Host Mike Graen is joined by Barbara Chase - Vice President, Global Sales Store Intelligence at ReTech Labs, Inc./SymphonyAI - for a conversation about leveraging computer vision to identify OSA alerts for resolution including:

• Customer perceptions of store intelligence solutions. (3:31)
• What is store intelligence and how does it work? (7:54)
• Hardware options. (11:22)
• How to identify out of stock products. (15:29)
• The criticality of recognizing out of stock products. (19:38)
• The problem with phantom inventory. (24:14)
• How to be customer facing. (27:33)
• Reverse planogram for cereal. (31:48)

Greetings. My name is Mike Graen. Welcome to

another edition of the podcast focusing on on shelf

availability sponsored by the Walton Supply Chain Center at

the University of Arkansas. Today, we are excited to have

Barbara Chase with us. Barbara is a longtime professional in

the on shelf availability area within the industry. She is

currently Vice President of Global Sales for ReTech or

Symphony AI. We talked to her about leveraging computer vision

in store to understand things like out of stocks, or incorrect

products or pricing discrepancies, etc. Let's join

our conversation with Barbara.

Yes, I am the Vice President of Global Sales

for store intelligence, originally ReTech labs, but

recently acquired by Symphony AI. And I've been in this

shelf-AI space as a believer for almost five years now. Coming

from the first stop was Bossa Nova. And so I learned so much

from that amazing company as AI and robots got, they got

started. And then from there, I've been at a couple of other

places in that space, and really wanted to join ReTech because of

the diversity of use cases that they that they solve. And so I'm

really fortunate to have been here now almost seven months.

It makes you an expert. You got it all figured

out. Right? That's fascinating. That's awesome. Well, thank you

for spending the time with us today. We're really excited and

everything. That's the one thing about on shelf availability, I

think whether we continue this podcast for one more week or 20

more years, it's always going to be a topic. Because every single

store every single day, every single item, in a normal big box

retailer, you could have 120, 530,000 SKU's at any one point

in time, you could have some of them out. And so the idea is,

can you deliver signals, whether the signals happen to be video

signals, or RFID signals or something that alerts people to

go, Hey, we have an issue here, let's get them resolved. Because

I think like people can solve the issues. But they they're not

they don't do a good job of auditing, and checking and

counting and things like that. So before we get into the

details, though, I have asked this question for the last year.

So I'm gonna put you on the spot as well, Barbara: at the end of

the day, we are retail technology, retail supply chain

professionals, that's what we do. But one of the things that

we also do is we're customers, so we go into stores, and we

grab our baskets, or we put products into our basket and

checkout. Or we log online and order something online and go

pick it up in the store. Sometimes it's available, and

sometimes it's not. So I always ask my guests before we talk

about this solution and the opportunity put yourself in the

point of being a customer, have you had a bad experience of

something that you expected to see in the store and didn't

find? Don't throw the retailer under the bus? Don't want to get

it all that but just give us the example because it's really did

it hasn't happened to you? And number two, how does that make

you feel? And you know, what were some of the customer

perceptions of that?

Yeah, and it's not just once it's actually

frequently I will go to numerous retailers. It's not one - I

think everybody struggles with that problem. I've, you know,

I've had situations where I'm about ready to host a party and

I am getting everything aligned and something obvious, like

paper plates forgotten. I got all the food looks great. But I

don't have everblades So yeah, you start with one store and

think that that's something easy to grab and put in but then

they're not there. And so then it becomes even more stressful.

Because you're you're driving to the next and sometimes the next

so yeah, it's not only frustrating but it but it is

stressful. And and I will say I'm an ardent supporter of brick

and mortar. So I do everything in my power to resist the the

800 lbs gorilla of amazon.com and instead support

brick-and-mortar and that's here and it's some sometimes on some

items that makes it difficult so I, I want to be the calvary to

the existing brick-and-mortar because I am so in awe of those

operations. And it's so important to me to be able to

have something that I can just walk down the street and get to.

So I want to do everything that I can to help them stay not only

survive, but to thrive.

