For e-commerce retailers, a strategic approach to both personalization and high quality product imagery can help a brand define its culture, create a better customer experience, and ultimately drive more sales. Crossing Minds’ Co-Founder and CEO Alexandre Robicquet and The Line Studios’ Co-Founder and Brand Director Lindsay More Nisbett sat down to chat about how to optimize conversion rates using the latest advancements in AI and cutting edge trends in product imagery. Check out highlights from their conversation below to get insights and best practices on how to supercharge conversions through personalization and brand imagery.
The following has been edited and condensed for clarity. To hear the entire conversation, check out the full video on YouTube.
ALEXANDRE ROBICQUET, Co-Founder & CEO of Crossing Minds: When we think about conversion, we've seen that images are the pillar or the cornerstone of the first emotional reaction of the customer interaction with the brands or the product itself. It's something that's so key to the translation of a brand and a product. How do you approach that whole concept with the customers? How do you define what approach is best for them?
LINDSAY MORE NISBETT, Co-Founder & Brand Director of The Line Studios: We always start with the overarching vision of the brand. So what is the brand DNA, what are those visual cues that they might use in their campaign imagery first and foremost that make their brand really unique and allow their customers to recognize their imagery, maybe stop scrolling for a second and say, "Oh, this resonates with me." So, I think first it really starts with understanding what those brand attributes are that we can kind of filter down through their campaign imagery into their e-commerce assets, and ultimately, you really do want those assets to be high quality, whether they're imagery or video, but you also really want to have those assets represent the brand.
What's been super cool to see in the last few years is how it's evolved into being more representative of that brand voice. That's everything from how you're lighting it, what models you're choosing, what's the hair and makeup, what is the styling, because a black dress style can be styled really conservatively for work wear or it can be styled in a completely different way to go out at night.
AR: You mentioned in this conversation the term DNA, which is very interesting because this is something that we use too, but we do it in a very data-driven approach, and you do it in a very artistic, creative approach, which I think is a super interesting overlap. It's like this importance of understanding and seizing the DNA of the customers as well as the brands. When we do personalization, there are two types of DNA that we really need to capture, first of which is the DNA of the product - what makes those products unique. So we look at the text, we look at the images, the tags, the prices, why this product is interacted with so often, etc. And then second, the DNA of the customers, and how those two types of DNA would then interact with one another.
In your case, when it comes to the images and the products, you were right on it in the sense of, each product could have different variants and the black dress could be very much formal. Then the same black dress with a different lighting, at a different angle and using a different model could be worn to a casual party among friends. For us, when you give us that input, they would be two completely different products, which is why the recommendations we provide to users have to be so contextual. But even in the Shopify ecosystem, there could be two variants of the same product. How do you decide which one to use? Do you provide both to the customers, and then it's a matter of us as a personalization platform trying to find which one to show to who at what time? How does that work?
LMN: I think that is also really dependent on the brand and how they want to inspire their customer, and I also think it always kind of comes down to e-commerce imagery always hits this balance between transactional and having that emotional connection with the customer to inspire a purchase. Ultimately, it has to show what it is, but a lot of what we buy, especially in the fashion world, is based on an emotional reaction to an image or video. So going back to your question, all of those have to be considered. Let's just use the black dress as an example. Perhaps it's a very simplified version where the customer might want to see themselves in the dress, so maybe the head's chopped off so it becomes more transactional versus a brand that is selling that black dress, but they want their customer to understand how to wear it with sneakers, but also how to put a jacket on it. So we might be shooting multiple images for them. I think it's super interesting when we show them all of those images, they might use all those images, but each customer might have a preference for different ones. So the customers’ behavior at the end of the day I think is a really interesting thing for us because we don't always get that feedback. I know that behavior-based recommendations are kind of a new thing for the personalization space, so maybe you can talk a little bit about that.
Good e-commerce imagery has to show what a product actually is, but a lot of what we buy is based on an emotional reaction to an image or video. –Lindsay More Nisbett
AR: It’s all about understanding the "why." Why someone likes something and what they want, and it doesn't matter who they are. I always give this example of someone entering a store and you as the salesperson, you have the choice of either knowing where they're coming from, what their gender is, what's their address. Or you could have the option of witnessing the first three items the person spends time interacting with. You don't need to have an MBA from Harvard to know that the second option is the better one from a sales perspective. Similarly, we do a lot of work on trying to capture what makes any given product unique, and then after that we try to capture in real-time what the behavior of the consumers around those products. Each click is very different: you can have a hyper-clicker that clicks on absolutely everything and then the signal is very, very noisy. Or you can have someone that is just a sniper and comes in, clicks, buys, leaves, and that's a completely different type of pattern to interpret. This is what we try to capture to then understand, why did they click on this product? They saw another product but they disregarded it - that's already a lot of implicit feedback about what has been ignored and not ignored.
I’ll give an example of where we had a very interesting realization when it came to one of our customers, which is a book publisher. We had very good models that were capable of looking at the story and the summary of a book - the models were actually looking at and interpreting the story and plot of the book. Then we had another model that didn't look at the story at all, but just took into consideration the images and extracted the information from the images. Which of those two models would be the most suited to predict what people would want? It ended up being the one that was just looking at the images and not at all looking at the story had a much higher accuracy in predicting someone's taste, actually. So people do actually care more about the cover of a book than the story itself. Which is a good way for me to then segue back to you because, for me, that was the moment where I realized how fundamental images were to creating customer excitement and predicting taste.
