Design That Sticks: Applying Neuroscience to Predict Attention

Daniel Rowles: Welcome back to the Digital Marketing Podcast, brought to you by TargetInternet.com. In this episode, we have a fascinating conversation about applying neuroscience to predict attention. So basically, how can we use neuroscience to help you create higher-performing digital products and assets?

So in this episode, I am joined by Kramer Reeves, who's CEO of EyeQuant, uh, and Professor Peter Corrick, who's one of EyeQuant's founders, and a world-leading expert in neuroscience. Now, if you're not familiar with EyeQuant, they take the concept of an eye tracker study, and that's where we get loads of people to look at the same content, and then try and create a heat map of where they actually look at that content to work out what's cutting through, but then using AI to really effectively predict where people will look instead.

So the idea is you don't need to spend hours, days, or even weeks collecting all that data, getting people to look at something, but actually can predict to a really high level of accuracy where are people going to look at something, that being a webpage, a social media post, a design, or anything else. There's some really brilliant insights in here from Peter about how our brains work, and how actually we're really poorly adapted to interpreting stuff that we see on webpages.

But what's brilliant as well is they've set up a two-week exclusive trial for listeners of the Digital Marketing Podcast, so the theory's brilliant. There's some really great insights, but then you can go away and you can test this out in practice, and I've played with it, and it is phenomenal. So if you go over to targetinternet.com/podcast, you will find it there in the show notes.

But without further ado, over to the interview Okay, so I am here with Peter Anne Kramer. And let's start off with why do you think so many well-designed web pages still fail to convert and to deliver? Peter König: I think we have to step back and consider what we are built for. And my take is we are, concerning our brain and visual system, hunter-gatherers.

We are optimized running around in the woods. You always know where you are, whether are you in a hut with your tribe in the woods, hunting a wild boar, or looking for an eagle. And when you click on a website, bang, something new is in front of you completely. So you are out of context, and you have to orient yourself, and this is very, very difficult essentially.

And to top that, the designer knows this website very well, and I believe you see some great websites. But you look at your website, which you know very well, with very different eyes than the end customer, for which everything is surprising. And so far, different mechanisms in our brains kick in, and a different mechanism kicks in, in your brain designing it compared to the one who is looking at it the very first time. Daniel Rowles: That's an interesting thing, this out of context, idea is really interesting.

So when someone lands on a page, what's really happening in those first few seconds then? Peter König: You try to orient yourself, which means that you look at the page and you are input-driven. Let's say it was surprising to me when I learned about it, our brain is mostly talking to itself. It is not sitting idling, waiting for the input finally to arrive.

To give you some numbers, there's roughly one million fibers from each eye to the brain, but the two hemispheres we have are connected by two hundred million, and connectivity inside each hemisphere is even more intense. So there are several orders of magnitude difference between how much information we get from the outside compared to what is going on inside. You might compare it with a huge discussion room, one hundred people in the congress and one external is joining the party and tries to communicate a message.

This is really a hard task. And to guide this process, what to attend to are typically very simple stuff. So for example, traffic signs are bright red and white in order to catch your attention, and you have to use such methods in order to establish a context, guide the viewer what it is all about, and guiding along the story you want to tell. Daniel Rowles: So Kramer, give us some kind of business context for how IQAN is kinda helping with this then, 'cause I think it's really interesting.

' I've kind of introduced what IQAN is, but I think it's good for people to see how it relates to this in a business context. Kramer Reeves: Sure. You know, we go back to this key point that many of the designs are built based on aesthetics, but the brain, as we're talking about, sometimes just can't actually convert that or put the context together to drive the user to the action that the designer wants. So what we've done at Eyequant is to model out what the eye actually sees.

We've created models so that we can predict What the user will see and likely do, and we've put that into our platform. So you can test your user experience, your creative, your ad, whatever it is you want that person to see, you can test that in advance. The first thing that every designer needs to do with their creative or their, or deliverable is we need to make sure somebody sees it. What do they see, and how do they see it?

And that's what EyeQuant does. Before the actual end user sees it, we can predict what they will see and how they will see, and then we provide all kinds of data about that particular prediction at analysis, so the designer can adjust and optimize before they put the creative or deliverable into market riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: And like it leads really nicely in as well actually, 'cause Peter, you often talk about behavior happening below conscious awareness.