Yeah. And even in a practical environment, if you've

got guests showing up in three hours and you need paper plates,

Amazon's not an option. That gives you a couple days that's

party be over by that. I don't know if those little drone

things are ever. We keep hearing. Yeah, but they moved to

Northwest Arkansas. There are people that are literally

getting rotisserie chickens delivered for dinner via drones

from Walmart. They're pretty fast. To show you that show you

that next time you're in town. That's pretty fascinating. How

do you I don't know what we're having for dinner tonight?

That's okay, the rotisserie chickens flying in now.

OK. I have yet, I have yet to see that. So I will

have to go to Bentonville and check it out.

Yeah, for sure. For sure. Well, I don't know you've

had I have no idea how to segue into retail technology after

that one. But thank you for that little laugh. It is Friday,

let's have a little bit of fun here. So. So I've got a store,

this store has got 120,000 items on it. Okay. You've been

involved in this space for a long time? How do I figure out

what's there? What's not there? What changes? And you believe

you have - you coined a phrase, which I'd never heard before,

called store intelligence solutions. That sounds

impressive. I'm not exactly sure what that means. So why don't

you elaborate on that one a little bit for us.

It can be taken in a couple of different

contexts, like SmartStore, we've used in our languaging.

Retailers have also said, the intelligence store, we're

working, you know, for the intelligence store. So it's kind

of a an amalgamation of those, of that terminology of store

intelligence, because it's never previously been available

before. I mean, yes, there's intelligence and data from a

point of sale and supply, as, as we all know, but it is like an

old newspaper a day late and a dollar short, even if you can

get something in, you know, 6, 8, 10 hours, it's still not at

that time. And there are certain instances, demographics,

categories, where you really do need to understand from a

customer satisfaction, and not not letting you know that person

down, letting them not be stressed out for their party and

making sure those paper plates are there, that you need to have

the eyes of the customer. And so store intelligence really

encompasses that now for the first time. With artificial

intelligence, and computer vision and optical character

recognition, we can be the calvary. Sometimes I say, you

know, turn your staff into superheroes, because it is it is

amazing what's expected, you know, of an in-store staff, and

increasing demands, you know, from customers themselves. But

it's it this is, I guess, a long answer what it is, is exactly

having the eyes of the customer and the lens of the

merchant, and so that, that by being able to see that and have

that data, then the retailers or CPGs can gain more more

information in regard to how their customers are shopping.

And therefore does it does it make sense to add more facings

does it add sense you know, make sense to add more end caps or

displays or you know, there's there's millions of dollars

spent on on on on best guesses you know, I mean that

optimistically and hopefully however, from a board

perspective, or a CEOs perspective, having intelligence

to validate that you know, that display or that pricing or that

positioning is resulting in the in in this these kind of

results, you know, and always to increase sales and customer

satisfaction is the goal.

So, give me a practical example I am a

associate that works in health and beauty and department in a

store and a Walgreens or a CVS or a Walmart or Target or

whatever. I've got cosmetics, I've got toothpaste, I got

toothbrushes, I got deodorants, I got scope mouthwash. I got all

these products here. How do I know what's there and what's not

there? How does the store intelligent application help me

to tell me? Because if I go and scan every one of them, I guess

I could probably figure that out. I don't have time for that

I got 40,000 things other than to do plus, my job is really to

interface with the customer. So how do you, how do you automate

that and deliver that to me in a way that's, that's easy to

ingest to be able to potentially fix?

Yeah, and actually, I do think a visual

would would help here because the great thing is with with our

solution, you do have that choice. So let me see if I can

bring that bring that up here. And let me know if you can, can

you see my screen?

Not yet. There we go.

Okay. So that's a great, that is a great question.

And with a lot of solutions out there, there's usually a

singular answer, that it's from the robot or from a camera. The

great thing about our store intelligence solution is it's

hardware-agnostic. So that means that the image-capture options

is the retailer's choice. And then is also available for CPGs.

For instance, a CPG is not going to invest in any store hardware,

but they typically have phones or tablets, for their, for their

reps or for their brokers. And so those can be used. Because

when you think about category management, you know, it's

hitting multiple people's, it's not just the staffer that you're

talking about, that has those responsibilities in store, but

it's the category management all the way up to the headquarters

and to the CPGs, and then talking and negotiating about

that. And so what's great about where this starts, which is the

image capture, is having that choice. And having that control

allows you to not have to put your career neck on the line.

