The life of an e-commerce manager has gotten much more complicated over the past 18-months. Between fixing compliance issues and uploading new content, it requires different skills and solutions from the insights industry.
Different solutions have been coming into the insights world to help organizations develop their e-commerce strategies, but these tools focus on text analysis and image size. There are no options for brands seeking to optimize the images on the product detail page (PDP).
In the latest podcast episode of Our Best Behavior, Adrian Sanger (Innovation Director, Verve) and Ruben Nazario (Vice President, Digital Shopper Innovation, Behaviorally) explain why Flash.PDP™, Behaviorally’s AI-powered tool for optimizing PDP Images, can help our clients excel at the digital shelf to drive shopper growth.
Hi, everyone. I’m your host Matt Salem, and you have tuned in to another episode of Our Best Behavior, a podcast brought to you by Behaviorally, winner of the 2020 market research podcast award.
Behaviorally, formerly PRS In Vivo USA, helps brands improve shopper and consumer experiences by defining and diagnosing the behaviors that drive shopper growth. Each month we produce a podcast to share industry insights on trending topics designed to help you make better shopper marketing decisions.
Today, we are joined by Adrian Sanger, an insights consultant and expert in developing insight solutions and go-to-market strategies for market research organizations. As well as Ruben is our VP of digital shopper innovation here at Behaviorally. Welcome to you both.
Glad to be here.
Thanks for inviting me.
pleasure, a pleasure. Thanks for coming on. Really appreciate it. Great to have you here. Adrian, great to have you, Ruben. Rubens on is like, you know, 11th show I’m losing count already. That’s right. That’s right,
Adrian, first time here with us, definitely excited to have you on board, I guess. To start, can you just tell the audience a little bit about yourself?
Yeah, I’ve spent my career in the marketing and market research area, in particular, at that intersection between insights and technology. And most recently, in the last few years been helping small and medium-sized, even large agencies kind of scale up, generate new growth through new products and services, something that I’ve done for many years and kind of corporate roles across dunnhumby and Nielsen and then GFK. And in recent years, I’ve been doing on my own
Excellent, and our loyal listeners might start piecing together the puzzle here that with Ruben on board and him being so involved in digital aspects of our digital-first organization that is Behaviorally that Adrian is partnering with us and helping us to deliver some of these solutions. And he’s been a great asset to us. And we’re thankful that he’s been willing to partner with us. And when we think about digital, and we think about what digital means Behaviorally.
Clearly, there’s a link there to E-commerce, and E-commerce is a high-growth area. I mean, the numbers, the charts, the everything you look at online, it’s just purely up and to the right. Every bar goes higher every year; every line continues to go up and to the right.
Let me hear a bit Adrian about what you’ve seen over the past couple of years in E-commerce and how different solutions have been coming into the insights world in order to help organizations further develop their E-commerce strategies?
Well, look, firstly, I got to agree with that. That’s absolutely right. E-commerce has been growing for some years across all sectors, not only CPG. And we’re seeing the growth of E-commerce as a channel long before the pandemic, but over the last 18 months or so, it’s just accelerated.
Some of the recent data I’ve been looking at from IRI suggests that in 2021, it’s going to be as high as around about 12% in the edible categories use online as a channel and as high as about 33 35% in the non-edible. And that’s a meteoric growth. It’s certainly for CPG manufacturers no longer a channel that we can kind of afford to think about as an afterthought. But it’s also been changing hugely in terms of its maturity, the way the whole of the channel is organized.
So many things are different in the life of an E-commerce manager, the way that they organize their time, the way they have to really adjust to daily, sometimes hourly changes, that require new content, fixing compliance issues, it’s dealing with literally hundreds of different content fixes, if you will, across their portfolio. And that’s a very different world from the world of bricks and mortar where, you know, there is slightly more a difference, let’s say to a plannergram and a calendar.
In the world of E-commerce, as many of your listeners will know, this is a daily changing event. And so, it requires a different set of skills from our clients and a different set of insights from the industry.
yea, it truly is a different world use that phrase a couple of times, and I wholeheartedly agree. I mean, a different world in terms of how shoppers’ shop, day in and day out, a different world in terms of how insights providers think about the shopping experience, a tool that we developed recently on the path is a tool that is meant to help bridge that gap between brick and mortar and E-commerce.
And then, going further, we recently developed a tool Flash.PDP, which is really homed in with laser focus on the E-commerce environment. And there’s just so much to keep up with. And part of what we wanted to do as Behaviorally is help our clients keep up. And I was hoping you could talk a bit to how teams keep up right now and where their focus lies. And where you think there can be a value add, that’s not currently addressed.
Well, look, there is a very different kind of set of maturities out there from some of the bigger organizations that have been, you know, working to develop their E-commerce tools, their processes, their staffing over the last few years, and others that, you know, just beginning to kind of get some specialism around equal, that’s a very, very diverse set of capabilities that we’ve got across the CPG world.