LMN: Right, so our customers are the brands, but their customers, you're asking them to make a decision very quickly and to understand a lot about your brand, your product, the why. What is that emotional feeling they're going to get by purchasing your product, and the number of places you need to do that is just becoming more and more - there's social media, there's your website, there's your digital advertising. It's just a lot and so I think helping our brands and clients understand that their imagery is really representing them. Every piece of it is important and that investment will pay off in the end if you just make sure that you're approaching it in the right way. What we do is try to convince them to look at all their channels and really think about what's going to benefit their customer, what's going to showcase your product, and are you using video in the right way and on the right channels, and how do your customers behave on mobile versus on your website.
I'd love to hear from your perspective in terms of a brand that is omnichannel, how are you guys approaching that and what are the differences you see in customer behavior across those channels?
AR: How a customer behaves on mobile is fundamentally different from what they do on the website, which is even more different than what they do in a brick-and-mortar store. For one, the attention span and the way that people behave on the phone is significantly different. Now, this should not be completely disregarded, because that's the same person at the end of the day. This is still a piece of information that's going to be valuable to interpret and translate the value of the clicks or the intention that there is when you recommend a product. So, going back to your question: 1) Yes, the behaviors are fundamentally different but they should not be disregarded, it's just another facet of someone. 2) It is impossible nowadays to think that you can serve proper recommendations if you don't have a system that starts when someone brand new arrives.
This is one of the biggest difficulties for recommendation, what we call the "cold start," when someone is engaged, to when they leave, and how to follow up with them, and when someone comes back and you have to rekindle their interest. This also ties directly into which images are presented to them. There's the image to seduce, there's the image to convert, and there's the image to retain, and there's the image to rekindle again. Like, "Hey, have you forgotten about this item?" I would imagine from your perspective there needs to be some diversity in imagery depending on which stage the customer is in.
LMN: Quite often, we're thinking about the core customer of our clients, which is the ones that have already converted, like you're saying, and it could be potentially a big missed opportunity to really think about, okay, how is the brand introduced versus how are we showing the clothing when it is a repeat customer. That said, we do do marketing campaigns which are really like more brand awareness, and so a little more romanticized, a little less focused on the transactional part of it, and so those are meant to be more of, if you saw this image, you would get a really good understanding of the whole world this brand is trying to live in and wants you to live in and communicate. And so those type of campaigns are definitely the top of the funnel, and then as you get a little bit more able to peel back those layers, those images are definitely approached a little differently. More commercial-minded and more focused on the product and the specific functions of the product. I guess in that way, yes, we do think differently depending on what type of imagery we're creating, whether that's living on a product detail page or it's living in a look book or a campaign.
AR: I actually have a question then for the campaign thing, because one of our main pillars when it comes to personalization is that personalization or recommendation needs to actually be personalized, and you have no idea how many of these recommendation systems are not truly personalized. Just showing the most popular items is something that 99% of the recommendation systems do nowadays as an answer to “personalization.” We actually focus completely on the individuality and the personality of the users. Now, that being said, this is something that is across every single one of algorithms and/or pipelines, but you can't deny that there is also a trend and an evolution of the taste on a global level based on the location or based on a given year. You mentioned at the beginning of this conversation that for a long time, e-commerce was just white backgrounds and the product, and now you're seeing these things evolve. Do you see changes in trends or what resonates with the users?
Just showing the most popular items is something that 99% of the recommendation systems do nowadays as an answer to “personalization.” We actually focus completely on the individuality and the personality of the users. –Alexandre Robicquet
LMN: Yeah, it's a few things. I think it is that injection of a unique brand voice, whether that comes in the environment that you're shooting in - so perhaps it's that that brand has a unique background color, uses modern propping, and has a harder light, and immediately those sort of help define this image as it relates specifically to that brand. But I think it also comes through in other ways. What we've seen especially now is there's just a lot of opportunities to utilize different kinds of assets. What we saw during the pandemic and then maybe we'll see a little bit more of moving forward is sort of a little bit more of brands willing to leverage UGC content - user-generated content - maybe on their product detail pages, where previously they wouldn't give up that kind of control, and I think what we're seeing sort of creep in now, is almost a more controlled version of that.
Maybe the brand is ultimately controlling and providing direction for those images, but they're not necessarily all needing to be highly produced. So we do what we call "low-pro" content, which is more of a content creator with the product that's going to get “real-life” photos, and those could be really beneficial for a brand to utilize on their PDP, in addition to the more refined and perfectly lit product photos that they're also capturing. I think both serve a purpose, but I think the other one - the videos, TikTok has to be mentioned. Bringing those sort of client testimonials to life on the PDP is something I think we're going to be seeing more and more. What is your approach to UGC in terms of personalization?
AR: We're tracking a lot of pain with the user-generated content. So the principle of recommendation is when you have an item that arrives, you have also the same problem with the cold-start for a product. How do you know who you’re going to recommend that product to if no one's ever interacted with it? So you look at the images, the text, everything is like some way to jump start the ability to recommend, but when it's user-generated content, it's savage. You may have a product where it has the wrong image, the price is put in the title, and the title is in the description. It can be really difficult to navigate. Then after that, you have to actually provide that recommendation, not only clean it in real-time, but then provide it in less than five minutes, which is absolutely bonkers. It’s why most of the recommendation systems are actually shying away completely from marketplaces. However, we definitely see the inherent value of UGC in helping us make the most accurate recommendations, which is why we choose to tackle that challenge as a personalization platform even though it can be daunting.
LMN: We'll definitely be doing more of these chats - thank you so much! This was a really enjoyable conversation.
AR: Thank you, Lindsay. Really amazing work that you guys are doing at The Line with product imagery. It's something that we would love to embed and push our customers to embrace more. Because, as I mentioned, sometimes the book is judged by its cover!
