Mike Graen dicusses Computer Vision with Pierre Marie Rallu - CEO of 1retail AI - and topics including common mistakes retailers make when implementing new technology, how do you track and measure inventory in the back room, digital twins and the opportunities in supply chain warehouses and distribution centers that CV can provide.

Hello, my name is Mike Graen. Welcome to the

Supply Chain Management Research Council focused on on shelf

availability. I'm your host today and I'm joined today by

Pierre Marie Rallu, who's the CEO of 1retailAI. They are a

company that works very closely with Ses Imagotag to understand

on shelf availability using computer images. Please join me

for with my conversation with Pierre. Well, let's go back to

let's go back to the solution that you offer. You've obviously

done a lot of research about this tell us somewhat the and we

already talked about the fact it's not just technology,

there's a people component, there's a process component,

there's a technology component. What's best practices, if

somebody was trying to do computer vision, to generate a

list of out-of-stocks or incorrect products, etc for

their team, whether it's your solution or somebody else's,

what are the best practices to utilize to put that in and make

that work?

So in the in the, in the world of SES, for

example, there is the world of, I mean, do we have share? Do you

have for example, ESL or not on the on the on the on the shelf,

and that's, that's fine. But having ESL, as you mentioned

earlier, we'll get the price and promotions, velocity and things

like that, but there's a sync with the cameras. And, and it

helps the cameras to to always stay in sync. In the meantime,

the AI algorithm are now using the products in near real time

as well. So can work in both both both worlds, in terms of in

terms of how do we make that happen? The technology itself

works. I mean, this is a product, but on the fact that

it's going to find your your is going to find your hole is going

to work like this for the one I've tested in there. And I know

pretty well, the one within the SES framework, they work pretty

well. So the technology works. But as I mentioned earlier, it's

people process and technology and having the you can have the

best tool in the world if you don't know how to use it. And

it's pretty useless. So so what we've seen is it's very

important for, for, for a retailer, and also for the

vendor to work together. I mean, my team spent a lot of time in

the store with the store managers try to understand, for

example, what's the day in the life of the before, before the

technology, what's the day in the life look like? And you

could imagine, like every grocery store the same, so

slightly different. And so the devil is in the details. So

spending the time to understand the process, the existing

process, and then looking at how the technology can actually

bring the value. You'll need to bring a revolution. I mean that

you can start with an evolution. And then then looking forward

with multiple level shows you can bring a revolution

eventually. The other thing, which is super interesting with

those technologies, and then I'm looking at most of the in store

IoT isn't it we have, you can test them pretty easily. It

doesn't require to have IT teams involved having hundreds of

interface and integration built. I mean, the light touch on the

technology, it will be light touch on the technology. And

then you can test it one store two stores, a couple of stores

to see to see the value in the end. And today. A lot of me if

you look at the spectrum of the retail in North America, what

yeah, the biggest retailer in the world. But you also have a

very large number of local retailers, I mean small to mid

size, we might think they don't have the bandwidth to do to to

go with that kind of technology. But when respectively. I mean,

I've been in the software business for 20 years. And it's

much easier to install, for example, a computer vision, that

camera vision than having an capital inventory and story

punishment in place. Just because in terms of technologies

much in terms of data, it's much lighter touch to start with.

It's easy to install, it's physical, you can see it, and

you can see the result quickly. So So best practice, I would

say, well, first of all, try it. Don't don't feel free to try and

do the thing is that cost wise. We're not where we were five to

six years ago. The technology if I was so good, was three times

more expensive, and was watching the color a 10th of what he does

today. So we're in a much better place than we were five years

ago. And and here's the thing and again, minimum investment

Have a demo of it and data. So I think I think that's a, that's

something that's very, that's very important to keep in mind

for all the retailers out there.

Okay, excellent. So you've told us what the best

practices are. Again, promoting your Coca Cola from your wife,

that's awesome. A little friendly sponsorship there. What

are the watch outs?

