#064: Analog (In-Store) Analytics with Gary Angel

Back in the day, we explained the difference between a visitor, a visit, and a pageview to stakeholders using an analogy of a person walking into a physical store. Now, digital channels are dominating, and physical stores are struggling…which is an opportunity to apply what we’ve learned about behavioral analysis on the web to in-(REAL)-store consumer behavior. Gary Angel from Digital Mortar (@digitalmortar) returned to the show (our first ever repeat guest!) to walk us through the many, many similarities, as well as to explain some of the unique challenges and opportunities of in-store analytics.

People, Events, Products and Posts Referenced

 

Episode Transcript

Transcript

[music]

00:04 Announcer: Welcome to the Digital Analytics Power Hour. Tim, Michael, and the occasional guest, discussing digital analytics issues of the day. Find them on Facebook at facebook.com/analyticshour, and their website, analyticshour.io. And now, the Digital Analytics Power Hour.

[music]

00:28 Michael Helbling: Hi, everyone. Welcome to the Digital Analytics Power Hour. This is Episode 64. You know, when I was a much younger person, I worked for a while as a Starbucks barista. It was at a time that when we used to correct people when they ordered a large coffee. “It’s a venti.” In our small store, there was a very specific idea about what each person should be doing at each station, at what time. And there was a dance and a rhythm of getting people coffee in that morning rush. And what became really obvious to me over time, is that somebody somewhere had spent a lot of time thinking about the flow of the store, and who should do what, and when. In both the digital and the retail world, great experiences don’t happen by accident. And most of us take in retail concepts, and bridge them in some way over to the digital world that we inhabit today. But on this show, we’re talking about flipping the script, and taking some digital ideas, specifically some digital analytics ideas, back into retail. Our guest is none other than Gary Angel. And Tim, you and I have been on the show a bunch of times. Welcome back.

01:41 Tim Wilson: Almost every one.

01:42 MH: Almost every single one. And on a somewhat interesting note for this show, this is our first ever return guest on the show. You might remember, he also joined us to talk about digital transformation. He’s the founder and CEO of Digital Mortar. Prior to that, he was the leader of the digital analytics practice at E and Y. And prior to that, he was the founder of Symphonic. Welcome back to the show, Gary.

02:08 Gary Angel: Hey guys. Thanks a lot. Great to be here.

02:10 MH: So, Gary, I think, first off, I’d love for everybody to understand the idea behind Digital Mortar, and what got you headed that direction to start this company up.

02:22 GA: Well, I think from my perspective, a lot of it, like most businesses, happened pretty accidentally. It really came about in early 2016, we had a client at Ernst and Young that we’d been engaged with for a number of years, was one of our larger clients. We’d done a lot of e-commerce and direct-to-consumer work for them. But some of our stakeholders had actually expanded their responsibilities a little bit, and were starting to do analytics for their own stores. And every once in a while, they’d run across a problem that they didn’t feel like they were getting enough support, or didn’t fully understand, and they’d bring us in to take a look at it. Which was interesting. They knew we weren’t experts on the store side of things. We’re digital analytics folks. But sometimes the analytics is similar, or sometimes you just trust people, and you hope they can do some interesting work or help you out of a problem. And really what happened is, these guys had gone out and hired one of the existing vendors in this space of tracking the way customers actually behave in stores.

03:23 GA: And what they’d done was, they’d wired up a store, they’d put a whole bunch of cameras into the ceilings, they had this whole array of like 40 cameras tracking the way people move through the store, tracking everything. How they moved, where they went, how long they spent, and they had a system, a software system that actually allowed them to take that data and do analytics with it. And so they’d done a pretty substantial pilot project around this, but they were struggling a little bit to really get interesting stuff out of it. And these were guys who were pretty sophisticated from an analytics perspective. They came out of the digital world, they understood how analytics worked, they understood how behavioral analytics worked, and yet they were looking at this data, and really finding it hard to, one, make sense of it, and two, use it to actually derive value. So they asked us to take a shot at it. And I think, for me, it was kind of revelatory.

04:13 GA: On the one hand, it was like, “Wow, I had really no idea that you could collect that kind of data.” Didn’t know it existed. I mean, obviously, I think we all get used to the fact that when you walk into a store, there are security cameras, you figure people are tracking in some way or another. But I didn’t really realize that that kind of data source, about the way customers were behaving in stores, existed. I thought that was super exciting. I also thought that it was surprisingly analogous to digital analytics. As I looked at the data, one of the things that really struck me was that, as a journey, if you broke it down into individual components, and you mapped it to the store, and you said, “Well, customer came in the door, they went to this place first,” that’s like a page view in this place, right? And an entry. And then they spend a certain amount of time there, and then maybe they went to another place in the store. That’s that journey and path analysis across the store. And then maybe they converted, or maybe they abandoned and they just left the store, right?

05:10 GA: It felt like there were digital analogues to all of those things. So from that perspective, it was like, “I get this, I get this data, I get how it can be used, it feels very comfortable to me, and super interesting, because it’s about something different than a website.” At the same time, I was really unimpressed with what the vendor was delivering, in terms of the reporting, the analytics, and the data quality. I say this a lot, but it’s totally true. It felt like going back to digital analytics, really web analytics, like 1998. It was a really crummy interface, the data had all sorts of problems with it, it was just laden with data quality issues. It was all really aggregated data, you couldn’t get a low-level data feed out of it, if you wanted to pop it into Tableau or do some real analysis with something. So, it was limiting in a lot of ways. And as we got into it, we realized, “Well, it’s no wonder that our client was struggling to get information out of it,” it really was hard.

06:05 GA: And the system made it a lot harder than it needed to. That felt like an opportunity to me. It was one of those things where looking at it, it was like, “What I know, and what I know how to do, is totally applicable here. At least, I think so. The technology is mature enough to actually be interesting.” But nobody in the space, it felt like, really understood how to put those two things together to actually derive value in an organization. And, hey, we all know how hard that is. It’s not like that’s a trivial endeavor, right? We’ve seen over the last 18, 20 years, tools like Adobe Analytics and Google Analytics go from being awful, to okay, to really darn good. But that’s a long journey, and I think from my perspective, there was an opportunity here to short-circuit that. We knew how to do it better ’cause we’d seen people who’d done it better over a period of years, and I kinda felt like, hey, I could take that, and apply it to this business. So that’s really how it came about.

