Can Mobile Conversion Optimization Be as Simple as Deleting Copy?

Deleting a few sentences of explanatory copy from a mobile modal produced a double-digit conversion lift — proof that on some pages, the explanation is the confusion.

TL;DR

• On a mobile zip-code entry modal in a large energy retailer's pricing/enrollment flow, we removed the explanatory text describing what would happen after zip entry. Nothing else changed.
• The stripped-down variant won by one of the widest margins in this experiment set: a lift range of 10% to 20% (Exp-050).
• The mechanism is cognitive load and the simplicity effect — added text doesn't automatically resolve ambiguity. On a high-traffic, low-attention mobile micro-interaction, added text can _be_ the ambiguity.
• Isolating one variable (the copy) on one micro-interaction (the modal) was a deliberate methodology choice, not a shortcut. It's the reason we can attribute the lift to a specific mechanism instead of a bundle of unrelated changes.
• The strategic takeaway isn't "write less everywhere." It's that decision-critical friction points deserve isolated diagnostic experiments, because intuition about what counts as "helpful" is frequently wrong.

The decision nobody wants to spend three weeks on

Here's the pitch a growth lead has to make internally: "I want to run a three-week experiment on a few sentences of copy in a modal." Most stakeholders will ask why that's worth the calendar time. It's a fair question, and it's the wrong instinct to act on.

The modal in question was the first micro-interaction in a mobile pricing and enrollment flow — a zip-code entry screen at a large energy retailer. Every visitor who wanted a quote touched this screen before touching anything else. That's what made it worth the rigor: unlike a mid-funnel page that only a fraction of traffic ever reaches, a top-of-funnel micro-interaction is seen by effectively everyone. A small effect at that position compounds across the entire flow beneath it, because every downstream step is conditioned on the visitor getting through this one first.

The size of the change — a few sentences — is exactly why it's tempting to skip testing it. Copy edits feel too minor to justify a dedicated experiment cycle, and teams default to shipping small text changes on instinct. The size of a change and the size of its effect are not correlated, and a checkout page's cognitive load is set at exactly this kind of small, easy-to-overlook moment. Treating "it's just a sentence" as a reason not to test is how teams accumulate friction they never measure.

There was also a legitimate hypothesis behind the original copy, not a careless oversight. Whoever wrote the explanatory text believed it would reduce ambiguity about what zip entry would trigger next — a reasonable design instinct. The experiment existed to check whether that instinct held up against actual user behavior, not to fix a mistake.

The mechanism: cognitive load and the simplicity effect

The behavioral principle at work is cognitive load — the idea, formalized by John Sweller in his research on cognitive load theory, that working memory has narrow, fixed limits, and every additional piece of information competes for that same finite processing capacity. Each sentence a user has to read before acting is not free; it's a withdrawal from a small, shared attention budget.

Steve Krug made the practitioner version of this argument in _Don't Make Me Think_: every extra word on a page is one more thing a visitor has to parse, and parsing is the exact activity you're trying to minimize between a user's intent and their next action. On desktop, with more screen real estate and a slower-paced browsing posture, a few extra sentences might get skimmed without cost. On a mobile modal, where screen space is compressed and the interaction is often a thumb-scroll away from being abandoned, the same sentences carry a heavier tax.

There's also a decision-science layer here worth naming precisely: added information doesn't only cost processing time, it can independently lower decision confidence. Research on preference fluency (Novemsky, Dhar, Schwarz, and Simonson, _Journal of Marketing Research_, 2007) found that increasing the perceived difficulty of a decision — even by adding accurate, relevant information — can reduce a person's confidence in their choice and increase the odds they defer or abandon it. The explanatory text in this modal wasn't wrong. It was accurate. That's precisely why the finding matters: more accurate information is not the same thing as less cognitive load, and conflating the two is the single most common misread of what actually helps a user convert.

Put together, this is the simplicity effect at work on a checkout-adjacent micro-interaction: reducing the number of things a user has to hold in mind before acting can outperform adding clarity, because the clarity was never the bottleneck. The bottleneck was processing capacity.

Why we isolated one variable instead of redesigning the modal

There was an easier, more satisfying option on the table: redesign the whole modal. New layout, new button copy, a progress indicator, maybe a friendlier icon. It's the version of this project most teams would default to, because a full redesign feels like more work is being done to earn the result.

We didn't do that, and the reason is methodological, not stylistic. If we'd changed the layout, the copy, and the visual hierarchy simultaneously, a win would have told us almost nothing about _why_ it won. Any lift could have come from the new layout, the shorter copy, the different button treatment, or some interaction between all three. We would have shipped a better modal without learning anything transferable to the next flow, the next client, or the next micro-interaction on this same site.