Hoping that this is the right thing for your organization, it

even gets farther than that, you may want to do a hybrid

solution, where maybe you're going to do a certain way of

capturing in this banner or in this demographic, or even maybe

you wanted to be a hybrid in a store. If you've got a

hypermarket with to your point 100, 110,000 SKU's like a

typical Walmart's gonna have. There's high velocity items and

categories, and then slower. So it might make more sense to have

a fixed camera in some places of the store, and robots roaming

the rest of the store. So and there's so many things that we

together don't know that we don't know, until we start

working with an organization and understanding their operations.

It takes a minute to really make those kinds of recommendations.

And I think that's what makes me feel good working here. Because

we can work in that consultative perspective, because we have a

mobile app, we honestly do not care and do not have bias on the

hardware itself. So with that said, if you There's choices in

regard to a tablet, or mobile, and then there's also choices in

regard to that staffer that you were talking about, or a third

party broker, taking those images for you. And again,

depending on ROI and number of stores, you know, those those

decisions will happen differently by a customer. So

you've actually got four decisions, you know, in this

one, am I going to do a tablet, am I going to do a mobile, you

know, existing Zebra device or whatever is there? Am I going to

have my people or different people do it? Then here with

this choice, you can do a shelf camera, a shelf camera, which is

here. They're about this size, and they're battery operated. Or

you can do a ceiling camera, which you can barely see up

here. This is one of our convenience store customers a

picture in here. And so and this is DC, so this is electrical. So

you have to have low enough ceilings, you know, something

lower than 12 foot so there are some things there. And then from

a robot perspective, Brain Corp is our partner, proud partner

and so you can either use a multi-use robot, so a robot that

not only cleans the floors, but then can do the scanning. And

that's what this arm is here. Or if you want a smaller format,

then they have just released their scanning-only robot. So

there's a lots of choices and decisions with hardware. And

because it's it's a larger investment, I believe strongly,

it's important to have those choices.

Gotcha. Terrific. So let me make sure that I'm clear.

So regardless of what the data capture is, I'm looking down at

the shelf. At the bottom of this page, I see a shelf label, like

up on that top shelf on towards the right hand side a shelf

label with no product above it. So I'm assuming you're telling

me that particular label whatever that product happens to

be, whatever that UPC? is, there is an out of stock for that

particular associate to fix? Is that example one of the things

Yeah, yes, it is. And that's a nice segue.

you can do?

Segue into what else can you tell me? Ok, cool!

Voila! While you, yeah, so you people may have

heard the term, realograms and planograms. So realograms is a

picture of what's really there. And then I always say, now put

your, your computer vision goggles on. And this is the way

computer vision operates. So in this in this picture, green is

good. That means that, you know, yay, we accomplished the right

product, in the right position at the right price. So that gets

the green color. The purple color up here is oh, we don't

identify that we don't see that. So we're going to need a we

don't recognize it. So we just need to do another scan, and

make sure that that then we can recognize, know what it is to,

to create action.

Let me stop you there. So the purple one could

mean, I read it too fast, it didn't get a good quality image

of it. So I can't really look what it is right? Or I would

imagine it also says, I don't have that image or that

particular picture in my image library. Right?

It's the latter.

It's the latter? Okay.

So that's what we need is a match.

So somebody from a software company is blaming the

software, not the hardware, this is awesome. This is fantastic, I

love this!

We're good, guys! We're good guys!

I'm gonna stop you, slow you down here a little bit.

Because I know working on computer image recognition,

suppliers, I believe get paid for marketing and constantly

changing products. It could be the exact same selling unit. But

it's a bonus pack or 50% more. You know, Wheaties does that

great thing, which is I get a different athlete out of the box

every single week. So how in the world do you keep that image

library up to date based upon how fast the suppliers are

changing their artwork?

Yeah. So that is an integration and collaboration

with the retailers and CPGs because yeah, new item cut ins.