06:58 TW: Nice. So obviously, right there when you start talking about the technologies behind it, immediately to mind comes the Amazon Go store idea concept. It’s how maybe future retail will be this transaction-less kind of thing. Is that… Do you see it going there and where do you see Digital Mortar intersecting with that idea, or future state?

07:26 GA: Yeah that’s a big question, and I, heck if I know, right? In one sense. I’ll say this, yes. I do think stores are gonna go in that direction. A couple of things are pretty obvious. One is, this is a massively challenging time for physical retail. I saw an article, I think, last week, that store closures for 2017 are going to be at a historic all time high. Higher than the highest period of the Great Recession, and by a significant chunk, too. So retail is struggling. That’s first fact, is that there’s massive disruption right now, and I think because of that, people have to change, they recognize that. I’m writing a blog right now, and the title of it is, “Change or Die.” And I think that’s right. This is a time where if you don’t change, you will die. And that’s harsh, that’s unusual, that’s not typically the way things are in the world out there.

08:20 GA: The thing about change is, when you do a lot of changes, when you’re trying to do new stuff, and when you’re trying to do fundamentally different things in the store, it’s hard. Change is really hard for everybody. It’s hard for us as people, it’s hard for organizations, it’s hard to get right. You mentioned, I’d actually kinda forgotten what my topic was, it’s been a long time since I was last on the show, but, we talked about digital transformation, and just how hard it is for companies to get good at digital. Well, with stores, hey, stores are pretty good at what they do in many respects, but now they’re being asked to do fundamentally different things. Used to be the store was all about selling people stuff, but nowadays if you want to be successful, the trick is you gotta make people wanna go to the store. ‘Cause they can buy the stuff online just as easily, right? So you’ve gotta deliver a good experience out of that.

09:06 GA: Delivering an experience is, in a lot of ways, a new kind of business for stores. What I see is what we do is we provide the measurement infrastructure that brings what the customer is doing to the table, and lets you understand as you make all those changes, be they little changes, big changes, the way you staff. I thought your Starbucks example was great. If you’re adding the digital experience to the store, if you’re adding a climbing wall to the store, if you’re trying to do clienteling, where your associates are taking a different role with people. All of those things are really interesting and important changes to the way the store works. Nobody has any way of measuring any of that stuff right now. Nobody has any way to really measure anything other what you sold. The experience goes largely unmeasured. I see our system as basically being that glue that brings together all the experiences in the store whether they’re digital experiences, or associate experiences, or just the customer browsing and looking at merchandising.

10:03 GA: And lets the analysts, and the guys who plan the store, and the guys who do the promotions really understand how that complex ecosystem is working. What’s working and what isn’t. And as they change things, is it getting better or worse? So I do see us as that glue that brings those things together, and I do think that stores are gonna go more and more like Amazon Go, but I think it’s gonna be a whole range of things too. I think you’re gonna see stores that are super hands-on in terms of associate interactions, and it’s all about talking to people, right? And I think you’re gonna see stores that are really deeply integrated with digital experiences. I don’t know you are familiar with what Oak Labs does. But a wall that goes in the changing room that interacts with you, and allows you to ask the sales associate to bring you stuff. Like, if you’re in there trying on jeans, and the jean kinda fits but you don’t like it, you don’t like the exact cut when you see it, you just press a button, and they’ll bring you a different cut of the jeans. That kind of experiences is, I think, the future of retail across almost every kind of store. People are going to find different niches, but I think everybody is going to have to change significantly from what they do right now.

11:08 TW: How far are we out from the grand debate of whether time in store is a good metric or a bad metric? [laughter]

11:17 GA: Absolutely. Right? It’s funny. One thing I realized, as I got into this space is, nobody’s ever done this kind of measurement. There’s a few companies, small, that are struggling along to do it. We’re now one of them. A small, struggling along to do it. But there’s no real vocabulary in the industry, and that’s interesting. And one of the things that we’re actually trying to do is name stuff. Like, what does it mean when someone actually spends time in a place in the store? What do you call that? Is that a dwell, it that a linger, is that engagement? What’s the vocabulary for that? And you’re right. We’re gonna have all those same debates, right? But it’s funny. I’ve been so influenced by the digital paradigm, that when I brought over, and we built the platform we named a lot of the metrics after things that happen in the digital world.

12:03 GA: Some things are different and we had to make them up, but we have things like bounces. If you go into a store, and you go to one section of the store, you linger there a little bit, and then you go, that’s a bounce, and we just ported that vocabulary over. So, I gotta say we stole all sorts of vocabulary from the digital space. And part of that is, I genuinely believe that it’s interesting and useful, and it’s a little more mature. But heaven knows if any of it’ll stick. It’s still a space where the measurement is so new, not only are they gonna have those debates, but no one even knows the language about how to talk about this yet.

12:35 TW: My take on the web analytics, I think it’s still a challenge we’re dealing with in digital analytics is that, sometimes it is, the technology is the easy part. The data capture. And that tends to lead the thinking about, “What are we going to do with this?” And even as you’re talking about… If you’re fundamentally measuring behavior in the store, but part of what you’re talking about seems like it’s, “Are they going to come back? What was their attitude? What was their enthusiasm?” I guess there are two things. One, do you foresee, or have you already run into, we’re trying to measure if this experience, if this store works better, but we aren’t getting the people in? Like, it’s chicken and egg. When I’ve had people, “We wanna know what people who aren’t coming to our site are doing on the site, or want on the site,” as well as… I’ve seen you talk about combining behavioral with attitudinal data for digital. Does that fit in it as well? Is there an exit survey, or observational stuff, are they smiling? Does that sort of stuff fit into it as well, or is that, I’m jumping the gun?

13:47 GA: To the last part, absolutely. It’s funny. I actually can’t teach an old dog new tricks, ’cause I am just stealing everything that I knew on the digital side. But stores actually have done a lot of attitudinal work. One of the core information sources they actually have is, they’ll go in and they’ll do some things that, on the one hand seemed wildly obsolete to me, right? They actually have people go in sometimes, and they’ll sit there, and they will chart how someone walks through the store. So they’re sitting there with a little chart of the store, actually drawing little pencil lines on it where somebody went. Well, that seems like an awful way to do behavioral analytics to me. That’s very cumbersome, very prone to error, and very difficult to analyze.

