A homepage built to serve five kinds of visitors was, in practice, being used by one.
- A large energy retailer's homepage offered several next-step paths beyond signup — but heatmap data showed the large majority of visitors ignored them and went straight for signup anyway.
- The team paired that qualitative diagnostic with competitive research on early plan-personalization before writing a hypothesis — not after.
- The experiment (Exp-051) stripped the alternate pathways and rebuilt the page around the user intent conversion path the data had already revealed.
- Result: a 5% to 10% lift, and the variant shipped as the new default.
- The lesson generalizes past this one page: more apparent choice is not the same as more real choice, and a page can look inclusive while quietly working against the one user intent conversion path most visitors are already on.
| What the page assumed | What the heatmap showed | |
|---|---|---|
| Visitor intent | Several distinct next-steps, roughly equal weight | One dominant path, chosen regardless of alternatives offered |
| Design response | Surface all options for inclusivity | Distraction from the path visitors already wanted |
| Winning move | Add more choice | Remove the choice nobody was taking |
The homepage was designed for a visitor who barely showed up
Most homepage redesigns start from a reasonable-sounding premise: visitors arrive with different needs, so the page should offer different doors. A large energy retailer's homepage followed that logic — the primary signup path sat alongside several other next-step options, each built for a plausible-sounding visitor segment who might want more information, a different product angle, or a slower path to purchase before committing.
The instinct isn't wrong in the abstract. Segmentation theory says visitors arrive with heterogeneous intent, and a page that only serves one intent will lose the others. The problem is that this instinct gets applied without checking whether the assumed segments are actually there in the volume the page design implies. A homepage can be architected around four or five hypothetical visitor types and still be functionally serving one, if that's what the traffic actually does. Heatmap analysis on this page showed exactly that: the large majority of visitors, regardless of which alternate pathway the page offered them, gravitated toward the signup path anyway. The alternate doors existed. Almost nobody walked through them.
That's the decision at stake in this experiment, and it's a decision every growth leader eventually faces: when your qualitative data shows visitors converging on one path despite being given several, do you take that as evidence the other paths are underused and need better placement — or as evidence they were never the paths visitors wanted, and the page should stop pretending otherwise? The team here took the second read, and the reasoning behind that call is the real substance of this case study.
Choice architecture doesn't reward more options — it rewards the right one
The behavioral principle at work is choice architecture — how the presentation of options shapes the decision itself, independent of what the options actually are. Richard Thaler and Cass Sunstein's foundational treatment of the concept, in _Nudge_, makes the point that every choice environment is designed whether or not anyone designed it deliberately, and that adding options is not a neutral act — it changes what gets chosen, including by making the "obvious" choice harder to find among the noise of alternatives that look equally legitimate. A homepage with five next-step options isn't a more generous choice environment than one with a single clear path; it's a differently designed one, and the heatmap data here suggested the multi-path design was the worse of the two.
The second mechanism compounds the first: commitment and consistency, the behavioral principle Robert Cialdini documented extensively in _Influence_ — once a person makes a small, active choice, they tend to behave in ways consistent with that choice going forward, including following through on a larger commitment they might otherwise have abandoned. This is why the competitive research mattered as much as the heatmap did. Visitors who personalize a plan choice early — selecting a usage level, a household size, a service preference, anything that counts as an active decision — show measurably higher downstream commitment to completing the purchase than visitors who arrive at the same information passively. The alternate pathways on the original homepage weren't offering visitors a different kind of choice; they were offering a way to delay the one choice that actually drove commitment.
Put those two mechanisms together and the redesign logic follows directly: a homepage that routes visitors toward an early, active, plan-specific choice — rather than offering a menu of equally-weighted browsing options — should outperform, because it is the version of the page that both matches observed intent and triggers the psychological mechanism that predicts follow-through. That's not a hunch. It's the specific combination of two well-established, independently-replicated behavioral principles applied to a specific, already-observed pattern in the traffic.
Why the heatmap came before the hypothesis, not after
The methodology choice here is worth naming explicitly, because it's the part most teams skip. The team didn't start with a hypothesis and go looking for data to support it — they started with the qualitative diagnostic and let it constrain what hypothesis was worth testing at all. Heatmap analysis is not a substitute for an A/B test; it can't tell you what will happen if you change the page. What it can tell you, reliably, is what visitors are actually doing on the page as it exists right now — and that's a different, earlier question than "what should we test," one that most experimentation programs skip past on the way to running a test.
That ordering matters more than it sounds. Running the qualitative pass first meant the team wasn't testing a redesign built on an assumption about visitor segments — they were testing a redesign built on an observed pattern in how visitors already behaved, then reinforced with competitor research on why that pattern (early personalization, higher completion) tends to hold. A hypothesis built this way survives a null result more gracefully than a hypothesis built on a stakeholder's intuition about what visitors "should" want, because even a flat or negative result tells you something specific about the diagnostic, not just that a guess didn't pan out.