We've got a CPG who does cards like Pokemon cards or baseball

cards, they're coming out with new products every week, and

planograms every week. And so, it's nothing short of a miracle

for them to hope that that that everything gets on the 1000s in

the 1000s of stores that they have and so - hard to sleep at

night. So now that we're showing those visuals of you know, uh

oh, we got a missing price tag here and uh oh, you know, we

don't recognize, and uho h, this is in the wrong place. It's it's

a it's so exciting to work with them because to see their faces.

They're so excited to, finally for the first time, be able to

see this data and then therefore, from that, have more

poignant conversations with their individual retailers and

where those fixes need to occur.

Well I slowed you down? Okay, that was the purple

box. You were about to explain the yellow box because I saw

because I want to double click on the purple box first.

Okay. Yeah. And so the yellow we actually it is

it well, it could be out of stock. We say not on shelf. And

the reason we say not on shelf is because of this little guy

down here. So the blue is a wanderer. So it's, it's supposed

to be in this planogram. But uh oh, not there. So maybe it

actually should be here. Maybe this is a PowerPoint, so I can't

click on it, but in the app, I could click on it, and it would

tell me what the expected position is. So then these

out-of-stocks then aren't really an out-of-stock, if it's just

misplaced, and that's the criticality and the importance

of being able to recognize a product that not only is a

Cheerios box, not to Kix box, but that a Cheerios box is not a

Honey Nut Cheerios box. Because, if you can imagine, again, the

just, I again, I will say it a million times, I'm in awe of

grocery operations, and I feel like, you know, where they might

have 100 employees in a store, they really need 500 employees

in a store, but then the, the vendor wouldn't be profitable,

you know, so it's it, or the prices would be too high, and no

one would shop there. And so it's this constant dance of, you

know, how do we do everything that we need to do; order

products, stock products, you know, make sure that

damage/expired is gone. And, and that's why it gets overwhelming

to, you know, every retailer I've ever spoken to in the last

eight years I've been dealing with retail, is they're still

doing that daily walk. And they'll call it they're going to

shoot holes, or they're going to do gap scans. And you'll hear

different terminology for it. But the reason that they have to

do that is for that very reason:

they can have all the data they

do that is for that very reason:

want from what has sold and inventory, but it doesn't take

do that is for that very reason:

into account shrink. Which is that expired, damaged, or stolen

do that is for that very reason:

or misplaced items. And so then when you're trying to reconcile

do that is for that very reason:

your inventory, that's again, why this is so critical, because

do that is for that very reason:

you can have that Honey Nut Cheerios box is is not selling

do that is for that very reason:

from your point of sale data is there's, why isn't why did Honey

do that is for that very reason:

Nut Cheerios Stop, stop selling all the sudden, but because

do that is for that very reason:

there's computer generated ordering, which is great from a

do that is for that very reason:

from an automation perspective, but it thinks that there's five

do that is for that very reason:

boxes somewhere, so it's not going to auto generate an order.

do that is for that very reason:

Because the data says that there's five, but the customer

do that is for that very reason:

sees zero, and therefore the point of sale, data equals zero.

do that is for that very reason:

So everybody loses from a dissatisfied customer to you

do that is for that very reason:

know, you know, lower sales, you know, et cetera, et cetera. And

do that is for that very reason:

that's why, you know, this is, this is just so on fire right

do that is for that very reason:

now, because everybody wants to be able to get access to this to

do that is for that very reason:

know then what to do reconcile their inventories, so that their

do that is for that very reason:

orders can be more accurate. And they can achieve those goals of

do that is for that very reason:

truck to shelf so that they have limited, you know, items sitting

do that is for that very reason:

in the back. So that is the that's the yellow, and then the

do that is for that very reason:

red is what we call an invader. I feel like I'm playing a video

do that is for that very reason:

game, but it just means that oh, this doesn't belong in this

do that is for that very reason:

planogram. So it needs to go into a different planogram. And

do that is for that very reason:

again, if I if I click on here, it would, it would tell me that.

do that is for that very reason:

And so that's what all the difference colors mean. And

do that is for that very reason:

again, it's not only recognizing the product, Honey Nut Cheerios

do that is for that very reason:

versus Cheerios, but then it's also comparing the price. So is

do that is for that very reason:

this the right price? And I think that's what's really

do that is for that very reason:

exciting is the more use cases that a retailer or CPG can have

do that is for that very reason:

to defend the business justification on technology. The

do that is for that very reason:

then you'll have success and reasons to buy it.