14:28 GA: The flip side, though, is they also do a fair amount of attitudinal research. And I do think that, just as in the digital realm, that pairing of people’s actual experience; what they say about it, what they thought about it, how they experience the brand, there’s real learnings to be had there, that really add color to anything you can get out of the behavioral side. The behavioral side has a lot of the same limitations that we see in the digital world. It’s largely anonymous. You don’t necessarily know whether people enjoyed the experience or not. So a lot of times, you’re looking at what they did. You can see whether they were successful or not, but you’re guessing statistically at whether what happened really was causal or not. And the attitudinal kind of analytics can add a lot of that. So that’s definitely there, and I think a really important part of what people are doing. That’s something that we’re also trying to bring to the table. I think the integration of those two things, well, heck. In our space, people are barely doing the behavioral at all. So the voice of customer out there, that’s fairly well established. Part of we wanna do is let people know that when you do this behavioral stuff, you can still that use attitudinal, and it actually is really valuable. That’s a lot more established on the store side than behavioral analytics is.

15:41 TW: Yeah, it makes sense, and the idea of somebody standing there with the clicker just counting the number of people going past, and it’s…

[laughter]

15:50 GA: It’s terrible right?

15:51 TW: Yeah, it’s, “How may hits did we get today?”

15:55 GA: Exactly.

[laughter]

16:00 TW: Am I right in thinking that, at least for large big box retailers, that when they do… In this, I’ve had limited exposure. I remember Coles was a client years ago, and I remember going to… Where’s that, Menomonee Falls, Wisconsin, maybe? One of those places. And they said, “Oh,” it was in the area, they’re like, “Oh, we have our test stores or our pilot stores.” And that was where they would say, if we’re testing displays or new layouts or new treatments, and I guess it didn’t really occur to me then to actually say, “It’s great… ” Or even, I’m in Columbus, Ohio, or Dublin, Ohio, where Wendy’s is. And so, they’ll have their test stores where they’re trying new food items or even new layouts of the stores, or the QSRs, but what’s… How are those companies evaluating what’s working or not in those cases now? Or maybe, let me flip it another way. When you’re talking about doing in-store measurement, is that often gonna be at a handful of test stores, or is that often intended to be, no, at all your locations, so you get smaller changes you can make and see. That’s like six questions all jumbled together.

17:14 GA: Yeah. So I’ll take that last one, ’cause I think it’s hugely important. And frankly, I’m still trying to figure it out. I think when I looked at what people in this space were doing, it mostly broke out into two categories of analytics companies, and I’m probably over-dignifying them by calling them analytics companies. But on the one hand, there are door counting companies. People whose essential business is, just what we were just talking about, when you do that clicker, how many people came in the door? When you do door counting, that’s to generate a KPI for the organization, because what they’re doing is, hey, they know what they sold, right? So they know how many people they had at the cash register, and they know total goods sold, and then that door counting gives them the opportunity, they divide the two, and they get their store efficiency metric. It’s not a very good metric, but that’s what people have used. So door counting is something that if you do it, you put it in everywhere. Because it’s not a test thing, it’s a KPI thing.

18:05 GA: The flipside is, as companies got into the space that Digital Mortar’s really tackling, which is understanding how customers move through stores. Mostly, in fact, I shouldn’t say mostly. Almost completely, it’s been deployed on a single store, or a small store basis. So, maybe you deploy it in three or four stores, maybe you deploy it in as many as 10 stores, but you’re doing it on a test basis. And frankly, that’s what we’ve done, too. I’m not convinced that’s right. I think that, inevitably, with the technology starting to mature, one thing that’s a little different about this space compared to digital, is there’s more cost in the upfront. In digital, you wouldn’t put a tag on three pages of the website. The expense of putting the tag on is pretty much the same, regardless of what you do. But with a store, every store that you wire up, because you have to put some actual technology in the store, is an expense. And certainly, with big box stores, it’s a pretty significant expense. So that means people are reluctant to go out and do fleet-wide. And I totally get that.

19:05 GA: So, what they try to do is, they try to say, “Well, we can get learnings by looking at specific stores.” And I think that’s kind of fair for a certain class of problems. You know, if you’re looking at store layout issues, and if you’ve got six different layouts of stores, well, you have to at least wire up each of the different six layouts, right? Because if two stores are fundamentally different in layout, it does you no good to infer learnings from one to the other, based on that. So, you at least have to cover your layouts. The other thing I think is super important is at least doing some regional coverage, right? Because stores are really different, and a store in Wisconsin is not going to behave like a store in Manhattan, which is not going behave like a store in Miami or San Fransisco. So, I think some regionalisation is super important. Seasonality’s different, your customer base is different, the surroundings for the stores are different. So from that perspective, I think you have to go beyond one store, I think one store’s awful. But is the right strategy covering key stores by configuration and region? Maybe. I’d love to get to the place where it’s truly comprehensive. I think that’s really where measurement has value. I think that allows you to take that measurement, and not only understand what the broad trends are, but what’s different about each individual store. What really works differently?

20:23 GA: And I think the other thing is, there are aspects to the measurement. Things like, one of the things we try to do is not just measure customer behavior, but measure the staff and associate behaviors. Which teams are performing well, which people are good? If you train people, what kind of difference does it make? If a customer walks into a section, how long should you let ’em browse before you actually approach ’em? Well, the truth is, those things are going to different, probably, across almost every store. So there are cases where I think, depending on your business problems, sometimes you need to have comprehensive measurement. There are analytics problems you can tackle without it, most of what we’re doing is still very much in the proof of concept type stage. Where it’s, I really try to talk people out of doing one store, but a lot of what we’re doing is like four or five stores, up to about ten stores. I think that’s valuable, I’d still love to see… Not just from a business perspective, as a measurement guy, I’d ultimately like to see people do this everywhere.

21:17 TW: So, you’re saying we’re a little bit away from two-tiered segmentation in…

[laughter]

21:23 GA: Yeah, we are. Again, I’m trying to work those segmentation concepts into what we’re doing, and sometimes it has real applicability, but boy, this is a lot, lot more primitive than digital analytics right now.

21:38 TW: Yeah, but, I mean, if you think about it and just keep going down that path, eventually the application of both behavioral and potentially demographic data, at the point of interaction, is all stuff that could, in theory, be part of the store experience, down the road.