Isolating a single variable — removing explanatory text and changing nothing else — meant the result could be attributed to one specific mechanism: a reduction in cognitive load at a high-traffic decision point. That's a finding a team can carry into other parts of the funnel, other products, other channels. A bundled redesign produces a one-off artifact. A single-variable experiment produces a piece of evidence about how users actually process a checkout page's cognitive load under real conditions — evidence that compounds in value the more places you can apply it.

This is a judgment call that requires resisting the urge to "improve everything while you're in there." Rigor here means intentionally shipping the less exciting version of the project — one variable, one hypothesis — because the payoff is a clean, reusable insight instead of an untraceable win.

The result

The variant with the explanatory text removed won, and by one of the largest margins across this entire experiment set: a lift range of 10% to 20% (Exp-050). Removing text from a single, high-traffic micro-interaction outperformed the original hypothesis that more clarification would help. The result was strong and consistent enough that the variant was scaled to full traffic.

The diagnostic most teams stop short of

Here's the part of this experiment worth sitting with, because it cuts against the default instinct in most marketing and product organizations. When a team suspects users are confused at a step in a flow, the reflexive response is to add explanation — a tooltip, a sentence of reassurance, a "here's what happens next." That instinct treats confusion as an information gap, something more words can close.

The diagnostic catch here is recognizing that on a high-traffic, low-attention mobile micro-interaction, added text can itself be the source of the confusion, not the cure for it. The explanatory copy in this modal wasn't ambiguous or poorly written — it was competing for a sliver of attention that the user hadn't budgeted for reading anything beyond the zip field itself. The text raised cognitive load precisely at the moment the user wanted to act, not think, and that added friction registered as hesitation rather than clarity.

Most teams stop at "users seem confused, let's explain more." Diagnosing this correctly means asking a harder, second-order question: is this a genuine information gap, or is it a low-attention moment where any additional processing demand — however well-intentioned — is itself the obstacle? Those two diagnoses call for opposite interventions, and getting the diagnosis wrong means you ship more of the thing that's hurting you.

FAQ

Does this mean we should strip explanatory copy out of the entire funnel?

No. The finding is specific to a high-traffic, low-attention, early-funnel micro-interaction on mobile. Deeper in a flow, at higher-consideration decision points, or on desktop, added context can still reduce hesitation rather than cause it. The transferable lesson is the diagnostic question — is this an information gap or a processing-capacity problem — not a blanket "delete copy" rule.

How do you know the lift wasn't just noise, given it was a small change?

The change was small; the traffic through that modal was not. Every visitor entering the pricing flow passed through this exact interaction, which gave the experiment strong statistical footing despite testing only one variable. The margin here was also one of the widest in this experiment set, which is a further signal it reflects a real behavioral effect rather than sampling variance.

Why not just redesign the whole modal while you were running the experiment anyway?

Because a bundled redesign would have made the result uninterpretable. A win on a multi-variable redesign can't tell you which change drove it. Isolating the copy as the single variable is what let us attribute the lift to a specific, reusable mechanism instead of a one-time page improvement.

Does a result like this generalize outside mobile or outside this industry?

The mechanism — cognitive load competing with conversion at a high-traffic, low-attention decision point — shows up across mobile flows generally, and in any vertical where an early funnel step asks for a small piece of information before the value exchange is clear. The specific lift range is tied to this flow and this audience; the diagnostic approach is what transfers.

How long should a team expect to run something like this before trusting the result?

This experiment ran about three weeks, which was long enough for the modal's high traffic volume to produce a stable, wide margin. Duration should be set by the volume moving through the interaction and the size of effect you need to detect confidently, not by an arbitrary calendar default.

Bottom line

A few sentences of accurate, well-intentioned copy were adding more cognitive load than clarity at the single highest-traffic step in a mobile pricing flow, and removing them produced one of the strongest results in this experiment set. The lesson isn't to write less everywhere — it's that "more explanation reduces confusion" is an assumption, not a law, and the only way to know which side of that assumption you're on is to isolate the variable and test it.

If you're deciding whether your team's growth experiments are generating real evidence or just shipping opinions with a p-value attached, that's the kind of program design and experimentation leadership I take on with founders and growth teams directly. If you want to talk about building a rigorous experimentation practice inside your organization, get in touch.

Evidence sources and free next step

Nielsen Norman Group's cognitive-load guidance helps explain the mechanism without pretending it predicts this result. Compare the case with the content testing framework and sample-size guide. Then try GrowthLayer free to predeclare the copy change, transaction metric, and stopping rule.