What does that really mean in a digital context? riverside_peter_könig_raw-audio_digital_marketing p_0232-1: As in our research field, consciousness is the big question. Everybody's buzzing about it. It's, it's enigma.

But when we are honest, we do a lot, and consciousness is like on the passenger seat, nodding, saying, "This was correct, this was correct," or saying, "Oh, what did I did just now?" So it is like, after the fact, and from our evolutionary origins, we have to act quickly And this does not allow for huge deliberation, So there's a concept of system one and system two. System one is doing the stuff, and system two is on the passenger seat, can do in-depth reasoning, deliberation, and all that stuff.

But most of the action is done by system one very quickly. And so far when we are in discussions, how do we do it? So system two is over-represented. It feels super important and always wants to be in the spotlight.

But sorry, system one is a real one. The s-soccer player who really scores the goal and shoots it is present at the moment when it's important. riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: The other thing that I've kind of come across a lot and you talk about is saliency, and I think that it's not something a lot of people will necessarily be familiar with, but why does it matter so much for design as well? riverside_peter_könig_raw-audio_digital_marketing p_0232-1: Wonderful question.

Saliency is a property of the image to attract attention, to be selected for focused attention and to be a-acted upon. Which means as it stands out, there is not a single color which is salience per se. So if you could see the video right now, I have wonderful neon green glasses, in front of a pale skin and a pale office environment. They stick out.

If I would be, for example, an alien with a green head, the green glasses would, well, yeah, there's a strange green alien, and it has green glasses, but the glasses would not be that remarkable. So it's a property of the image which captures your quick attention, and there you start your analysis. So it is quickly guiding and thereby often predetermining what you will do at the end. So for system two to change what's going on with this quick reaction is much harder than when you are already on the correct road towards the correct goal.

Therefore, salience, properties of the image of your visuals which determine how the uninitiated viewer is processing the visual is super important, and this is what the model is doing. It is, so to say, predicting which parts of a visual are attracting how much salience. And for example- A typical sequence which we see is you make a great ad which is nice, aesthetic and your boss comes in and says, "Yeah, but there is Black Friday. We have to add fifty percent reduction here."

And then he comes in back two hours later and says, "Yes, but we have too much stockpile of product Y. We have to advertise it to get rid of it." And so you overload your ad with a lot of stuff, and this dilutes the salience, and you leave it to chance where the viewer looks first, and essentially you leave it to chance what's happening. And so far, this is something we often see, and an objective model which so to say models the viewer and is ignorant of, well, your plethora of visuals, everything you want to communicate in parallel is a real good base check and leads to much better, much more focused advertisements.

riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: That's a great example. So what are the common mistakes that teams are quite often making when they're trying to guide attention? riverside_peter_könig_raw-audio_digital_marketing p_0232-1: They are overloading To say too much. And understandable because you as a designer, you know it all.

You want to plurp it out in parallel, but you know your ad. When you are accustomed to it, you you don't find it that much loaded. But the other one who is looking at it for the very first time finds it super crowded, and our model tells you that. Then there is a clarity and the storyline where you start.

Where is the first fixation from where you move on? So if you, let's say, the click here button, of course, it's important. You you want to have a nice salient click here button. But if you start with it and I see a webpage where the main message is, "Click me, click here," I swipe left and go somewhere else.

And then so far, you also have to think about the storyline, what is viewed when, and adjust it accordingly. riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: So with that in mind, what kind of role do color, contrast, visual hierarchy play in what users actually notice? 'Cause obviously I can see that's gonna be highly connected. riverside_peter_könig_raw-audio_digital_marketing p_0232-1: Yeah.

So here you a-asking for a simple recipe. Let's say a-again, how we are built, we would like to interact with what is close. So if you have the visual effect that something is close, most primitive is if it's large. So if something is large, probably it's close.

We pay more attention to it because everything which is far in the distance, well, we have some time to adjust our reaction to that. And then, of course, I mentioned color. Also here, color contrasts are more important So that my green glasses, they stick out of my face because I'm not an alien from Mars. They are super salient, and you will start looking there and seeing, well, this is a nice guy.