21:55 GA: Oh, no question. No, that that is really true, and I’d also say that there’s a lot of additional analytics that we never had to think about from a digital perspective that are in play here. I mean, I think the associate interaction I talked about is part of that, right? That whole human element adds a whole bunch of different data points, and different kinds of data points that we’re used to dealing with, that we’re still learning about. But it isn’t just there, things like weather are much more impactful in the online world than they are in the digital world. The music that’s played in the store right? We never had to think about that, websites aren’t backgrounded with music by and large, right? But there’s music playing when you walk into stores, and what’s played actually makes a…

22:32 MH: Not since 2003, right?

22:34 GA: Exactly. We all did try that for a little bit, but the store’s a very complex ecosystem, with a lot of different factors going on with it. I think analytically, it’s super, super rich. And of course, the other thing… And this is a place I do get asked about a lot, partially because I come out of the digital space, but there’s tremendous interest in that omnichannel join. You know, can I take this in-store data, can I pair it up with the digital data? Well, sometimes you can, there’s still real challenges to doing that. But when you can, that’s really fascinating, because now, all of a sudden, when you do things like display remarketing, you actually know what someone looked at in the store, right? And so, things like show-rooming are a huge issue for retailers. They have a hard time capitalizing on that. But if you can actually understand what customers did in store and use that as part of the communication strategy, you’re likely to get a lot more value out of it. So, I think there’s incredible untapped opportunities in the data, but there’s still a lot of work to be done just to get it to the point where you can actually use it effectively.

23:33 TW: Yeah, and certainly, if our listeners read Kevin Hillstrom, who is on another aspect of the retail side. He goes on epic Twitter rants a lot, about how omnichannel is not so great, but actually, digital needs to help drive to retail. And like what you just said, Gary, is, “Well, how do we join that data so that digital is both informing retail and retail is informing digital, so that you can take your best customers, and give them a reason to be there in the digital… In the retail store?” So it’s a very interesting shift of paradigm, and it comes back to what you said at the beginning of the show, which is the news about store closings and bankruptcies in retail, is just at epidemic proportions right now. And it’s kinda crazy, and I think there’s… Obviously, there’s a ton of factors, and I can’t pretend to know all of them, but not all of it is that people aren’t shopping, but it’s certainly that it’s so easy just… I can Amazon Prime now most of the things I would go into a store to buy, and, of course, I am probably not the person, ’cause all the analytics you’ll collect on me in retail is, “Leave that guy alone, he knows what he wants. You’ll just make it worse if you try to talk to him.”

[laughter]

24:51 GA: I’m kinda in that segment too, I gotta say, yeah.

24:55 TW: But wouldn’t it be great if the store knew that about me? I would be okay with that. And then, if I sort of give you the, “Hi sign,” then come over. I’ve looked on every aisle and I can’t find what I’m looking for, now I’ll go to my last resort, which is… And it’s like the navigation versus on-site search, right? The, “I’ll find it,” versus, “Go ask somebody in the store,” and people self-select into one of those two navigational approaches.

25:23 GA: I’m totally on board with that. From my perspective, I’m like you, I’m one of those guys where, by and large, if I walk into a big box electronics store, I know what I’m shopping for, and I probably know more about it than the person who’s walking the aisles. And I don’t wanna hear them, I don’t wanna… And, you know, I’m kinda shy too, I don’t even wanna talk to them, I don’t wanna have them come up and pressure me to buy something. So it actually keeps me out of those stores, right? And yet, those same techniques can be really effective with a different class of customer, and that ability to segment those populations, and understand which is which, that could legitimately create a really much better store experience for everybody, where the people who want the staff attention are getting it, and they’re getting more of it, because the staff are freed up to concentrate on them, and the people who wanna be left alone are getting a better store experience for them too, and I gotta believe the store as a whole is gonna benefit from that. So no, I think it’s very similar, and I think that one nice thing about this is that in some respects, there’s actually more data in-store than there is from a digital perspective. It’s a little harder to glean, and it’s a little harder to use, but it is a really rich data environment.

26:27 TW: People in the store, they can’t clone themselves. The equivalent of opening up multiple tabs on one website, and totally bastardizing their session.

[laughter]

26:35 GA: That’s right. Maybe we’ll figure out a way to do that in-store, but I don’t think so, that feels like it’s a long way away, right? [laughter]

26:44 MH: Some path analysis will work.

26:46 TW: Yeah. Well, it’s… There are few more finite paths. And I do, I laughed at… I mean, I didn’t create it, I think a lot of us use the explaining the difference between a visit and a page view with the bell on the front of the door, so it’s come full circle. I remember ten years ago trying to explain digital metrics in terms of physical store behavior.

27:08 GA: Yeah, it has come full circle. And I gotta say, that’s something… Obviously, hey, most of our audience are probably digital folks, but you know what? I look back on that, and I legitimately think we have a lot to be proud of in the digital analytics community. We took a discipline that was incredibly immature, had no vocabulary, no understanding of how it worked, it was new and really different. And over the course of a pretty substantial amount of time, turned it into something pretty mature, where people are actually doing consistently interesting work that actually does make experiences better, and more profitable.

27:43 GA: And I think that’s pretty great, and I look at the store side of things. And it’s not like they don’t know what they’re doing. People who run stores, they’re sophisticated, they’ve had a lot of experience, they have that hands on part of it, but they’ve never had the analytics and the measurement to do this kind of stuff. There’s a big missing piece out there, and it’s refreshing that digital’s the sophisticated guy now, and we can actually be proud of that, I think, and bring this stuff over, and say, “Hey, you’ve got some stuff to learn here.” Because we’ve actually spent a lot of time figuring out how to think about a website as a funnel, and how to drive people through it, how to segment them, and how to personalize that experience, and there’s just a heck of a lot of learnings there, if we look back on it, we’ve actually done some pretty interesting stuff.

28:26 TW: So how do you start? Like you’ve said a lot of things. You’ve got associate interactions and associate training, you’ve got the layout of the store, you’ve got placement of products, you’ve got the music. So, what is the… What’s the approach, if you’ve got a prospect or a client who says, “I wanna do in-store analytics.” How do you figure out which aspect of it to chase? And, I guess, on one front, does it actually… Do you have to figure out which technological solution to get the data capture, based on what their budget is and what problem they’re trying to solve? How do you approach it? If they say, “We can’t train our associates,” you don’t need to go trace that, or, “We can’t retrain them.” Where do you start?