There's a brand here, by the way.

What are the watch outs. What are the things that

people do that get them into trouble? They try and leverage

the technology, but they just make mistakes. What are the

components, some of the common mistakes you've seen them ma\ke

when they're trying to implement the solutions?

Well, I think that I mean, I've talked

to I've talked to retailers that have a very preconceived idea of

what they wanted to achieve. And that was not exactly what the

technology was designed for. And so trying to and try to get out

of the boundary is possible. But it's, it requires a partnership,

then you have to go, you have to be ready to go through the

journey of exploring emerging emerging technology boundaries,

and it can be painful. So So one of the other is to seek to

understand, okay, what are the existing business cases, what

has happened out there, what works, good deals, or retailers

try to understand what to do and what is being done. I think it's

one thing. The other thing is to, to go a step further and

making sure you work with a partner, a vendor, that is going

to spend time with you, again, as much to get into store, know,

your processes know what happened. I mean, again, that

type of technology, we're still we're not in the in a cookie

cutter type of technology, where you can pretty much do it, it's

still a bit tailored, doesn't have to take months and weeks

and months and things like that, but at least doing the initial

assessment of what's your day to day, what's the impact on your

process doesn't have to be huge, but needs to be done. I think

that's important. The other thing is, now, as I mentioned,

we can start with very limited data set. But what comes with

very limited data set is limited business case in front of it. So

for example, if you put the camera, you feed the you feed

the system with your list of products, okay? And then you're

going to start start crunching the data and give you Okay,

where are the hours? Okay? Well, technically, you can go on the

shelf, scan yourself, and you see where the outs are, right.

So it gives you that visibility without having to work your

shift, which is a great, great value already. Now, if you have

the capability to bring to feed the system with your inventory.

Well, it's good to benchmark what you have on the shelf, and

what you're supposed to have on your shelf. And it's going to

tell you, you're going to check those 50 products, because you

have you have stock on that product, whether it's empty on

the shelf, so that they give you a much better business case,

because now you're narrowing the scope of the product that you

need to check. Now, if you are self-feeding, this is and I mean

these are these are relatively simple set of data. But if you

feed in the system with average sales, for example, now the

system instead of only telling you measurement around out of

stock out of stock, sorry, and out of stock duration is going

to give you potential loss sales, which is emerged as a KPI

than just the out of stocks. Because having a an out on a

high velocity product, high margin product is much more

impacting than having an out on the product that doesn't turn.

And so when you want to check is focusing on what is essential.

And what makes sense. And therefore, I mean being able to

get some of this data already, you can unlock more business

cases. And then after you can integrate the planogram to the

appliance, you can do the price and promo checks and saying that

much more you can you can involve with that technology as

well.

Yep. So I know that you know this person pretty

well. So she's asking this question. I'm not but Susie

Mumford has asked how do you track inventory that's in the

back room?

Well that's a very very good point.

So that's probably why she, that's probably why she

asked it. She's wicked smart in this spot she spent a lot of

time with Gartner. She knew that that was going to be a very

tough question. So take it over to you.

I mean, of course I mean Susie knows

written pretty well.

Yeah.

Euphemism. The, the number one the number

one problem in the grocery store that you might know your

quantity on hand, more or less usually, usually bounce on and

is no more accurate than 75-85%. If you can, you can if you can

tell you can know what you have on hand, but you don't know

where this is in the store, and you have the back and the front,

obviously. And in the front, you will have multiple locations;