This is the senior-practitioner move that separates a mature experimentation program from one that's just running variants: qualitative diagnostics inform which quantitative test is worth running. They don't get run after the test to explain why it won or lost. By the time this experiment shipped, the team already had directional confidence from the heatmap and the competitive research — the A/B test was there to confirm the magnitude and de-risk the decision, not to go fishing for a pattern that might or might not exist.
The result: a 5% to 10% lift, and the page stopped hedging
Exp-051 ran for about six weeks on the homepage. The variant — stripped of the distracting alternate pathways, rebuilt around the single user intent conversion path the heatmap had already surfaced — won, producing a lift of 5% to 10%. The result was clean enough, and the qualitative reasoning behind it solid enough, that the variant shipped as the new default rather than staying in a holdout or requiring a second confirmatory round.
What's notable isn't the size of the lift — ranges like this show up across plenty of homepage tests, for plenty of reasons. What's notable is what the result confirmed: matching the page to the path the data showed visitors already wanted outperformed offering more apparent choice, in a case where "offer more choice" is usually treated as the safer, more defensible default. That's the finding a founder or CMO should take away from Exp-051 — not the specific number, but the direction it points.
The diagnostic catch most teams stop short of
Here's the part that most teams — and most homepage redesign playbooks — get backwards. The default instinct when a homepage isn't converting well is to add: add a path for the segment you think you're missing, add a comparison tool, add a "not sure yet, learn more" option, on the theory that more doors serve more visitors and removing any of them risks alienating someone. That instinct treats optionality as inherently safer and more inclusive than restriction.
The heatmap data in Exp-051 said the opposite. The apparent choice on the homepage wasn't serving a real segment of visitors who wanted it — it was distraction sitting on top of a path the large majority of visitors were already choosing, alternatives offered or not. Most teams stop at "the page has several options, so it should be serving several intents" and never check whether the traffic actually splits that way. The diagnostic catch here — reading the heatmap as evidence about what to _remove_, not just what to _add_ — is the difference between a page that looks thorough and a page that converts. Removing the alternate pathways wasn't a simplification for its own sake; it was the page catching up to intent the data had already shown, several weeks before the redesign existed.
FAQ
Why not just add a clearer call-to-action to the existing paths instead of removing them?
Because the heatmap evidence wasn't that the existing paths were poorly labeled — it was that visitors weren't choosing them regardless of prominence. A clearer label on a path visitors don't want doesn't change the underlying intent; it just makes the distraction easier to find. The diagnostic pointed at removal, not emphasis.
Isn't offering more pathways safer for visitors who genuinely have different needs?
Only if a meaningful share of visitors actually use those pathways. Heatmap and behavior data are exactly how you check that assumption instead of guessing at it. In this case, the "safer" multi-path design was quietly working against the majority's already-established behavior — optionality that looks inclusive on a wireframe can still be a net drag on conversion in practice.
How do you know the lift wasn't just a seasonal or traffic-mix fluctuation?
The experiment ran as a controlled A/B test over roughly six weeks, which is the mechanism that isolates the redesign's effect from seasonal or mix noise in the first place — that's precisely why the heatmap and competitive research fed a controlled test rather than being treated as sufficient evidence on their own. The qualitative work told the team what to test; the test itself is what confirms the effect is attributable to the change, not the calendar.
Does this mean homepages should always minimize options?
No — and that's an important distinction. The finding here is specific to a page where the data showed one dominant user intent conversion path and several unused ones. A homepage genuinely serving multiple high-volume segments would show that in the same kind of diagnostic, and the right move there is different. The principle isn't "fewer options always win." It's "test what the data shows visitors actually do, and don't assume the page's original design got the segmentation right."
What would you look at first on a homepage that feels like it's underperforming?
The same two things paired here: qualitative behavior data (heatmaps, session replay, click distribution) to see what visitors already do regardless of what the page offers, and outside evidence on what predicts downstream commitment for the specific mechanism you're considering. Guessing at a redesign without that pairing is the single most common way homepage tests waste a quarter.
Bottom line
A homepage designed to accommodate every plausible visitor intent isn't automatically the more thoughtful design — it can just as easily be a page hedging against a segmentation assumption nobody checked. Exp-051 held a 5% to 10% lift by doing the less intuitive thing: reading the heatmap as permission to remove, not just add, and pairing that qualitative signal with competitive evidence on commitment before writing the hypothesis. The generalizable lesson isn't about homepages specifically — it's that apparent choice and real choice are different things, and the gap between them only shows up if you go looking for it before you test.
If your homepage — or any high-traffic page in your funnel — was designed around assumptions about visitor intent that haven't been checked against what the data actually shows, that's the kind of diagnostic work I run before I ever propose a test. Let's talk about what your data already knows.
Evidence sources and free next step
Baymard's homepage and category-navigation research shows why observed navigation behavior can inform a hypothesis but cannot replace a controlled test. Use the landing-page navigation review and A/B testing examples evidence table to grade the two evidence types separately. Then get started with GrowthLayer free to connect research observations to the eventual experiment.