29:15 GA: Yeah. Well, I’ll tell you where I start, and I think it’s a pretty plausible place to start. I think the core of it is the customer journey. That the core technology, from my perspective, is how the customer moves through the store, that’s fundamental. Now, you gotta keep in mind that there’s at least one other data source here that’s equally fundamental, but retailers have already got it, and know how to use it, and that’s Point of Sale. That’s what they sold, right? And in digital, it’s not that dissimilar, right? In one sense, cart checkout was often handled by back office systems, so you might not have a digital analytics system, but you might have known what you sold on your website, right? Digital analytics gave you all the color, all the other picture, how many people we lost.

30:00 GA: Who bought, what their experience and what their journey was like. The core technologies here start with the customer, and that’s what they give you. You don’t get music, you don’t get weather, you don’t get associate tracking. What you get is the customer journey through the store. I think that is a good place to start. I’m not sure it’s enough, and there’s certainly questions you can’t answer if that’s what you have, but I think that’s the core data source, and that’s usually what we start with. To get that core customer journey, there’s really, even there, there’s core technological choices you have to make. There’s really maybe three or four different ways that people have hit upon to do this, from a technology perspective. The big ones are camera, which is pretty easy to understand, I think. We all understand how cameras work. I think what I’d say about the cameras is, what they do is, they essentially just detect that there’s a human being in their field of vision. Blob it, and then track that blob as it goes across their field of vision. And then they just report that data source up to the cloud.

31:00 GA: A second way to do this is with WiFi access points. So, interestingly, if you walk around with a phone, certainly if you connect to a store’s WiFi, you’re issuing a constant series of electronic requests to the WiFi that allow them to geo-locate you within the store. It’s not super accurate, and there are issues with it. And there are issues particularly for people who don’t connect to the WiFi. One of the things that changed, maybe about a year and a half ago, is that Apple changed the way their phones worked. It used to be that every 15 seconds or so, an iPhone would ping out with its MAC address, saying, “Hey, I’m looking for a WiFi network.” And what people did, and especially what all the door counting companies did was, they actually used those pings that said, “Hey, I’m here, and here’s my MAC address,” to track phones. Well, Apple felt like that was PII information, and they changed the iOS so that basically, every time it does one of those pings, it randomizes the MAC address. What that meant was that all of those solutions basically went out the door. They don’t work anymore. It’s funny, ’cause a lot of them still provide measurement, but what they do is they just count all the Android phones and then they multiply it by some amount [laughter] and say that’s the total number of people who came in the door. That’s a terrible system, right?

32:15 TW: Perfect for any luxury goods.

[chuckle]

32:18 GA: Oh yeah, great, right? You’re totally dependent on that population staying fixed for all eternity, for your measurement. Not a very good kind of solution, but then people came along with these network sniffer devices, that look across a whole bunch of bandwidths, do a much better job of fingerprinting iOS phones. Those work significantly better. And people also use, obviously, things like their mobile application. If you’re in your mobile application, you’ve permissioned that, that can issue out pings, and you can positionally track people that way. So there’s actually a range of different technologies for doing just the physical tracking of the customer. And they all have drawbacks. None of them are perfect. I think they range in cost from very inexpensive, to quite expensive, and they all have data quality, and in some ways, data collection issues. One of the big things we had to do when we first sat down in this space, obviously we’re not engineers, we’re analysts, right? We looked at every frigging hardware solution out there. I looked at… There was a guy out there, and it’s actually a pretty interesting solution, but they have these pressure sensors they put in floors that they use for door counting. So as you walk through the entrance way, the pressure sensors pick up how many people are walking across the floor, right?

33:30 GA: It didn’t really meet our needs, ’cause you couldn’t really track people’s journey over time, but there’s a whole bunch of technologies out there. As we got into ’em, boy, I just found that everything had challenges, and we’re still living with that. We’ve basically decided on a particular hardware configuration that we think works the best. But even that changes sometimes from client to client. Really large footprint clients, we might do it a little bit differently than a small footprint client. Sports arenas, we might do differently from retail stores. There are reasons why the technologies are more applicable to one kind of space than another. And it’s still just a big giant headache. I gotta tell you, If I could do this business without the freaking hardware, I’d be delighted. But we just couldn’t figure out any way to do that.

34:11 TW: Do they actually track… Put stuff on their shopping cart? So that anybody who gets a shopping cart, they have very…

34:18 GA: Absolutely. You know, in grocery, that has been done. And again, some kinds of shops are much better for that than others. If I walk into a Costco, chances are, I’m gonna get a cart. If I walk into a grocery, chances are I’m gonna get a cart. But, you know, if I walk into an apparel store, that ain’t gonna happen, there’s just zero chance. Yeah, it’s actually way easier to track the cart. You can put a device in the cart, it’s your own property, so that is actually a really excellent way to do the tracking. But it only applies to a limited set of stores, which just goes back to the fact that, right now, unlike, say, digital, where… We saw the evolution from web-logs to tagging, but at most, there were really only two technologies, and pretty much everybody settled on one. Maybe that’ll happen in this space, but right now, it’s much more complex than that, with a lot of different competing technologies, none of which are really all that great. And there’s absolutely no standardization. And worse, there are real reasons why people with different footprints and different business questions might pick different technologies. So you can’t just ignore the technology on this one. You really have to pay attention to it if you’re gonna get it right.

35:27 MH: Well, when the arguments do start to happen in earnest, can you invite us to the Yahoo! Forums so that we can laugh at all the people who are arguing about it?

[laughter]

35:37 GA: Yeah, you don’t wanna re-live the good old days of arguing about that stuff?

35:39 MH: That’s right. Like, “Oh, I remember when… ” Yeah.

35:43 GA: That used to be, like, half of what we talked about, right?

35:46 MH: Yeah, sure. I remember, early in my career, putting together a PowerPoint decks showing, “Well, with log files you get this, and with JavaScript tags you get this,” and so…

35:55 GA: Yeah. Yeah, I don’t miss that at all, man.

35:57 MH: No, I don’t either.

36:00 GA: That’s the stuff you have to do to get to the point where you can actually do reasonable analytics, but man, this stuff is definitely still in that space.

36:06 MH: No, that’s awesome. If the technology is so unsettled, so, what’s driving… Obviously, cost is gonna be a huge driver, ’cause, to your point, getting it installed across an entire company’s footprint of stores, there are companies with 800, 1200 stores, the cost associated with deployment would be monstrous.

36:29 GA: Yup.