But pretty much all of them are trying to navigate that by kind of embracing systems that help that whole process of content management. And by that, I mean making sure that you’ve got the right digital assets, you know, with the right retailer in the right format, at the right time. If you think about that, for a moment, you know, the hundreds of different.com sites through which trade comes across many different markets and many different SKU. And then you multiply that by all of the content that’s required just to deliver for one SKU, you got a hero image, you’ve got some text, you’ve got some bullets that describe the product benefits, you’ve probably also got a carousel of different images that describe different parts of the product. And so, you’ve got an incredibly large amount of things that need to be maintained all the time.
I mean, what we want to do with within that is really build on some of the capabilities that are already out there and offered by kind of analytics platforms that are doing a really good job at highlighting some of the failings, we think that you know, as Behaviorally, there’s an opportunity here to, you know, really double dive on this whole area of kind of image quality. And perhaps say a little bit more about that in a moment. But that’s an area which perhaps hasn’t had quite as much focus from the E-commerce world as some of the other areas around compliance and fixing and keyword searching. And, and all of these other aspects that are essential for E-commerce success
And certainly, when you think about the world of compliance in E-commerce. I think text comes to mind a lot, right?
So, the text analytics and understanding how we could best position via the written word. But what you brought up and what we are extremely interested in is optimizing images online. And my understanding is that there are certain guidelines that retailers may have, I think I’ve heard it referred to as a plus guidelines where there are a number of suggested rules for images, but it still seems that it’s tied more so to either compliance or the cost of entry for driving conversion online rather than truly optimizing images so that they’re most successful for a given brand.
Yeah, well set Matt, I think the issue that out there is that the guidelines will give you a tick list of things that you need to do in order to make a kind of minimum basic level. You’ve got to have the product name, you’ve got to tell the reader how many units are in there, you’ve got to be clear about the variant. And you’ve got to have an image. So, you can almost go through a tick list.
Once you’ve done that, you are a compliant image, so to say for many of these platforms, but for sure, that doesn’t make it a good image. And that’s where we think the opportunity lies in trying to go beyond the more basic guidelines and try to provide more direction to our clients around whether that image is fully optimized. And if it isn’t, how do we go about optimizing it.
And with everything that’s there in terms of the text piece, why are images so important? What is it about images that really helps facilitate the shopping experience for those who are leveraging E-commerce platforms to shop?
And if we think about how people shop increasingly nowadays, whether that’s on a mobile or on a tablet, you know that they’re not giving, just as they don’t, when they approach a fixture, a huge amount of time to any individual skew, it’s got to make an immediate impression. And for most CPG brands, that immediate impression comes in the form of a little thumbnail image, a so-called hero image, and a few words of text.
And in those few words of text in the image, you got to do enough to be able to get people to click and learn more about production. And hopefully, bye, that’s a lot of requirements that have got to be satisfied. And the truth is that when you’re trying to do that for the number of skews and the number of images that doesn’t have And as perfectly as it could, you only need to scan across any website you choose.
While there may be some standardization in the way in which an image and text is laid out, the quality of those images, their clarity, their ability to be able to kind of instantly give you what you need, hey, I can see that a lemon variant, for example, is incredibly mixed. And this is an area where we believe we can provide much more guidance around what is good, what is deficient, which are the images that we need to pay attention to.
So, with that, let’s talk a little bit about Flash.PDP. Because the idea here from Behaviorally is to help in this regard with monitoring image content and optimizing image content to help facilitate that shopping experience. And on the one hand, folks may say, Oh, well, why can’t we just test as we normally would with Behaviorally, perhaps some sort of control test scenario and a B test of sorts. But really, that wasn’t an option in this case. And I think that ties to the sheer amount of content that would need to be monitored, online, and tested online. So would love to hear about the birth of Flash.PDP and why it is this different approach we’ll look over.
The first thing to say is that this is, again, intended to build on all of that testing, and monitoring, and optimization that is already going on across many different CPG companies, you know.
This is not to be thought of as instead of AV testing; it’s just trying to help to provide better insight around the images.
If you test two images, A and B. At the end, we don’t know whether those are two good images or two bad images. We just know that one performs slightly better than another. And it takes a while to make that learning, you know, typically, A/b tests, you know, are set to run for some time. They’re also done in a very public way. So, it isn’t always the ideal way to kind of get it right the first time and make it make an immediate impact on that shopper.
So, we wanted a different start point. I think we wanted to move away from a process which kind of interrupted that flow and went and asked a ton of people. A ton of questions are reported back on those results. To our mind, that just didn’t fit into the workflow of any comm team. We needed a totally different approach.
