We Removed Our Signup Input Field to Reduce Friction. It Backfired
The hero section on a SaaS product I work on has a classic layout: a headline, an input field where you type something about yourself (in our case a URL), and a button. To start, you fill the field and click.
Every mobile UX article I'd read that month said the same thing: that input field is your biggest enemy. Typing on a phone keyboard is the highest-friction interaction on the page. Replace the field-plus-button with a single fat thumb-friendly button and let people type later, inside the app, once they're committed.
So we tested it. A mobile-only experiment: the current stacked input-and-button versus a single "start free trial" button with no input at all. We also ran a desktop version as one arm of a larger CTA test, same idea: button only, no field.
The button lost. On both devices. The version that asked visitors to do keyboard work on a phone, before they'd even signed up, beat the version that asked for a single tap.
The theory we were testing
To be fair to past me, the theory is coherent. Conversion optimization has a hierarchy-of-effort model baked into it: a tap costs less than a keyboard entry, so a page that requires only a tap should convert more visitors into the funnel. Front-load nothing, collect everything later. Countless teardown posts point at landing pages with input fields and prescribe removing them.
And the button variant did do part of its job. More people interacted with the hero. Taps are cheap, so we got more of them.
Then everything downstream got worse. Fewer of those visitors made it through signup, and fewer still reached a trial. The single most "frictionless" hero we could build produced the least conversion of the options we tested. Roughed out, the input-field control outperformed the button variant by a meaningful margin on the metric we actually cared about, and it wasn't close enough to call noise.
Why the input field wins: it's not friction, it's the hook
The mistake was categorizing the input field by its interaction cost instead of by its function. Three functions, in our case:
1. It's a micro-commitment. A visitor who types something about themselves into your hero has started using the product before signing up. That tiny investment changes their relationship to the next screen. They're no longer evaluating a marketing page; they're finishing something they started. Clicking a bare button carries no such momentum. The tap is cheap, and cheap actions are cheap to abandon.
2. It personalizes everything downstream. What the visitor types is the seed for the whole first-run experience. When they arrive in signup, the product already knows something concrete about them, and every subsequent screen can prove it was listening. The button variant delivered visitors into a generic flow that had to start the conversation from zero. We had, in the name of removing friction, removed the information that made the funnel feel personal.
3. It self-qualifies. Typing into the field is a small declaration of intent. The people willing to do it are disproportionately the people the funnel converts well. The button collected intent-free taps: more entrants, worse entrants. If you only measure clicks on the hero, this looks like a win. It's a win the same way inflating your top-of-funnel with an ad for free money is a win.
The measurement trap that almost fooled us
This experiment is a case study in why you never judge a landing page test on its first metric. The button variant genuinely did increase hero engagement. A dashboard scoped to "CTA interaction rate" would have declared it the winner and we'd have shipped a worse funnel while high-fiving.
The ranking only flipped when we followed both cohorts to the end: through signup, through onboarding, to trial start. Everything I've written before about reading the whole funnel applies double to landing pages, because the top of the page is where vanity metrics live.
If you take one operational rule from this: the metric for a hero test is trials (or revenue), never taps.
Reconciling this with "reduce friction"
I wrote recently about how shortening our onboarding lost twice. This is the same lesson wearing a different shirt, so I'll draw the distinction again, because it keeps earning its keep:
Dead friction does no work for the user or the product: broken layouts, redundant questions, slow pages. Remove it and conversion goes up, every time we've tested it.
Productive friction does work: it builds commitment, collects the seed for personalization, filters for intent. Remove it and you often remove the reason people convert.
An input field on a hero can be either. If your field asks for something the visitor doesn't care about and the product doesn't visibly use (a phone number for a demo, say), it's probably dead friction and the teardown posts are right. Ours asked for the one thing the visitor cares most about, and the product used it immediately and visibly. That's productive friction, and cutting it cut the funnel off at the knees.
The way to know which one you have is not to reason about it in a meeting. It's a two-variant test and a couple of weeks.
What we'd tell you to steal
Don't remove an input field just because a checklist told you to. Ask what the field feeds. If the answer is "the entire personalized first-run experience," touch it carefully.
Judge hero tests at the bottom of the funnel. Engagement on the hero is the easiest metric to move and the least meaningful.
Watch for the qualification effect. Any change that makes entering your funnel easier changes who enters it. Volume up, quality down is the default outcome, not the exception.
Run the mobile and desktop versions separately. We assumed the effect would differ by device (that was the whole premise). It didn't, and knowing that it replicated across both is what made us trust the result.
The frictionless page is a seductive idea: every barrier gone, nothing between the visitor and the product. But a landing page isn't a barrier to your product. Done right, it's the first screen of your product. Ours starts with a question, and it turns out people like being asked.