36:29 MH: So obviously, the cost is a huge piece, but then, what are the other drivers of that? Is it really just what can provide that best view of what’s actually happening with an individual at the point that they’re interacting, and…

36:42 GA: Yeah, there’s a couple of drivers that turn out to be really interesting, and I’ll walk through ’em. One is, that some systems count universally, and some don’t. So for instance, cameras pick up everybody, right? They pick you up whether you have a phone or not. They pick you up whether you’re a woman or man. They pick you up whether you’re an adult or a child. So, cameras are great for getting that universal count. And that can be very important.

37:08 TW: Except for vampires. You’re gonna miss out your vampires.

[laughter]

37:12 GA: Yeah, okay! And who knows how big a segment that is in some places?

37:16 TW: Well, it’s much bigger for you out there near San Francisco, of course.

37:20 GA: Yeah, yeah, absolutely.

[laughter]

37:24 GA: Okay. I take that back. But cameras by and large get a better sample than most technologies.

[laughter]

37:32 GA: If you’re using any kind of electronic sniffer, you’re largely dependent on people, A, having the device, which is pretty ubiquitous, but B, also having things like WiFi or Bluetooth turned on. They gotta have some kind of electronic signal turned on. So that’s important. The size of the sample you’re getting, and how representative that sample is. If you’re using something like WiFi access, and you’re only measuring people who opt into your WiFi, you’ve really reduced the sample of people that you’re measuring a lot. And you also have to ask yourselves, “Is that sample even remotely representative of my overall shoppers?” So there really are issues around that.

38:04 GA: Second big thing is positional accuracy. Again, cameras have some advantages when it comes to positional accuracy. When a camera takes your picture, you can pretty much position that person to that spot in the store. With electronic signals, basically what they’re doing is they’re triangulating on the signal from multiple collection points. And the more collection points you have, the better that triangulation is, and the better the positional accuracy. A lot of people who do this in a bad way, they have, like… They just use the store’s WiFi access points, maybe there’s two of them. The triangulation’s awful. And it gets you down to something like a 10 or 15 meter circle of air. Well, you think about, a 10 meter circle of air, you could be anywhere in the frigging store, right? You could be in totally different departments, which doesn’t make for very good journey tracking, right? So you wanna get that positional accuracy down to something like a meter, a meter and a half, something in that range, where you basically know where the person was in the store and that’s super important.

39:00 GA: So, so far it sounds like camera would be the ideal technology. It turns out, cameras have a huge issue too, though. One of the things about cameras is, that the way they’ve been architected is they have a real challenge actually following people from camera zone to camera zone. So, most cameras cover a zone of about 20 by 20 feet, and that’s all they can do with accuracy. So if you’re gonna do even a typical mall store, you’re gonna have to put a lot of cameras in. And God help you if you’re gonna do a big box store, it’s gonna be a lot of cameras. The problem is, that these camera systems, they don’t actually do face recognition in the cloud. The way they work is, as I mentioned, they blob the individual, right, when they’re following them.

39:38 GA: Well, if you think about that, what that means is as you walk from one camera zone to another, the other camera picks you up, blobs you as an individual, and starts tracking you, but there’s no way to marry those two blobs! So what happens is, with camera systems, you often can’t answer even really simple, basic questions like, “Hey, of all the people who went to jeans in the store, what else did they do?” Or, “The people who went to this section of the store, how many of them checked out?” Or, even if you have… One great case for us, we were working with a store that had a long wall of product that was about 60 feet long. So it actually had three cameras to cover the zones. One camera said we had 180 visitors in an hour, another said 240, and another said 200. Now your guess is as good as mine whether that meant they had about 600 visitors, or whether they had 240 visitors who crossed the whole thing, right? There was absolutely no way to reconcile that. So that ability to track the journey is something that electronic devices actually do really well. They pick up the journey, and they track the whole journey seamlessly. For a lot of the camera systems out there, it’s fundamentally broken.

40:46 GA: So bottom line, every one of these collection technologies has some issues. We actually like to deploy a mix of them. We like to do camera on entry, so we get a 100% count of everybody who came in. We get the basic demographics, things like gender and age. But then we like to tie that up to an electronic signal so we can follow the rest of the journey, and we like to put enough devices out there so we can do it with a high degree of accuracy. That’s also, by the way, cheaper. ‘Cause cameras are, by the way, the most expensive part of this whole thing. But is that the way every measurement system is gonna look? Or is that what we’ll be doing two years from now? I doubt it. I think this is still something that’s in a high state of evolution. And I think for us, we spent a lot of time trying to figure out what made sense right now. But it would not shock me at all if two years from now it was totally different. And really, our objective mostly was to architect a system that could take feeds off a lot of these different technologies, and still work pretty seamlessly and well. I expect to have to use that capability routinely as we go forward. [chuckle]

41:48 MH: Yeah. So what you’re saying is, if somebody out there can invent a technology to reacquire the same customer as they crossover cameras, that would be a killer app.

41:56 GA: Doesn’t seem that hard, does it?

42:00 MH: Well, I don’t know.

42:00 GA: You know Homeland Security has this, right?

42:03 MH: Yeah.

[laughter]

42:06 GA: You know, I think part of the problem here is not that… I think we all know that things like facial recognition have gotten way better in the last couple of years. And the analytics exist. The problem is that the way these camera systems were architected doesn’t allow you to take advantage of it. ‘Cause they were architected with all the processing onboard the camera. And that camera has enough smarts to do the blobbing, but not enough smarts to do the facial recognition. And I think that’s a real issue, I think that will change. I mean, I’ll be shocked if… Give it a year, give it two years, if most of the systems out there aren’t able to do that cross-segment join much more seamlessly from a camera perspective. It’ll still be really, really expensive, that’s still gonna be a drawback to camera that might put you off it from a big box perspective. But I gotta believe they’re gonna solve that problem. Because it just seems so easy to solve, really, given today’s technologies.

42:56 MH: I can’t wait ’till there is, literally, a camera based DMP in the sky where you could figure out what other stores this person’s been in today.

43:05 Announcer: Yeah, yeah.

43:06 MH: That kind of stuff could happen.

