Key takeaways:
- Treat explanations of search data as hypotheses, not facts.
- Measure conversions and ROI alongside rankings and impressions.
- Investigate ranking shifts from more than one angle.
- Consider site experience and credibility when a page underperforms.
- Create helpful content and revise your strategy as evidence changes.
Search performance rarely comes with a clear explanation. A page loses traffic, rankings shift, or impressions rise without a corresponding increase in leads. It’s tempting to point to an algorithm update or a recent content change, but either conclusion may leave out part of the picture.
Straight North Senior SEO & GEO Strategist Bryan Paravola sees search data as a starting point for investigation. To understand what’s happening, teams need to consider the page, the wider website, the audience, and the business results together. They also need to be willing to change their explanation when new evidence appears.
That’s why storytelling is indispensable to SEO and GEO strategy. Data can reveal a pattern, but explaining it requires a hypothesis about what caused it. We have to begin with the possibility that the search ecosystem can be understood well enough to guide a decision, even if we can’t identify every factor or be certain our explanation is right. A good story connects the signals, gives the team something to investigate, and changes when the evidence calls for it.
Why SEO Data Needs Interpretation, Not Just Reporting
A search report can show that traffic fell or impressions rose. It can’t, by itself, tell you why. SEO teams need to look at the signals together, consider more than one explanation, and revisit their conclusions as new data comes in.
I think the key is that we just want to remember that data itself isn’t some, you know, perfectly objective source of truth that’s, you know, perfectly, you know, unbiased or anything like that, because we’re always sort of looking at it through the lens of our own bias. You know, we always have to rely on, sort of, a little bit of storytelling, you know, a little bit of, creating narratives, and trying to figure out, you know, what hypothesis or what explanation best fits the data. And so I think the key is just to acknowledge that there is sort of a bias to it, and, you know, have a little bit of sort of humility towards it. And be willing to update our explanations as new information comes in, as we get new data. We don’t want to be too locked into, you know, one explanation as, you know, the explanation that can never change, you know, that’s objectively true, or anything like that.
Interpretation is unavoidable. The goal is to make it more disciplined: identify what the data shows, be clear about what you’re inferring, and stay open to another explanation. That matters especially when a report highlights a promising number that may have little connection to the result the business wants.
So which numbers deserve the most attention? Rankings and impressions can show visibility, but conversions and return on investment help reveal whether that visibility is producing value.
Which SEO Metrics Matter Beyond Rankings and Impressions?
Rankings and impressions show whether people have an opportunity to find a page. Traffic shows whether they visit it. Those numbers matter, but they don’t tell you whether visitors found what they needed or took an action that matters to the business.
Bryan references Goodhart’s law, which offers a useful warning: when a measure becomes the goal, it can become less useful as a measure of success. If an SEO team focuses only on increasing impressions, it may celebrate a bigger audience even when that audience is unlikely to become customers.
For example, a page might attract 50,000 visitors while only 200 take action. The 200 could be the more meaningful number if those visitors become qualified leads or customers. Looking at conversions, lead quality, and return on investment alongside visibility helps the team ask a better question: Is search performance improving the outcome we set out to achieve?
The same caution applies to newer visibility signals. Being mentioned or cited in an AI-generated answer can show that a brand is part of the conversation, but like an impression, it shows presence rather than value. Pairing visibility signals with outcome measures, such as form submissions, phone calls, demo requests, and the quality of the leads that follow, lets the team see whether that presence is leading to business results.
How Do You Diagnose Ranking Fluctuations Without Jumping to Conclusions Too Quickly?
Choosing the right metrics is only half the work. Even when a team tracks the numbers that matter, those numbers will still move in ways that are hard to explain. When rankings change, it’s tempting to name a cause right away: Google updated its algorithm, AI changed the search results, or a recent edit hurt the page. It’s important not to stop at the first explanation that seems to fit. Search performance depends on several moving parts, and a change in one metric rarely proves what caused it.
Start by identifying the pattern. Did the change affect one page or many? Did impressions, clicks, and conversions move together? Then look for other plausible explanations, including changes to the website, competing content, and shifts in what people are searching for. Google’s guide to investigating search traffic drops also recommends checking multiple possible causes rather than assuming every decline follows an algorithm update.