And so, we landed on the idea of creating more learning through kind of training a dataset to understand what good looks like in an individual category. If you understand well if we’re talking about, say, the cat food category, what is a good image, what is an image that customers are going to respond to when they see that and let’s run that through hundreds of thousands of times so that the algorithm gets really good at recognizing, which are good images, and which bad images for that category.
So that when we come to scanning many other images, hundreds of thousands of different images, also in the cat food category, it can provide a categorization of whether those other images are good or not so good. And that’s what we’re doing nearly thousands of times in the background in the product Flash.PDP.
So, you mentioned the example of cat food, of looking at not only the images of interest but clearly all of the other images, essentially, in quotes that are out there, we’re talking about 1000s, hundreds of 1000s of images. You mentioned the word algorithm. Something tells me that Rubin isn’t sitting there manually doing all of this. So, I’m guessing there’s an AI component involved. And I would love to hear how that’s wrapped into the Flash.PDP offering.
Yeah, it was absolutely driven by AI. And maybe Ruben will come in and just a moment to add to it. But yes, it is driven by AI. We’ve done the work in order to kind of better understand how the category performs.
It created a kind of learning data set, which then enables us to be able to do all of this work by reapplying that that algorithm to new images and using the data that we’ve already collected to provide a prediction on whether that image is going to create the right impact message in the right way, be persuasive, all of these kinds of things. So, it isn’t just a kind of pass-fail, but a diagnostic on what elements of that image are successful or not.
To provide a bit more guidance about, you know, if you’re going to fix that image, if you can optimize it, change it in some way. What do you need to do, and what we do on top of that is then run kind of visual analytics, which is become quite well established actually, in recent years to be able to predict, again, using AI, how an eye would likely scan that image? And what would it see first, or the heat map be? What parts of that image with this scene, and which would be missed in that all-important kind of two or three seconds? While you’re scanning that image for the first time?
Yeah, I’ll add that the AI that we’re using is cutting edge. It uses image recognition AI to detect individual characteristics and features within an image. So that’s how it can then tie it back to the images in our database and predict performance based on those patterns that it detects.
But it’s all an automated process. I am not sitting there, Matt, doing that one by one. That would be impossible. So, the AI not only detects the individual characteristics of the images, but it actually automatically categorizes images, whether it’s a hero image, whether it’s a size image, and they can also categorize. You know, as we monitor images over a period of time, it can also categorize automatically if images have changed or not over time. So, it’s a very, very advanced technology that we’re using, as you can, as you can hear the excitement in my voice.
It sounds almost like Flash.AI and Flash.PDP are cousins of sorts, I mean, they’re rooted in similar ideas in terms of leveraging AI in order to help our clients and leveraging a database and linking image recognition of packaging of multiple packs, maybe on e-com, you know, there may be displayed at times together with the database to see trends, yet, they are different in that one is exclusively running with that E-commerce focus, whereas the other is more so your traditional brick and mortar focus?
Absolutely, I like that idea of third cousins. I’m gonna start saying that.
You know, maybe it’s also worth adding Matt, that when we’re talking about this number of images, and really this amount of complexity that comes from managing so many assets and navigating so many kinds of rules, I guess from the different retailers and the different marketplaces, we’ve got to have something which is doing a lot of scanning in the background, rather than perhaps how the model has worked in the past where you have to almost nominate those brands and those cues that you want to learn a little bit more about from a shopper insight point of view.
What we’re saying instead is, look, let’s figure out which of your categories Do you want us to kind of monitor and take a very, very broad cast? Yep. So, in other words, that stay with cat food, everything for that client, including hero images, including carousel images, and anything which fits amongst our kind of, let’s say, top 30 salads. And let’s highlight across all of those because if we’ve got a failing in any of those kinds of images, we’d want that highlighted because that would be worth the return on investment.
We’ve also at the same time got to recognize that, you know, there’s a lot of fixes that are being kind of requested from various sources. And so, we need to be super selective about which images to optimize and put our focus on ones that are going to make a real difference, you know, where we can drive a much better conversion and a much better ROI.
And so, this idea of, you know, taking a broad cast, and then literally sort of setting it up and almost forgetting it, so that monitoring is kind of quietly going on in the background and pointing out areas that need further attention is, we think, quite an important difference in the way in which we need to think about these tools.
Yeah, I would agree. The whole idea of being able to have ongoing monitoring to help drive optimization, to me, is groundbreaking. And I love the tool so much because of that. Well, look, I definitely appreciate both of you taking some time to join us today. It’s been a great conversation, and I’m really excited about what the future will bring for the digital shopper. I’m excited for Flash.PDP and seeing how we’re going to help our clients optimize their digital images online for that shopping experience.
Once again, to our audience, thanks for tuning into Our Best Behavior, brought to you by Behaviorally. I’d like to thank our guests Adrian and Ruben, and we’ll catch you next time.