you have your end-caps, you have your main location, your primary

location, and maybe to a secondary location. So you might

them in mulitple locations as well. So in knowing what you

have in the, in the, in the, what you have in the back is

interesting, obviously, because like, I mean, that tells you

what to transfer when to transfer it and things like

that. Now, if you're in the if you're in the warehouse with a

very high bill, as you're sorry, in the grocery store, for

example, it was a very high velocity term, but it comes from

the warehouse or from the DSD. I mean, those pallets are always

moved to products, right. So, so they come in, usually, usually

they come in, they go straight to the floor, and then they

might come back with whatever leftovers are. But there are

multiple type of Fritos there. So if you stand up groceries I

work with, I work with many grocery store that have a very,

very lean back home, they really have don't have a lot. And the

policy is that, okay, if you have less than the case has to

stay on the shelf, or more than the case bring it back. If you

bring it back on the case, then it can be organized on the back

room. And if the backroom is relatively well organized, then

you can then you can have different systems to measure the

to measure the the level of inventory you have in the

backroom. And it doesn't have to be doesn't have to be at the

SKU, because it's not the case. But at least you can recognize

this case back as you recognize the the student on the front. So

that's one possibility. And then you also have I mean, depending

on the type of retail, for example, if you look at the

pharmacy, I mean, very close to a very large chain of

pharmacies, and their backroom are extremely well organized.

So. So it's kind of easy to, it's easy to track what's in the

back, as soon as you get organized in the backup. So

understanding what's in the backroom, there are multiple,

you can now then after you can use scale, depending on the type

of retail, you can use cameras, you can, you can there are many

ways to measure. But ultimately, it's a matter of labor rescues

return. So as a matter of labor. So for some categories can be

really interesting, or some others is zero value to try to

to count what's in the back. So that's only a bit and then the

fact of scanning, what's from the back to the front is

extremely consistent. And nobody does that. I mean, we've tried

that it's extremely labor consuming. And that's not

something which is possible. So so. So ultimately, ultimately,

the idea is that keep in the back, just your cases and things

like that. Keep on the front what you need, but more

importantly, get a much leaner and more accurate replenishment

based on your balance on hand so you don't have to keep much in

the back. And so if your backroom is not packed and it's

fairly empty, it's very easy to visually prepare everything. So

after it's a matter of organization. That's a tricky

question.

Yeah, no, it's a good question. Very good

question. So, I would also argue, and I'd love to hear your

reaction to this. I would also argue the retailer's not the

only customer of this data. But as you start, you mentioned

direct store delivery. If I was Coca Cola, I would love to know

how much Coca Cola is at the cooler at the front of the

checkout, how much Coca Cola was on the sales floor where it's

supposed to be bought? How much is in the backroom of the store,

before I ever arrived in the store? So do you see there's an

opportunity to share some of this data with suppliers whether

they're direct store delivery or otherwise? Because that's

insights of what's going on in the store that I don't think CPG

companies have ever had before.

That's absolutely correct. I mean, the

traction, the traction of the CPGs on that technology is

massive. It's, again that most of the big CPG I've used mobile,

Yeah, yep. Well, in a lot of cases, those direct

mobile, mobile CV mobile computer vision and AI and stuff

like that. But having a consistent visibility and, and

via an AI engine that recognize each location, each product and

each location in the store and then can monitor okay, what's

my, what's the velocity of my shelf has a turn was my level of

what's my level of outs. Another measure, which is really

interesting is the recurrence of the outs. Meaning is my

merchandising appropriate? If a product is out twice a day, we

have an issue if the product is is out three times a week, then

you have a replenishment issue. And then and then doing those

analytics is good to bring bring a huge amount of data to those

store delivery folks are paid for and compensated based upon

CPGs. The thing is, if you look at if you look at the most most

of the DSD, I mean I'm always shocked by the amount of DSD

how much product that I move off of my truck to the store,

products that are in the back rooms. This is, I mean, I would

say 50% of the back rooms that work. For all the backroom they

see almost 50% of the backroom is filled with DSD product. And

yes, some of, many of them are high velocity. I understand

that. But that's seems like a lot of inventory on the back

room compared to what's coming from the warehouse, for example,

or the distributor.

whether it's on the backroom or the sales floor. And I'm, I'm a

firm believer with the problem with reward systems as they work

right?