43:07 GA: You know that’s gonna happen, right? I mean… And it’s funny too, because I never took a lot of the privacy stuff on the digital side that much to heart. And I shouldn’t say that. I mean, it’s obviously important, most of the brands we worked with were super sensitive about it. But I was always getting asked, “When does it turn creepy?” And I’m like, “When you actually start using the data.” The problem was not the creepiness…

[laughter]

43:30 GA: The problem was people’s fear of actually doing anything. We were so far away from being creepy in most cases that I didn’t feel like it was a very productive discussion. But it is true. I think that this infrastructure, actually, is a little bit creepier. I’m glad it’s anonymous, I think people do have to be careful with it. And I, at least personally, find it a little more unsettling than I ever did, cookies on the digital side of things. But I would be, again, surprised if we don’t see very similar kinds of evolution with, eventually, systems being able to track people. There’s even some of that now. The guys who do the electronic stuff, and collect things like MAC addresses, they can do some of that, of tracking you from store to store to store. And that is part of the benefit of that technology, and part of what they sell, is creating that kind of interface, where you actually get not only what you’re doing in-store, but what those customers were doing out. So we’re definitely gonna see the same kind of third-party markets, and the same kind of bartering of information. I personally… I guess I’m okay with that, in some respects. I do prefer, and I’ve always preferred the owned analytics, where it’s your store, and you’re just optimizing your store. And I’ve always been a little bit uncomfortable with third-party marketplaces trading around people’s information. I’d rather be on the, “Hey, we’re optimizing our store,” side of things.

44:44 TW: Yeah, thinking of the personalization and the targeting, if you get to the point where you’ve got the feedback loop, where it’s like, “Oh, if a guy comes into this… If a middle-aged guy comes in and turns right, then make sure the associates know not to approach. If it’s a middle-aged female who goes straight ahead, then we’ve got a tracker into the associate that starts a three minute timer and says, “Go over and approach this person.” And provides a personalized experience that, probably, brilliant Nordstrom associates had internalized with just their intuition and practice, but trying to have the data feeding that.

45:27 GA: Makes it better. And you’re right, I’m absolutely sure that a lot of associates and store managers pick up and internalize on this stuff. Just because that’s what we do all the time, right? But you also know that anytime you’ve got a lot of people involved with something, there’s the people who internalize it because they pay attention and they’re smart, and then there’s all the rest of us.

[laughter]

45:49 GA: A part of what technology does is it brings all the rest of us up to the level of the people who are actually good at this stuff, right? And I think that’s where you can really take advantage of it. I also think, and this is really interesting, a lot of people who’ve gone into Digital Mortar’s space, the same kind of thing, have evolved in the direction of focusing on delivering coupons to people as they walk through the store, so the idea is, “Hey, we can see what they’re looking at, we can then real-time pop a coupon to them.” And that’s a great application. It’s kind of interesting, but I’ll tell you, I see two problems with it. The big problem I think we’ve seen almost universally is opt-in rates for those kinds of programs have been so low that they just don’t make a difference to stores.

46:32 GA: And I think the second part of that, which drives the first part, is I don’t know about you guys, but I don’t want a bunch of coupons bombarding me while I walk through a store. You know I mean? There’s a segment of people who care about coupons, and then there’s the rest of us who don’t want to be annoyed from an experiential standpoint. I don’t like coupons. So I think from that perspective, there’s limitations. But I love what you just said about… The flip side of that, though, is if we could take that same kind of analytics and use it to steer the human side in the store, the associates to the right place, to the right customer, at the right time, you don’t have all those issues. You don’t have the opt-in issues, you don’t have the customer gets annoyed issues, you’re making good decisions about how to deploy the human capital you have in your business, that’s not only one of the most expensive components of a store, but really the part where you can get the most competitive advantage. That’s more of where I think we’re gonna be focused, when we get to things like real-time messaging. It’s not driving coupons out to folks as they walk through the store, it’s letting associates know what the service opportunities are, and who they should be talking to, and what the right conversation is.

47:38 MH: Oh, that’s cool.

47:39 TW: And those associates will be AIs, anyway. They’ll be AI robots, anyway, so…

[laughter]

47:45 GA: All the rest of us may get replaced with AI robots, but those associates in stores will, probably, still be people. That’s one of the funny things about that, is that some jobs that you might’ve thought were immune to AI, may not be. And some jobs really are, even though they don’t seem all that hard.

48:00 MH: Man. So this is such a great conversation, Gary. It’s so great having you on this show. Unfortunately, we have to move towards the wrap-up, but, yeah, but it’s going to be so interesting just to watch how this develops. So, thanks again for coming. One of the things which we do on the show, we do a thing called the Last Call. We’ll just go around and talk about something we’ve found that’s interesting that maybe our listeners wanna check out. I don’t know, Gary, if you’ve got anything that’s a Last Call, but I’ll let you kick it off.

48:30 GA: Well, two things I want to say, I guess, given our audience. One is, if you’re on the digital analytics side of this, and you’re working for a place that has physical locations, whether that’s a retail or a hospitality, or banks and brokerages, this is incredibly interesting data. And it’s not like anyone else in the organization either is doing anything with it, or knows what to do with it. So, I guess, one thing I’d say is, there is real point in this, right? I mean, this is an opportunity for digital analytics teams to expand their reach and derive additional value into the organization. And I don’t want people to forget that. I guess if you’re out there in digital, my message isn’t, “Hey, you should be coming over and doing this kind of stuff.” Maybe you should, but part of it is, within your organization right now, there may be a real opportunity to do this stuff, and it’s fascinating. I just think it’s the coolest stuff around.

49:17 GA: As part of that, I’m going to be doing workshops for all you guys who know e-metrics. I’m going to be doing a workshop on this stuff at the Chicago one. I think that’s around June 19th. That would be great if people have an interest. It’s a really in-depth workshop. It’s an all day thing, where we walk people through how this data gets sourced, all the technologies, doing the omnichannel joins. You actually get hands-on a little bit with some of tools. So that should be pretty cool. It’s actually the first time. I said it should be cool. It’s the first time I’m going to be doing it, so I’m hoping it’s cool.

49:45 GA: So, that’s one thing. And then, I guess the other thing, I was thinking, too, about books and articles I read. I’m a big reader, but I mostly read novels. But I did read a book recently that I thought that most analysts would enjoy, and it’s called Lab Girl. It’s basically about a woman who’s a botanist. She studies trees, by and large. It’s her history and memoir. It’s just a fascinating read. Part of it is just a personal journey as a scientist. Part of is, every other chapter is a drill down into some fascinating aspect of trees. And I actually thought this was so cool. Some of the facts and some of the science about how trees and plants actually work were just amazing to me. And I think almost any analyst would enjoy it. I enjoyed it a lot. I think any analyst would get a kick out of it. If you’re into science, that’s a book I’d highly recommend.