Do I trust him? Maybe. And from there, the story unfolds. And similarly motion cues are very strong.

But you have to think about you cannot give first priority to five items. riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: Right. riverside_peter_könig_raw-audio_digital_marketing p_0232-1: but doesn't work that way. riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: So that leads us quite nicely because you've mentioned that visual confusion can reduce conversion.

What does that kinda look like in reality? riverside_peter_könig_raw-audio_digital_marketing p_0232-1: So we have to measure clarity, which is the opposite of visual confusion and determines how clearly, how unique the storyline, the foci of attention and the path is. I want to shortly discuss another issue. We are in modern AI times, so we have models and originally people thought, well, as you are asking which features are important, and I mentioned colored and so on, and we scientists feel proud if we can mention that.

We have to admit end-to-end training models surpass the performance, which means we take human behavior, we take all the visuals, we train the model to make optimal predictions, and the models are even better than the experts, which is great. We use such models end-to-end trained for superior performance. The downside we pay is they are black box models. If you have a visual, it tells you all the clarity and the saliency and where the focus is, but it does not tell you why.

Therefore, we need, again, the second type of model as well, where we have identifiable features and can say, "Yeah, it is crowded because," and say, "Okay, something here, something there, and here an addition. This you have to change in order to help you." This is like in school a teacher who says only, "Well, this was not a good answer," but doesn't help you to give a better answer. You you need both.

You do not need only the genius teacher, you also need the teacher who is helping you to improve. riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: It's fascinating. So, so for teams listening, what's something kind of practical, one thing they can go off and do to im- to improve performance potentially? riverside_peter_könig_raw-audio_digital_marketing p_0232-1: To get an objective measures.

We started with the designer looks at his own ad with very different eyes. Because the brain is talking mostly to itself, he sees his mental context in the visual, but the customer at the end does not have this mental state. He looks at it for the first time very very different. And that's the second issue.

The designer is accustomed to it. So take a well-known webpage like eBay or something. If you see it the very first time, show it to your parents, they say, "Oh, God, what's going on here?" And they use it 200 times, okay, then they know what to click where.

But f- for us, in principle, for the uninitiated user, it's not really well-designed, you must say. And for here, objective measures, models which are not trained to use the specific sites but behave like you see it the very first time are super helpful because they give you the reality check, how your visuals actually look like. riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: So it is really interesting to me because so much of the work we do is fairly subjective, and we talk about best practice and those kind of things.

How do you think the role of neuroscience is evolving in design and marketing? riverside_peter_könig_raw-audio_digital_marketing p_0232-1: So neuroscience and science in general loves data. And so far when we are modeling, we have our measures, how good the models are, variance explained and all that stuff in order to get to the best models so that we know what we are doing. That's super important.

The knowledge about how the brain works can guide us. For example, I've repeatedly said the brain is largely busy with it- with itself. The brain has mechanisms which are fast, unconscious, able to deal with sudden threats and new information, and we have systems which do deliberation. But these are not completely separate, but because the first one feeds the information to the second one.

And so far you cannot only say, "Well, wait till the user really thinks about it and understands everything," because a fast system is selecting what is fed to the slow system. And in so far- Neuroscience, or let's say cognitive science, helps dramatically in order to streamline these communication channels What I explained with the brain is also an opportunity because when you create a context, you can fit in what the brain typically considers as important in different contexts.

For example, movement. If something is moving, well, there's action, this is important. And we found that if you can have visuals only which are static because it's print press and there's just an image and you cannot make it move on paper, the effect is still of moving images uh, of moving objects which are photographed 50% of the real motion, which is quite a lot. So imagine you have a photo, a static photo of a car compared to a video which shows this car moving towards you.

The static photograph communicates still 50% of the saliency effect, which I is is great. You get it for free without any video technique. And similarly, the contrasts which are important, color, luminance, texture, and so on, guide the eye movement, and this can help you to make better visuals. And so far, cognitive science is helping here a lot.

riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: fantastic. So look I'm, I'm really interested, Kramer, We kind of understand the theory. We looked at some really good practical tips of how we might apply this. Talk to us about IQANT and, you know, what it does, who it's for and kind of where the product's going as well.