We've had this discussion 100,000 times

now. So yeah, that's funny, because every retailer was like,

"yeah, I know it's a problem." And, and that, that's

interesting, both. And anyway, traction from CPG is huge. And

what we see now and looking at part two, for example, one of

the things on the SES platform is having a portal for the

vendors to go and see the data and API's readily available for

them to digest those data. So all of that's already packaged,

I mean, where the retailer wants to see by department and

categories and things like that, for the CPG. We want to see its

brands. And so, so it's a different vision. And and for

the retailer, there is a history of monetization around that.

That's not free, right?

Yep, the way I think about I'm from Arkansas, so I'm

a pretty simple thinker. Here's the way I think of it: a

retailer should know exactly what they have, they should know

exactly where it's located in a retail store. But here's the

'but' without a single associate, or, or person at the

store, collecting data. So I'm not scanning anything, I'm not

wanting anything, I'm gonna let your system work in conjunction

with other systems to say, here's what I have. And here's

here's where it's located, at a high degree of accuracy - not

60% accuracy, but 95% accuracy. Now, running AI engines to take

that data and create tasks for the the employee to go get that

product in stock, because it's in the backroom, and not on the

sales floor, or the Coke person - your wife works for Coca Cola

- before I even get to the store, I should have a heat map

of the store that says exactly how much by SKU I've got. So

what do I need to bring into my truck? I don't have to come in

and take inventory first. I don't have to go fill out the

backroom with a product because I just want to make sure that

it's all there. I have the opportunity to create and

deliver product directly to the associate. That to me is one of

the visions that I would have if you share that data with the CPG

Yeah, it's it's really, really close. I

Okay, so another great question from our friend JW. Outside of

companies.

mean, I've seen, I've seen from the R&D team, I've seen, I've

seen search solutions of what we call the digital twin. So

really, I mean, that tells you I mean that's looking at the store

as it is really now with the cameras and the sensors, which

are remapping automatically the entire store in 3D. And that can

tell you where the hotspots are. But where are they where the

outs are. And on the shelves were but on the macro as well as

on the micro. And they're off that being redesigned completely

automatically. This is R&D. I mean, you have to see that, but

it's pretty awesome. But the but it's really like we have the

cameras taking the pictures. And then the cameras know that

positioning in the store on the footprint. And now building a 3d

view of the store where you can walk in and see the actual

shelf. So the fact that we can do that is what I mean it's it's

cool to see that I mean, but it also implies that we have a

level of data, which is can be shared and which are completely

actionable. Where in the store, think about different thinking

about your consumer well I'm looking for my Coke, I don't

know where it says and then I go on the on the retailer app and

as a type I type coke and then the system tells me turn right

on left and you're there. Yep, and it by the way, and it's on

the shelf. Yep. And so so that is a huge gain same thing for

the personal shopper as we talked about e-comm. One of, the

there are two areas where where the technology is bringing your

revolution to e-comm; first of all, I mean being able to tell a

consumer if a product is out at the time of placing the order so

I mean, that will be understanding it won't be

refilled by them. That might be strategy, but even more

importantly making sure that the personal shoppers is is notified

of all the exceptions before it started store. So so because

what's extremely expensive when you have customer shoppers in

the store which are building the baskets is every exception

requires a quite complex decision making process to

substitute. First of all, do we have it in the store? If we

don't have it, if we have it in a store, we need to call someone

go and check and say that if we and then and then customer

shopper is not producing some time. If the if the products not

in the store, what do we do? Do we? Do we just scratch it? Or do

we substitute? What do we substitute it with? So the

process is complicated. So removing that exception

management from the from the from the Personal Shopper store

is a huge benefit for the for the retailers and the shoppers

in general.

the retail store - think more upstream in the supply chain

warehouses, distribution centers, etc. - do you also see

opportunities AI or computer vision AI opportunities in those

locations as well?

