Your trial signup page is a validator. Everything vague in your positioning compiles fine right up until someone hits that form, and then it throws.
Which is why most advice about these pages is unhelpful. It is about button colours and social proof widgets, when the actual failures are upstream, in what you are asking someone to believe before they hand over an email address. Here is what the traceable evidence supports, and where it runs out, because a lot of what circulates on this topic is invented.
The benchmark, and the trap inside it
The best-documented recent dataset comes from ChartMogul, with Kyle Poyar and ProductLed: 200 self-serve business software products surveyed in January 2026, median revenue between one and ten million dollars annually, free-to-paid measured within six months (sourced from ChartMogul).
The median free-to-paid conversion was 8%. The more useful finding is the spread: a tenfold difference between the top fifth and the bottom fifth. Their benchmarks by model:
Freemium: good is 3 to 5%, great is 8 to 12%.
Free trial, no credit card: good is 4 to 6%, great is 10 to 15%.
Free trial, credit card required: good is 25 to 35%, great is 50 to 60%.
That last line is where people lose their heads. Card-required trials report about 30% conversion, more than five times the no-card figure, and the obvious conclusion is to add a card field on Monday morning.
DO NOT! This is survey data from self-selected products, and the comparison is almost certainly confounded. Products confident enough to gate a trial behind a card tend to have higher-intent traffic and higher contract values already. The card does not multiply your conversion. It moves your qualification step earlier, so a smaller number of better-qualified people reach the trial and a higher percentage of them buy. Your absolute revenue may go up, down, or nowhere. The percentage will look fantastic either way, which is exactly why it is a dangerous number to chase.
Worth knowing too: the credible datasets disagree with each other. ProductLed's own benchmark set puts freemium ahead of free trial, the reverse of the ChartMogul ordering, most likely because they define the metric differently. Anyone presenting either as settled has not read both.
What the page is up against
Two numbers frame the job. Unbounce's conversion benchmark study, covering more than 41,000 landing pages between 23 July 2023 and 23 July 2024, put the median landing page conversion rate at 6.6% across all industries and 3.8% for software (sourced from Unbounce). Those are Unbounce-hosted pages taking mostly paid traffic, so treat it as a floor for a cold-traffic page rather than a target for your homepage.
The second number is the one that should change how you write. TrustRadius surveyed 1,862 technology buyers in January 2026 and found that 83% shortlisted three products or fewer (sourced from TrustRadius). Gartner, surveying 646 business buyers between August and September 2025, found 67% prefer a buying experience with no sales representative involved at all (sourced from Gartner).
Put those together. Most of your buyers are building a three-item shortlist, most of them would rather not talk to you while doing it, and your signup page is doing the work a sales conversation used to do. It is not a form. It is the pitch.
Publish the price
TrustRadius reports that transparent pricing has been buyers' number one wish-list item for vendors four years running, and in its 2024 survey of 2,164 buyers, 53% agreed with "I wish all vendors had transparent pricing."
Now the honest caveat, because I looked for the experiment and it does not exist. I could find no published controlled test showing that publishing prices increases qualified signups. The survey evidence that buyers want pricing is strong and consistent. The experimental evidence that showing it makes you money is absent. Anyone quoting you a percentage lift from removing "Contact us for pricing" made it up.
What you can reason about is the shortlist. If 83% of buyers are picking three products and most of them will not call you to find out what you cost, an unpriced page is asking to be left off a list you never see. A published range, with the conditions that move it, is usually enough. You do not have to commit to a single number to stop being unshortlistable.
Stop obsessing over form fields
This is the received wisdom most worth throwing out. Zuko, a form analytics company, published an analysis of 739 forms covering 20,000 labelled fields and 93 million form views, and concluded that "the length of a form makes almost no difference to the rate at which people complete it" (sourced from Zuko). Their mix leans toward ecommerce and lead forms rather than software trials, so weigh it accordingly, but it is real data at real scale against an idea nobody has ever tested at that scale in the other direction.
The other pillar of the field-cutting orthodoxy is a HubSpot analysis of more than 40,000 landing pages, and its actual finding was that as fields increase, "conversion rates decrease slightly, but not as steeply as I expected." Every blog citing it has inverted the conclusion. What the same analysis did find is that the type of field matters more than the count, with multi-line text areas and dropdown menus showing the strongest negative association.
So the question is not how few fields you can get away with. It is whether each field is answerable in two seconds without the person having to go and look something up. "Company size" as a dropdown they have to think about costs you more than two extra text inputs they can fill in without pausing.
The technical floor
None of the above survives a page that hangs. Google's current Core Web Vitals thresholds are a Largest Contentful Paint of 2.5 seconds or less, an Interaction to Next Paint of 200 milliseconds or less, and a Cumulative Layout Shift of 0.1 or less, measured at the 75th percentile of page loads and split between mobile and desktop (sourced from Google). You pass only if you clear all three.
Google says these "are used by our ranking systems," and in the same breath that "there is no single signal." So this is a floor, not a growth lever. A signup form that shifts under someone's thumb on a phone loses conversions long before it loses rankings. If you have never checked, the free website audit walkthrough covers how, and the mobile-first piece covers what usually breaks.
What to test when you cannot actually test
Here is the advice that almost nobody gives founders, and it is the most useful thing in this article. If you get a few hundred signups a month, you cannot run a valid A/B test on trial conversion. The maths does not work. You will reach a "winner" that is noise, ship it, and wonder in six weeks why nothing moved.
So do this instead, in this order.
Instrument the form itself. Field-level analytics tell you exactly which input people abandon on. This is a fact, not an inference, and it needs no sample size to be true.
Watch recordings. Microsoft Clarity is free and PostHog has a free tier. Twenty sessions of real people failing at your form will teach you more than a month of a badly powered experiment.
Call ten people who signed up and never activated. Not a survey. A call. The reasons are usually embarrassingly concrete and almost never the button colour.
Fix the obvious things, measure before and after over a long enough window, and report it as a direction with wide error bars rather than as a declared winner at 95% confidence.
Come back to real experimentation when your volume supports it. GrowthBook, Statsig and PostHog all have free tiers. And if you inherited a page built on Google Optimize, note it was shut down on 30 September 2023 and Google now points to AB Tasty, Optimizely and VWO instead (sourced from Google).
Before you send the first onboarding email
A trial signup is not just a conversion event. It is the start of a commercial email relationship, and if any of your signups are Canadian, Canada's Anti-Spam Legislation (CASL) applies to every message that follows.
A trial signup generally gives you implied consent through an existing business relationship. Implied consent is time-limited and narrow: it covers messages related to the relationship the person actually entered into, not your general marketing list. Whatever the basis, every commercial electronic message needs clear identification of who is sending it and a working unsubscribe mechanism that keeps working. The practical version is to keep onboarding and product email separate from marketing sends, and to get express consent before you move someone across.
This is general guidance rather than legal advice, and anything consequential is worth putting in front of counsel. The reason it belongs in an article about a signup form is that the form is where the consent question is decided, and it is much cheaper to design it correctly than to unpick it eighteen months later.
The uncomfortable question
Most trial pages I look at are not badly designed. They are precisely designed around a value proposition the founder has not finished writing. The form is fine. The sentence above it is doing no work. That is not a page problem, and running the marketing audit first usually finds it faster than another round of form tweaks. Finishing that sentence is positioning work rather than design work, which is why redesigning the page rarely fixes it.
So before you touch the page: can you say, in one sentence a stranger would understand, what someone will be able to do fourteen days from now that they cannot do today?