Consider a theoretical example: Suppose clicks to a service page drop by a fifth over a month. If the decline is concentrated in a handful of queries where an AI-generated summary now appears above the results, the team has a reason to look at how the results page changed. If it began the same week a new page template launched, and other pages lost clicks too, the template becomes the stronger lead. Neither pattern proves a cause, but each tells the team where to look first.
The aim is a diagnosis you can revise. State what the evidence supports today, what remains uncertain, and what you’ll check next. There’s no single method that explains every fluctuation. A useful explanation should become more precise as new information arrives, and lead to a clear next step for the team.
How Can SEO Teams Test a Hypothesis While Improving a Website?
Once a team has identified a possible cause of a ranking change, the next question is how to act on it. A controlled test can help isolate the effect of one change. Bryan compares SEO testing to a medical trial. Researchers use controls and placebos to help isolate whether a treatment works. But a doctor caring for a patient cannot always wait for a perfectly controlled experiment before trying to help. A live website creates a similar tension: teams want to learn what caused a problem while making improvements visitors need now.
So, with SEO, that’s kind of how it is. You know, we don’t always want to create perfectly controlled experiments. You know, experiments and studies certainly have their place in the industry, you know, in terms of, you know, figuring out, you know, how the algorithms work, figuring out, you know, what’s going to be the most effective way to, to do things, but when we’re actually looking to solve problems, when we’re actually looking to improve websites, you know, we want to, sort of attack it from multiple angles at once. You know, we want to sometimes create even, you know, multiple ideas at once. So, you know, we don’t want to perfectly control experiments every time when we’re trying to improve, you know, the website as a whole. We want to look at the entire health of the website as a whole, and, you know, really use all of the tools at our disposal to try to improve performance.
Controlled SEO tests have value, but a website may need attention in several areas at once. If the team changes content, site experience, and supporting information together, an improvement is encouraging, but it doesn’t prove which change made the difference. That isn’t a reason to hold back improvements. It’s a reason to keep a dated log of what changed and when, so the team can line up later movement in performance against specific updates and rule out explanations as evidence builds.
A useful hypothesis is one the team can check against new evidence and revise. That takes patience. If a content update does not produce an immediate gain, the content may still be useful; a problem elsewhere on the site could be limiting its performance. Likewise, a tactic that worked in the past is not guaranteed to work again as search changes. Record what you expected, watch the results over time, and look for evidence that supports or challenges your explanation.
When Is Underperforming Content a Site Experience Problem?
When an update doesn’t produce the gain a team expected, the next question is whether the content is really the problem. Even a page with the right keywords may struggle to rank. That doesn’t automatically mean the keyword choices were wrong.
Before rewriting the page again, look at the whole picture:
- Is the site easy to use?
- Does it give visitors reason to trust the information?
- Are other issues limiting an otherwise useful page?
Look beyond the page to the site’s and brand’s credibility on the topic. Clear authorship, first-hand experience, and accurate, well-sourced information can help visitors see why the content deserves their trust. Even useful, well-optimized content may underperform when a broader site issue or lack of authority is part of the picture.
Those questions call for a wider review of the visitor’s experience. A slow page, distracting layout, unclear navigation, or difficult path to conversion could limit the value of otherwise useful content. Metrics can help the team spot where to investigate, but no single number explains the whole experience.
So, you know, we can start by looking at things like, Core Web Vitals, we can look at, signals from, you know, click-through rate, bounce rate, you know, the depth that people are clicking through, in the path they’re taking through your website, conversion rates, and things like that. But we want to be careful not to get too hung up on, sort of, trying to reverse engineer every single ranking factor, you know, exactly how the AI algorithms are looking at your content, because that’s just almost impossible, if not impossible, and it’s not really efficient. It’s… important that we can take sort of a design stance and assume that these algorithms and Google are being constantly designed to reward content that is human-friendly, that is helpful to real people, that people want to see.
If visitors arrive but rarely continue to another page or make an inquiry, trace the path they are taking. Look for friction in the content, navigation, and conversion process, then decide which issue to investigate first. A Core Web Vitals score can help identify a technical problem, but even a good score does not guarantee strong search performance.