Yep, that's awesome. So double click into the hay,

the inventory system says I have ten, but I don't have any. We

don't sell any because we don't have any and we don't reorder

any because I think we have ten. We're good, right? So that term

is called phantom inventory or otherwise known as ghost

inventory. And the reality is, that's really one of the big

reasons why radio frequency identification in apparel in

general merchandise for the most part, are so big right now

because unlike this where I may have a holding capacity of 20 or

30. In apparel, I have a holding capacity of two so if I'm off by

50% I don't have any it's bad. It's a bad deal or if I if I

decide to buy one of these that are yellow, but you don't have

it, I may switch to a different form flavor. But if I have an HP

printer with a specific printer cartridge that I need, I don't

care how many printer cartridges you that you've got, you need

the one for you, or you're getting on Amazon and ordering

it. So I think grocery is a different animal. But it's the

same reasons, you have an out-of-stock, you have what we

call a plug or an incorrect product, that's your red ones, I

have a situation where I have ghost inventory. The other one

is comparing it to the planogram; I'm missing a label.

There's not only the product is missing, but the label is

missing and, as you eloquently laid out, pricing is also a big

deal. Does the point of sale at the register versus what's on

the shelf label match? And God forbid that the higher shelf

price is at the point of sale. So I thought the shelf said $2 I

get to the register it says $2.80. Now I'm really

disappointed because I feel like you're being dishonest with me,

retailer. So this is this is really really cool. So what what

exactly ... so if I'm a store associate, what do I do with all

that? That's just a lot of things that tell me where it's

wrong? How do I fix these issues?

Well, and that's a great point, right? Because

you can have, you know, death by data. We can, if we can action

against it, then then what's the point? Right? Um, so again, it

choices, choices, choices. So we use the AI to create a

prioritized work list. You see how this says item list right

here? And that is set on the KPIs of an individual store, if

you want to. Not only a region or a banner or to the chain in

its entirety, but we can go that granularly. And so, you know,

there's lots of things going on in a grocery store. And they

might only have, you know, 10 minutes or 15 minutes in a day.

And what's great is we're going to do the thinking and the

worrying for that staff. So then when they look at their device,

they can see in order, what's the priority? Is it based on own

brands? Is it based on velocity? Is it based on profit margins?

Is it based on what's available in the back room? So all that

thinking and wondering, that's what's awesome about AI

-artificial intelligence - is it takes that mass amount of data,

and in seconds, make sense of it. And so you know, exactly I

want to be customer-facing, I need to make sure that there's

not a long queue in the lines, I need to make sure that I can

see, you know, someone's parent looking desperately for an item

that that I can come help them and offer that that service,

which, which we humans love. And that's a differentiator, I, I go

to my neighborhood retailer, so I can hug Cole every day,

because I just, I just love him, and he is such a special bagger,

you know, and that's something I can't get in the car. So it's

making those, it's making those differences so that those

retailers can can actually shine. So then therefore, it can

be the the employee that you know, has those 10 or 15 minute,

you know, availability. Or again, if there's a you know, a

relationship with a broker or third party auditor, then that's

something to that they can work against it to be that that

calvary and help as well.

Got it. So we have a question from the audience on

your thickets your second to last shelf - I love it, you call

them invaders, the red?

Invaders. Yeah.

So you've, you've also got one that's kind of a

purplish color. And their question is, is that a

designation with the color? Or is it just the fact that it's

pushed back a little bit? What exactly is that?

Yeah, no, that's that. Well, that's a couple of

things. Actually, thank you, whoever asked that question.

Because I actually rarely didn't ever point this out. So this

here is purple just like these are. So what that means is that

it needs to be scanned again. So that there's recognition;

probably is a new item. And so we just need to, you know, take

that takes about six seconds to scan an item and then make sure

that we know what it is and where it fits. But what's

special about this picture, right here, is the depth of

field. And so I know that but you know what, I never actually

talked about that. And I should because that's, that's really

important that we do have good depth of field. We don't have

X-Ray vision. We get that question a lot like you know,

can you count you know the number of things, are they in

his line? No, we can't. And if everything is, you know,

front-facing, you know, you don't know if it's one or 12,

you know behind it, but we do have that depth. So that kind of

shows that off that capability off there.