50:35 MH: Nice. That was a three-for. [chuckle] We’ll have Gary back on when he’s figured out in-store analytics, and he’s moved on to digital botany. [laughter]

50:44 TW: Digital botany.

50:45 MH: That’s right.

50:46 GA: But I was impressed, right? It was… Trees are surprisingly cool, was my take away from that. [laughter]

50:53 TW: Oh boy. You want to go next, Michael? I’ll have to make sure I don’t steal yours.

50:58 MH: So, I’ve got a couple. Well, obviously in light of this show, I’d be remiss if I didn’t talk about the new Amazon Echo Look, which they launched a little while back, and are… And, again, this is, Amazon is moving towards retail, a device that’s gonna tell you which outfit is the best outfit. That’s crazy. Anyways, that was just in the news recently, and I found it fascinating, it’s tangentially related. But actually what I wanted to talk about was, I read an article not too long ago, a while back, I don’t know, about a professor at the University of Michigan. And he wrote an article talking about how he banned his students from any projects that allowed exploration or free form exploration of data. He teaches a data visualization course, and he basically won’t let them submit projects where exploration of the data, they have to have an end in mind.

51:57 MH: And the article was profound, because it answered a lot of questions like, there’s all these things on the academic side that you have as intuitions over here in analytics, for those of us who aren’t very academically minded. Probably, Christopher Barry knows all about this. But that was really fascinating to me, really good article, and it cemented a number of things in my head, that was like, “Yeah, that’s why it’s really bad to just sit an analyst in front of an analytics dataset and be like, ‘Go find some insights.’”

52:27 GA: Have at it, yeah. Yeah. Not very productive.

52:30 MH: “What insights should I be trying to find?” It’s like, yeah, the promise of data is not that you suddenly get tons of access to it. And so, it’s really cool to see someone with an academic mindset explaining his philosophy behind it. And anyways, really good read. We’ll put a link to that in the show notes. What about you, Tim?

52:50 TW: I’ll follow the pattern and have a two for one that is related to the show topic. That was inspired. So, have you guys heard about the crime prevention robots that they’ve started to develop? So it is taking the role of the security guard in some locations. But really, specifically, I wanted to call out the drunken man who assaulted the robot, I think in late April. And it was a 300 pound K5 security bot that he pummeled, and apparently the robot did what it was supposed to do. It just fell to the ground and called for help. The quote in the one article I saw that was, “The robot has recuperated from his injuries, and is back on patrol, keeping our office and employees safe again.” [laughter] So that just cracked me up. My other Last Call is also not super deep, but it’s a podcast, Reply All. I don’t think I have brought it up here before, but it’s a couple of guys who have journalist backgrounds, and it’s a Gimlet Media podcast, which is a journalist-driven company with their podcasts. But they cover all sorts of stuff around the internet.

54:00 TW: And it can be amusing, it can be historically interesting, it can be informative. They wound up doing a couple of episodes around security, password security. I discovered the I’ve Been Pwned site, as they kinda tried to explore why the owner of the company had started getting Uber notifications for things happening in Russia. So it’s a good… They can be between twenty and forty minutes, and I think it’s every other week, but the Reply All podcast with PJ Vogt and Alex Goldman.

54:34 GA: Sounds cool.

54:35 MH: Nice. Well, as always, we can never get a show and all the things we want to talk about into less than about forty minutes. So, this has been a great conversation. Gary, thank you again for coming back, being our first return guest, which is, that’s a big deal for us. And obviously, as you’ve been listening out there, folks, you are thinking some things. Maybe you’ve got a brilliant idea that Gary absolutely needs to hear about to transform analytics for retail. So look them up on Twitter, or us on the Facebook page, or on the Measure Slack. We’d love to hear from you. I believe your Twitter is @digitalmortar. Is that right?

55:19 GA: That’s right, and I also just do @garyangel too.

55:21 MH: @garyangel. So either of those would be a great way to interact with Gary on Twitter. So anyway, thanks again, this is really fun. It’s sort of interesting how future and nostalgic this show actually was.

[laughter]

55:37 MH: And so, that’s pretty cool, anyways. For my co-host Tim Wilson, keep analyzing.

[music]

55:46 Announcer: Thanks for listening, and don’t forget to join the conversation on Facebook, Twitter, or Measure Slack group. We welcome your comments and questions. Visit us on the web at analyticshour.io, facebook.com/analyticshour, or @analyticshour on Twitter.

56:06 Charles Barkley: Smart guys want to fit in, so they’ve made up a term called analytics. Analytics don’t work.

56:14 MH: I saw a tweet where Charles Barkley was talking negatively about analytics again, and I was like, “Oh, is there audio of that? We need that for the show.” We go all over the place talking to anybody.

[laughter]

56:29 S?: I’m honored now, yeah.

56:34 TW: Wildly successful analytics talents, mostly people’s [56:38] ____.

56:40 MH: You can’t flake out on me. This is hard. Okay. We’ve got a show schedule. There’s a Tuesday morning coming, and we’ve got to put an episode on the air.

56:52 TW: I’ve gotta have somewhere to direct my anal retentiveness. It’s not at kind clients, I guess.

57:00 MH: Hence the name of our website for the show now: Analyticshour.io.

57:04 GA: Yes.

57:07 MH: But you know, io is ironic. So, it’s alright.

57:14 MH: People ended up on a gravelly beach in FEMA tents being served cheese sandwiches like it was… They were trapped on the island, they couldn’t get home. It is a tremendously beautiful story of big idea with absolutely zero thought or structure given to it.

57:35 GA: So that’s the podcast experience you’re promising people, is it?

[laughter]

57:40 MH: Come on down to the Digital Analytics Power Hour.

57:46 MH: [57:46] ____ flag and analog bounces.

 

2 Responses

Leave a Reply



This site uses Akismet to reduce spam. Learn how your comment data is processed.

Have an Idea for an Upcoming Episode?

Recent Episodes

#257: Analyst Use Cases for Generative AI

#257: Analyst Use Cases for Generative AI

https://media.blubrry.com/the_digital_analytics_power/traffic.libsyn.com/analyticshour/APH_-_Episode_257_-_Analytics_Use_Cases_for_Generative_AI.mp3Podcast: Download | EmbedSubscribe: RSSTweetShareShareEmail0 Shares