major behavioral motivator is healthy competition. Stores can actually be quite isolated from one another – on their own islands, so to speak. But if you can gamify the experience it creates so much healthy energy. One way Quorso encourages this is with leaderboards – every time we show you how much more you’re selling, we also tell you how much you’ve improved in the overall store ranking. You sold an extra $6,000 in lawnmowers and you went from being the 830th store to 567th. And then we give people the chance to pile in and comment and congratulate and emoji. All that kind of stuff.
One of the biggest things for us at the moment has been integrating with Microsoft Teams. Teams is the biggest commerce platform out there – 130 million users – and its presence at store level continues to increase rapidly. And now you can plug Quorso in, within a day, and have all of your Quorso Missions delivered via your Teams app rather than via Quorso, so your store managers can stay in the platform they are already using every day.
There are two moments I enjoy most when we start working with a new retailer. Most start off extremely skeptical, but the first really fun moment is about eight days in when I get a call from the VP of Store Ops and he or she says, “I’ve just taken your app into three different stores and looked at the Missions you’re suggesting to that store and you’re finding stuff I haven’t spotted in my last five years doing this job.”
We find the stuff that they would spot too, but we also find stuff like the pricing being wrong or an architectural problem with the store, or that a planogram doesn’t work in a particular store.
It’s so rewarding when they ring up and go, “You know, actually, I’m starting to really believe in what you do.”
Then the second big moment is at about 30 days. And at that point they’re seeing the basic improvement – e.g., this store has just sold an extra $300 of cookies. All good stuff they like to see. But then they start to see the bigger picture – that their stores are up 1%, even 1.5%. And they start to realize the power of compounding – what happens when lots and lots of stores are simply doing more of the right thing, every single day.
Each action might only add $200 in sales per week by itself, but when you multiply that right across the store network it stacks up in real money astonishingly quickly. And they’re shocked by that because they’re so conditioned to think about big central initiatives.
