Imagine walking down your local shopping street. While browsing through shop windows, passing by a few busy cafés and shopping spots, you bump into a huge 50% off sale sign.

There is very little chance that you would miss seeing this offer. With bold, white text on a red background hung at eye level, you can not avoid noticing the sign. Your attention is innately drawn to it.

You stop by the store and notice something: it seems that so many others have stopped in their tracks by the sign. Many have seen it, and now enter the store to claim their offer.

‍

When everybody’s looking

You choose to stand outside the store and look, and the more you look, the more it seems that almost everyone glances at the sign, and many passersby stop and go into the store. What seems to happen here is that, even for a brief moment, people's attention is driven toward the same place: the discount sign.

What you've observed is a very coherent consumer response. Consider this a gem! If you can identify consumer responses that seem to be very similar across different people, there's a high chance that you can use these observations to predict something about as large as a whole market.

Why? Because when a smaller group responds very similarly, there's an increased chance that adding more people to the mix will show the same pattern. And so, the question now is, how many people do you need to make an observation representative of the global population? To answer this, there's no way around some hands-on statistical craftwork: you need to test different sample sizes and determine what sample size you need to reach the same conclusion. 

Consider another scenario: you pick out another sign from the street. This one is a bit higher up in the window. It's dark blue with pink curly text and the slogan reads: Get your groove on. You watch the scene and observe people going past this sign. It barely gets noticed from all the surrounding visual noise!

Only a single person sees it (besides you). It can be hard to tell exactly why this happens. Perhaps the text is not standing out, and is not visually salient. The colors may not be attractive enough. Maybe the text is not engaging, or the copy is hard to understand. In any case, Get your groove on doesn't seem to have the same pull as the first sign.

Now, you realize that you're onto something. But how can you foresee what gets noticed and what doesn't catch attention? How can you measure attention and make predictions?

Sure, you just saw that you could infer others' attention by observing their head pose and gaze relative to the sign you wanted to investigate for attention. But all things being equal, this method is crude and imprecise, not to mention that it is also not scalable to test many people at the same time. 

The obvious next choice is eye-tracking.

‍

What is eye-tracking?

Broadly defined, eye-tracking is a way to automatically measure and track eye movements. Today, hardware-enabled eye-tracking is a solution growing in popularity, application, and usability. Whether you're an advertiser, a designer of apps, packages, or websites, or if you're a retailer, eye-tracking is fast becoming the industry standard for assessing visual attention. The field is growing so rapidly that innovations are making headlines every month. 

‍

‍

Eye-tracking is often done through hardware-enabled tools that use infrared light for optimal gaze detection. A more recent method uses webcams on computers and phones. Using webcams does, in principle, allow for a global and very scalable way to measure consumer attention. But as we have written about webcam eye-tracking earlier, it is typically limited and comes with a set of premises before you can use it.

For example, webcam eye-tracking requires high-quality cameras to collect relevant data. For accurate results, it is also necessary that participants sit in a well-lit environment without moving too much. Even then, webcam eye-tracking should still be used to track only gross eye movement (following a moving object) and not to make detailed conclusions about attention drawn to smaller regions of a screen.

‍

The basics of predicting market responses

Can eye-tracking predict group behavior as vast as market responses? Based on the assumption that attention acts as a “gatekeeper” for all consumer decisions, attention can definitely tell us something about how a market at large will respond to an ad.

All advertisements have a so-called “stopping power.” As the first crucial step of the 4-power model for advertising success, “stopping power” is an ad’s ability to stop you in your tracks and grab and sustain your attention. If this does not happen, people won’t remember your ad and, in scientific terms, you will not see any emotional or cognitive responses to the ad. Attention is the gatekeeper. This is also why attention is the main driver of in-market effects and can make an lasting impact within a single second.

Can we predict the behavior of tens or hundreds of millions of people by measuring only 50 or 100? Surely that sounds very optimistic.

It all boils down to how reliable the behavior we want to predict is. Some types of behaviors are more reliable than others. In other words, some of our responses to certain images, sounds, and videos are more similar to other behaviors. 

Imagine that you went back to the shopping street with the store sign. Instead of observing people, you now do an experiment. You try different things to see how people react and try to find coherent responses. More specifically, you try to see how similarly they react.

First, you try ringing a bell. A few people passing by you will look and smile, but it seems they are the only ones. Next, you start singing. Perhaps you're a decent singer, and there's a gathering of people around you to listen to you sing.