It looks like the it's kinda looks like the

invader product to the left of it, though, doesn't it? Or is it

not? Is it a completely different product?

I can't tell from the picture does look similar?

Absolutely. So maybe, but it could be gluten free. That's the

crazy thing. Like there's this little sugar free, gluten free,

you know, whatever, that at a glance, or even in these

pictures, certainly, even if you're working there, and it's

only two feet away, you really have to be like - it has lemon?

You know, it's hard to tell. But with computer vision, there's no

there's no bias, it just says, Hey, this is not identical. And

so you know, let's figure out, you know what that difference

is.

And by the way, I know the audience knows this,

but computer vision is great. But it has to have some kind of

source, source of truth to compare it to. It either has to

have a very accurate product image library, or things like

that. And when you go into shelf stores, some shores, especially

on a weekend, half of the bottles are knocked over, the

other ones are on their side, everything's pushed to the back,

it's really, really tough to be able to run this. So there is

some operational discipline to make sure you're facing the

product, it's correctly positions not knocked over, it's

not turned on its side. Because if you think it's hard to read

the front of the box, start turning it 360 degrees, it

becomes much more of a challenge, right. So there's

operational things that go with that.

That's right. And to add to that, we have one of

our very large CPGs that is, is not a cereal provider. And and

they in fact said that; if you give us a demo on cereal boxes,

just don't bother coming because that is the easiest, that is the

easiest from an AI to detect those really nice, you know,

rectangle. And when you get into, to your point, twist a

product or specifically with with chips or with candy or

confectionery, where there's glare on the packaging, or it's

all squished. They, they, they were quite pleased to see that

that's something that we could demo real time in front of them.

And that's the exciting, that's the exciting thing about AI is

it continues to improve every day, every week. So the you

know, the continued challenges are really fun to, to hit up

against and succeed.

Well, by definition, the algorithms get smarter and

smarter and smarter. But there is a requirement of a quality

image on the shelf, ie the realogram, and quality image,

which is the product, you know, product or library, if you will.

Because if you don't have that, it's really tough to run an AI

platform on it.

That is true. That said, we also have the

ability to do a reverse planogram. So we, that is that

is new. And so CPGs - and some retailers - because some

retailers have recommendograms, or they might not have

planograms at all. For whatever reason. And so what's great with

with our solution is we're able to reverse create, and then

export out. So then at least you've got, hey, if you want to

start with what the planogram is right here, you know, from a

historical reference, and then you can and then you could see

you could even, you know, see it an aisle and this is the button

actually right here. And literally just click on it and

in seconds, that reverse planogram appears. So it allows

you then you know for for this, it allows you to be able to

walk, you know any category in any store where we are to be

able to then see it for yourself and then help make

recommendations. It also saves in travel costs because there's

a lot of money spent on district managers and you know up the

chain flying and driving in, you know, trying to look at what's

going on. Well now you can be that armchair store manager and

having this virtual visit you can really understand and then

you can make a phone call to that store manager and say, hey,

that's actually should b in the candy category. You know,

Houston, we've got had a problem or if there is, you know,

questions just you know what what's going on right now there

was a you know - the Nuggets, we're winning the Finals and so

there was a lot of excitement about frozen pizza or candy or

whatever, because we're gonna win.

Chicken nuggets! Chicken nuggets for the Denver

Nuggets! I'm reaching here, I don't know.

Exactly.

So question for you. So Barbara, some retailers, not

all of them some retailers, actually when they stock they'll

open up a box, they'll put as much as they can on the shelf,

what doesn't fit, it actually goes on a top riser. I know

Walmart calls that top stock. Is one of the future applications

able to look at what's on the top stock and let an associate

know that hey, it wouldn't fit before but it could fit now. Is

that one of the use cases you're exploring?

Yes, absolutely. Yeah. Because it's not only

Walmart, but it's several, top stock, using that vertical

space. Yeah, absolutely.

Well, I hope you enjoyed that conversation with

Barbara Chase. Please join us next time as we continue part

two of this conversation regarding on shelf availability

and using computer vision in retail