Learn how to analyze behavioral data to improve your search performance. Discover the three levels of diagnostic tools — from basic GSC data to advanced neuromarketing metrics — and how to use them to optimize the entire search journey.

Click on the whiteboard image above to open a high-resolution version!

Hi, my name is Giulia Panozzo, and I’m a neuroscientist turned marketer, and today I want to talk about the behavioral data that matter for search. And why behavioral data? Because search has changed dramatically. Not only search as we know it, with the introduction of AI overviews, organic product carousels, and other features that have impacted both informational and transactional queries in the past 12 months, but search behaviour has changed too.

Users are already making searches that are more conversational and their search journey now spans across different channels, including socials and LLMs, and it is estimated that by 2026, traditional search engine volume will drop 25%, with search marketing losing market share to AI chatbots and other virtual agents according to Gartner.com.

SEO is now getting evaluated on journeys that are no longer linear and are driven by the user. And SEOs have always been shy talking about user behavior and relegated it to the UX teams because it’s not an official ranking signal, however, some recent data coming from the Google doc leak and Mark Williams-Cook’s research have highlighted the role of user signals in ranking and the importance of nailing user intent to benefit the overall evaluation of a website, so it’s time that we acknowledge that optimizing for search now includes more than just getting a click to your website, but encompasses the entire journey, something that is now referred to as SXO, the intersection of SEO, UX and CRO (something that Sara Fernandez often writes about). What all these disciplines have in common is the user as the end beneficiary of our optimization efforts.

So SEO as we know it might be dead, but the future of search is analysing and predicting user behaviour in order to optimise accordingly.

And when we talk about users, we are talking about humans who make decisions all the time and are very often biased. Familiarizing with what these biases are is important for everyone working in marketing, but in order to understand and influence users’ behavior, it really all comes down to understanding and mastering two main dimensions:

• Getting  Attention (to stand out in a sea of potential options)
• Fostering a Connection (so that users keep coming back to you)

Provided, of course, that what you have to offer is relevant to their search.

That’s why we need to include the study of other measures than just the traditional SEO ones. The new data we need to take into account spans across the entire search experience and multiple touchpoints and include behavioral data. 

Looking at behavioral data to inform your search strategy is what I imagine a doctor needs to do when examining a patient:

• You listen to complaints and symptoms
• You analyze data to diagnose the root cause
• You prescribe a treatment

The symptoms are the easy ones to start with, because they are relatively easy to spot and are usually quite uncomfortable to deal with from a business perspective, so they’re the ones that your stakeholders will care the most about and will bring to your attention.
These might be: Loss in traffic/low clicks to a site, lower impressions, lower average order volume or conversions. These are generally just an outer manifestation of something that might be wrong on the inside, so you’ll need to dig a little deeper.   

When it comes to analyzing the root cause, we have several diagnostic tools we can use, and they reside on three different levels of data that we can get: the basic behavioral data, the next level data, and the predictive data. Let’s get into each one of these.

1. Basic data

The basic data comes from tools that you don’t need buy-in or set up for. One of them is Google Search Console (GSC), which can reveal poor intent match when we look at CTR both from a branded and non-branded perspective. Most of the other data in this bucket is qualitative and allows us to identify common points of frustration both for a pre- and post-purchase journey, like surveys, CX logs, social mentions, and reviews, so make sure you collaborate cross-functionally to have access to what users are asking of you. There’s also live testing, which is the most time-consuming option, but also potentially one of the most rewarding since there’s not much to infer.

2. Next level data

Next level data is primarily quantitative and can be obtained with tools that need tracking set up, like web analytics and heatmaps, which record user behavior that might be less explicit because they are not actively communicating their frustrations, so there is a certain level of inference we have to apply to our findings. From web analytics, you can look for examples related to engagement time and engaged sessions or bounce rates and points of abandonment. Interaction heatmapping tools can integrate that information and everything that we've seen from web analytics, uncovering not only areas that might not be getting enough attention but also elements that don’t actually work (via dead clicks, rage clicks or error clicks for example).

In general, while we can infer why some journeys get cut short or don’t end as we expect via these tracking tools, I always recommend to pair it with qualitative data to really understand what is going on.

3. Predictive data

Finally, there’s the predictive data. This is the hardest to get, because it relies on special equipment and training to properly interpret it, but this is something that uncovers preferences and behavior that not even the user might be aware they have. For example, eye-tracking goes a level beyond heatmapping data and can show us attentional patterns and areas that are missed. This can inform the design of pages, which is important as attention is a precious commodity in the land of 24/7 stimuli. On the other hand, electrodermal activity, EEG and fMRI measure neural activation in response to marketing stimuli can help us understand and predict preferences in content even before the user is aware of them.

So, now that we know all of the diagnostic and predictive tools available, it’s time to plan your treatment. Depending on the size of the business, there might be an element of collaboration and prioritization needed, so to facilitate it, ask yourself these questions:

• What’s the time/effort involved in this fix?
• How critical is this fix to the success of the business?
  • This fix resolves a blocker to navigation or conversion = urgent
  • This fix is a nice-to-have feature = not urgent
• What’s the impact or ROI of my fixing this issue on the wider business?

This will inform how to populate a prioritization matrix, where everything that has high impact will get done either now or in the near future, and everything that has low impact will either get postponed or discarded.

Anytime you implement a fix, make sure you always record what problem it solves at a deeper level, so you can identify cross-domain opportunities for improvement.

For example, if we see lots of searches but poor CTRs for return queries, fixes can include making return policies more available both on-page and on the Merchant Centre. If we look at the user need and the underlying bias that this solves for, it’s the need to avoid losses, which means we can proactively address this need on other areas, like for example in the pre-sales messaging (e.g. “free trial,” “no credit card required,” which conveys to the user you save both time and money).

It is now our duty as search professionals to take into account behavioral data.

You don’t need to be a UX professional to investigate them, and your title shouldn’t be an excuse not to deliver better content or products to your audience. As SEOs, our job doesn’t end when we make them land on site, but it continues throughout the entire journey to make sure that user interaction is a positive one and doesn’t end in abandonment.