Finally, you try something completely different. You now blow up a balloon, and you make it pop! Most people around you are startled and stop in their tracks; some might even scream in shock. Now, you see a very different type of response. You've discovered a coherent behavior!

This means that when many people respond similarly, it's also very likely that a similar type of response will happen at the scale of the market. If an entire nation is exposed to exploding balloons, they will all show a startle response! Some will respond a bit more intensely, some a bit less. But they would all be startled. 

‍

Coherent and consistent responses

Identifying this type of consistent behavior requires a lot of work. First, you must focus on getting the best possible tools to measure the behaviors you want to observe. Preferably, you should use high-resolution eye-tracking when you are measuring attention. If you want to say something about consumer attention, then the visual materials you are showing to people should include consumer-relevant items and not general images only.

These materials could range anything from ads, packaging, websites, app designs, and store shelves to product and brand placements. You can run eye-tracking studies on different platforms like TV, print, social media, in-store, or even outdoors.

Finally, you need participants to run your experiment. You’ll need people. And you’ll need a lot of people to get statistically significant results. Fortunately, when you have a large sample size, you can calculate the optimal sample size for your experiment to come to valid conclusions.

There is a way to learn what the minimum sample size is for you to get reliable results. This method is referred to as Monte Carlo bootstrapping simulation. Basically, it's a systematic test asking what the sample size needs to be for you to derive the same conclusions in a reliable manner.

In relation to eye-tracking studies, most of these calculations are optimized for computer screens. Here at Neurons we have collaborated with Meta and Copenhagen Business School to expand on these calculations and estimate the desired sample size for eye-tracking on phones as well.

‍

Predictive eye-tracking

The latest development in eye-tracking is called predictive eye-tracking. This method relies on in-depth studies we have just covered above and is based on the fact that certain types of attention are very similar in people. We all tend to look more at delicious food and desserts (especially when hungry). We tend to look at faces first. We look more at well-known brands than unknown ones. Predictive eye-tracking uses this similarity in responses to create something new. Something magical, you might even say. 

To recap, we now know that some behaviors and responses are coherent and similar, and we also know how many people we need to test this. Based on this, we can collect the right amount of high-quality data to bring this to a new level: using machine learning and AI to create models that can predict the responses. 

‍

Why would you do this? What are the advantages of using an AI to predict something that we can measure ourselves? It has plenty of advantages, it turns out, including:

• An AI model typically gives the same results in seconds to minutes instead of weeks or months. 
• The AI never gets tired; you can test endless materials on it. 
• An AI model is resistant to the repetition effect, allowing you to test multiple variations of the same design. 
• It's a lot cheaper, as a single study would cost more than you would pay for a year's usage of an AI model. 
• It allows testing on the fly while a design is being made. That iterative process would not work even using online eye-tracking methods, since one would still have to wait days to weeks for the results.

‍

Marketing lessons from eye-tracking studies

As a marketer or brand owner, the new science of predictability offers many new and untapped opportunities. We have seen across multiple industries that using predictive eye tracking introduces a wide range of positive changes in how teams work, have discussions, and make decisions. Some of these changes include:

• Moving certainty upfront. There is a great deal of uncertainty when designing marketing materials. Predictive technologies allow decision-makers to be more certain about their choices up front and before launching designs and campaigns. 
• Front-loading of predictions as pre-insights. Brand owners, marketers, designers, and others use predictive technologies to get early feedback on how a design works. This allows for making decisions on the fly and during the design process. 
• Test, predict, and repeat. Users of predictive technologies use the methods to try things out again, again, and again. Since the results come in seconds, they can try many different iterations of a design and learn what works best. 
• Boosting experimentation. When predictive technologies are used to their fullest, designers try out things they would not otherwise do. They might have an idea or even an inkling of what could work, but the cost of failed experiments is traditionally several orders of magnitude above their pay grade. 
• An internal language. When teams and companies start using prediction technologies, it changes how they work, collaborate, and make decisions around the table. When teams get even more creative, we can sometimes see how they combine features that win on different aspects such as attention, emotion, and comprehension, thereby trying designs they would have otherwise never tried. 

As neuroscience and platform technologies allow new solutions based on predictive technology, a whole new branch of opportunities are emerging. We already see that prediction technology is the new frontier for businesses, and those who adopt and use the solutions are becoming more likely to succeed in the market. And while at the office, you’ll hear many more phrases like "have you predicted it yet?"