Oh, yeah, I absolutely. I think I mean, I think that computer

vision and AI is the the next revolution. I mean, the really

the it's a huge technical revolution for many, many areas.

I mean, I focus on the retail because I've been there for 20

years, and I'm lazy and all that. But the but the but if you

think about it, I will take the Tesla's I mean, they are like

they have massive computer vision everywhere. In the

warehouse. Yes, of course, you can go in the warehouse. For

example, we've, we've done, we are working with in little,

little auto stores. But for B2B, for warehouses. For example, we

have a client, which is a manufacturer of plumbing,

plumbing hardware, and they put those little parts, it's

actually the contractor, they put them everywhere in the

country where the plumbers can go and scan, get in, take

whatever they want get out. And the idea here is the fact that

we have computer vision, the scales, wedding, from the nuts

and bolts, because it's computer vision, that's the limit. So and

this is where we start having to develop technology. And we also

have the price tag on it. Because every time they scan in

less than 30 seconds, the entire store prices are refreshed to

their own price because they have negotiated price. So it's a

b2b to b2b exception. In the warehouses, you're gonna find a

lot of a lot of technologies which are omitted, again, the

scales, the tags, the blinking image on the blinking the pick

to light. I mean, there are a lot of love technology that can

be used. Computer vision can be used to data to detect defaults,

defaults on the shelf, to say, "Oh, that's cool that we are

seeing there are things missing." And this is when and

in a different environment, you can also think about different

process. Drones might be a good solution. If you have a drone

that have like a regular tour, checking for the fixtures,

checking the inventory on the on the third or fourth row. And

just kind of saying that that's absolutely. I mean, this is not,

it's not 10 years from now. It's now

Yeah, that's awesome. Well, believe it or

not, we're almost out of time. It feels like we just got

started. Let me ask you one kind of final question. The question

is and this is my one of my favorite questions at the end:

what's hot right now? What's the question that I should have

asked you, but I didn't right? So what's on your mind or any

closing thoughts that you might have or anything that you'd like

to share that wasn't necessarily a question that you got asked?

Fascinating. Yeah, we didn't even we didn't scratch

My conviction, my conviction is

that in the years to come, I mean technology will drive the

acceleration of the evolution toward a, a fully digital retail

store; like a really like a very highly digitalized physical

store, which is fast, efficient and convenient. We have today we

talked a lot about how we improve the process and the

store. I think we didn't talk too much about the customer

experience. And and what we imply that if you find your

the surface of the customer journey. But at the end of the

product on the shelf you have a better experience. But I think I

think a lot of the technology will will be geared toward

customer experience, the store, the store of the future will be

both a showroom for the clients to find new experiences and

storage for the e-comm to be able to fulfill the the orders

and therefore why you need to marry those two visions and how

do you, how do you, and how you transform your your customer

experience into stores is definitely what's going to

determine the winners from the losers tomorrow.

day, that's why we're all trying to make sure the products are

available available for them, when, when and where they want

them. So, Pierre, thank you so much. I know you've, you're a

little bit jet lagged, you've been on a plane all day you just

got back into Atlanta, you're home. We definitely appreciate

you taking time out of your busy schedule. Boy, computer vision

AI, the things that you're doing are really, really breakthrough.

And it's really exciting to see the digital digitization of the

store shelf really come to life. So we thank you for that very

much. We will have your LinkedIn bio as part of our podcast out

there. So if people want to get in touch with you for any

reason, they can do that. So thank you very much. Have a

great weekend. And again, thanks for your time.

Thank you, Mike. Have fun!

Bye bye. Well, I hope you enjoyed that

conversation with Pierre regarding computer vision and

what they're doing at one retail AI. Please join us next time

when we are joined by Steve Statler. He is the Chief

Marketing Officer of Wiliot Technology. He'll be talking a

lot about how Bluetooth enabled RFID technology can be used in

the online or offline retail supply chain. Join us then!