'Cause I think since I was first aware of the product a number of years ago, there's been huge evolution and things have really developed. So talk to us about that. riverside_kramer_reeves_raw-audio_digital_marketing p_0234-1: Okay. Sure.

So, as as we're talking about the core problems that we're trying to solve the biggest problem is our clients and the designers out there have to go faster. They have to do it more effectively, so in a cheaper fashion, less expensive fashion, and they have to be more accurate. So fast, cheap, accurate. That's what we're shooting for with our technology.

Can we help our designers get the objective feedback earlier in the design cycle so that they can optimize and improve and put into market what's more important? In addition to this, one of the key things, one of the key things we hear back from our clients is they are measured on outcome, on performance. How do we start with the end goal in mind? How do we start with the objective in mind?

So we've incorporated into IQuant intent-based outcome inputs. So the inputs start, what is your objective? Natural language, just write it and speak it. Tell the - Tell our platform what you're trying to accomplish.

A-and our model will then look at your creative, look at your design, look at your deliverable, whatever you want to test, and tie the analysis and recommendations to your intended outcome. And that is important because if we don't know the end objective, we really can't test for the full point of what you're trying to build. We can model out what the eyes will see and how they will see it, but we've taken it a step further to answer your question, how has IQuant evolved? Over the last year, we've taken it a step further, and that's because of the evolution of AI and the accessibility through things like MCP Model Context Protocol server integrations.

We can bring in the intent and tie that to what our analysis and our recommendations are providing. So that's one of the most important changes that's been made over the last year. riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: It's really interesting because we had - we did an interview of Adobe literally a few days ago, and they were talking about, "Look, we've got all this data now. We can bring it into one place, and you can take things out there way more quickly than you could ever before."

But the problem is, if you've spent months developing a campaign and you're very wedded to your designs, you know, going back to what Peter was saying, actually, we need to be able to test those before we even get them out there in the first place and actually understand how people are gonna react to this content. So I think it... In this very busy, noisy environment where everyone's generating stuff using AI as well, actually making sure it's gonna be effective is gonna be so key to that.

So if if listeners wanna go off and try and play around with this technology or test it out, what's the b-best thing for them to do? riverside_kramer_reeves_raw-audio_digital_marketing p_0234-1: Well, the listeners of your podcast are in - they're in good shape because we normally offer a free trial. We are going to extend for this group only to two weeks of free trial of EyeQuant. It's super easy.

We'll pop the link in. You guys can get access to it and take it for a test run, and it's it's a great way to see if this is something you can include into your workflow. And this is a key point. You designers we've been following in the design thinking process over the last 10, 15 years has evolved into this infinite loop.

We understand that. You want to slot verification, pre-testing, this type of system into your process, into your workflow. So whether or not you're in Figma, it's just a press of a button to test. Adobe, other tools.

And so the best way to learn about this is to take it for a test drive. Within about 15 minutes you'll know, does - is this something that can help me? That's how easy it is and how quickly you'll deduce, do I get the insights I need to hit my business objectives? riverside_daniel_rowles_raw-audio_digital_marketing p_0233-1: That's fantastic.

We'll put that into the show notes, so targetinternet.com/podcast, and you'll be able to find that special offer. I'll just say this from a real practical point of view. We don't have paid placements on the podcast.

Now, whenever we have software on, people are a little bit suspicious of like, "Have they paid to be on here?" We absolutely don't, and I would say that Icon... I've worked with an agency recently that were using it and it's made a huge difference to things because there's so m- so much subjectivity. If you've got three or four expert designers, they all have different opinions and different aesthetic styles and things like that, and actually what they are able to do is just take a bit of a step back and say, "Look, actually, this is what's catching people's eye.

This is where actually they're looking in the first place. If we reduce down the amount of noise in this design, it's having a bigger impact." So very much coming back to a lot of things that Peter spoke about, but very much in practice happening really quickly as well. Okay, Peter and Kramer, thank you so much for joining us on the Digital Marketing Podcast.

riverside_kramer_reeves_raw-audio_digital_marketing p_0234-1: having us, riverside_peter_könig_raw-audio_digital_marketing p_0232-1: Thanks a lot.