Beyond the metrics, consider what a person in the target audience would appreciate. That perspective helps the team decide whether the next improvement belongs on the page, elsewhere on the site, or in the way the content answers the visitor’s question.
What Makes Content Helpful for Both SEO and AI Search?
That audience-first thinking applies to the content itself, too. Helpful content starts with the audience’s question. A page should make its answer easy to find, then provide enough detail to help the reader understand the topic or make a decision. A clear opening summary and headings can help someone scan the page, while the full article gives those who need it room to go deeper.
For example, a page answering “How long does a website redesign take?” might open with a short, direct answer, then break the timeline into stages, explain what can speed the project up or slow it down, and link to related questions. The reader who wants a number gets it right away, and the reader who needs to plan gets the detail. The same structure that helps a person scan the page can also make its main points easier for search systems and AI tools to identify.
Content length should follow the reader’s needs and the complexity of the topic. A short answer may be enough for a simple question. A more involved subject may need examples, context, and answers from several angles. The aim is to cover the topic thoroughly without adding words just to reach a target count.
Brevity helps readers, but it should not come at the expense of a complete answer. Some topics need more depth and answers from several angles, especially when AI can help readers summarize longer pieces. Word count itself is not a direct ranking factor; the right length depends on what the topic and audience need.
That approach applies to both traditional search and Google’s generative search features. Google says foundational SEO practices still matter for AI search and emphasizes original, helpful content. It doesn’t prescribe a special content length for those features.
“What is a human user from your target audience going to appreciate?”
Senior SEO & GEO Strategist | Onsite SEO & GEO Expert
How Should You Report SEO Results Without Overclaiming?
Once improvements are in place, the next challenge is describing their effect accurately. A useful SEO report should explain what changed without claiming to know more than the evidence shows. Be wary of promises such as “this title change will improve rankings” or “adding 500 words will fix the page.” Search performance has too many moving parts for one change to guarantee an outcome.
Instead, separate the observation from the explanation. For example: “Clicks increased after we updated the title” describes what happened. “The title update caused the increase” is a stronger claim that needs more evidence. The report can identify other possible causes, explain what the team will check next, and say what result would make it reconsider its conclusion.
Here’s how that might look in practice. Instead of writing, “The new title increased clicks by 15%,” a report could say: “Clicks rose 15% in the four weeks after we updated the title. Demand for this topic also tends to rise this time of year, so we’ll compare against the same period last year and check whether the increase holds. If the gain fades as seasonal demand levels off, the title change is less likely to be the cause.” The second version is longer, but it gives the reader the evidence, the uncertainty, and the next step.
But we want to always make sure that we’re approaching it from a sort of a holistic perspective and thinking about every factor at play. And from a human point of view, from a user experience point of view, how would a real human interact with this? We don’t want to get too focused on, sort of, the algorithmic stance, because algorithms are changing so fast, and they’re being changed and adapted and reinforced by real humans. So, you know, at the end of the day, if we really want to future-proof our content, you know, we do have to create helpful content that’s people-first, that’s helpful for real humans.
That people-first perspective should also guide what teams do when search performance changes. It ties into Bryan’s closing advice: Rely on more than one method to explain what’s happening, rather than a single best practice.
So, I think my one piece of advice would be that, you know, we always want to try to use a variety of methods in trying to explain, you know, ranking fluctuations or performance in general. We don’t want to just be too caught up in looking at best practices, looking at what’s worked in the past, and say, well, this is for sure what’s gonna work, this is the one answer, because sometimes there’s a whole bunch of factors that could be playing into it. So we want to not get too caught up on one method, you know, there’s not just one right way of doing SEO and GEO, and sometimes we have to be a little creative, you know, we have to sort of think about, you know, how are different… what are the different angles we can come at this problem from?
Taken together, these ideas point to a repeatable approach. When rankings shift, look at the page and the wider site, consider changes in search demand, and check whether visibility is turning into meaningful action. Then form the explanation that best fits the evidence, make an improvement that serves visitors, and decide what result would cause you to rethink the plan.
No team can account for every factor in a changing search ecosystem. But without an explanation to test, it is difficult to know what to investigate or improve. The value of an SEO and GEO story is that it gives the team a defensible next step while leaving room for a better explanation to emerge.
