Users abandon signup when the form demands more than your product has yet earned: too many fields, opaque password rules, verification walls before value. Most of the hard evidence below comes from checkout research, chiefly Baymard Institute’s, because nobody has studied SaaS signup friction with the same rigor. Checkout is the closest measured proxy, and we’ll say so each time instead of passing its numbers off as signup benchmarks.
For the scale of the problem: Baymard’s running average across 50 abandonment studies is 70.22%. Seven in ten people who started, gone. Your signup faces the same physics: low commitment, a tab bar full of alternatives, and every extra question a fresh exit. This teardown walks the 5 steps of the standard SaaS signup gauntlet: at each one, what products get wrong, the receipt, and the ticket-shaped fix.
The abandonment number everyone gets wrong
The most repeated stat about forced account creation, that 24 to 26% of users abandon because of it, comes from Baymard’s older survey rounds and is no longer the published figure. Baymard’s cart abandonment page, updated September 22, 2025, now reports that 19% of US shoppers who abandoned a checkout cite “the site wanted me to create an account” as a reason. Self-reported, multi-select, and the base excludes people who were just browsing.
Every post that quotes 26% copied a blog that copied a blog. Nobody clicked through to the source, which updated years ago.
The 24% figure is from Baymard’s 2021 to 2022 survey wave, and a later round said 26%. The current figure on Baymard’s own page is 19%. Still roughly 1 in 5 abandoners naming your account wall as a reason they walked, so the correction weakens nothing. It just makes you the only person in the meeting citing a live number.
Baymard · checkout survey, updated Sept 2025That habit of tracing numbers back to primary sources is a whole discipline. We did it for the other famous UX stats in where the famous UX numbers actually come from. Now, the gauntlet itself.
Step 1: the form asks for more than it needs
The first screen of a typical B2B signup wants first name, last name, work email, phone, company, company size, and a password. Baymard’s 2024 field benchmark found the average checkout contains 11.3 form fields across 5.1 steps, while most sites need only about 8. Their sharper finding: field count hurts usability far more than step count does.
The micro-failures are measured too: when name is split into 2 fields, 42% of users type their full name into the first one. One honest caveat: the field-count relationship is not linear, and CXL’s review of A/B data shows motivated users tolerate long forms. So the rule is not “fewer at any cost.” It is: nothing the first session doesn’t need.
And whatever fields survive should validate inline. Luke Wroblewski’s 2009 study with Etre measured inline validation against after-submit validation: +22% success, 22% fewer errors, +31% satisfaction, 42% faster completion. Small eye-tracking sample, not an academic trial, but nobody has published better since. One nuance from the same study: validate when the user leaves the field, not per keystroke, because mid-word red text reads as being yelled at.
Step 2: password rules revealed one failure at a time
The user types a password, submits, and only then learns about the uppercase rule. Submits again, meets the symbol rule. This is the step where products convert a committed user into a former one, and it is one of the best-measured frictions in the pack.
The same Baymard research puts abandonment from complex password-creation rules at up to roughly 19% of returning users, and their guidance is blunt: any requirement beyond a 6 to 8 character minimum will cost you completions. Every symbol rule you add today is a reset flow you force next month.
Your password policy is not securing accounts. It is training users to pick Password1! and to remember you as the product that made them feel stupid.
WCAG 2.2 made this a compliance line, not a taste question: no recall-and-transcribe hurdles, and paste must work. NN/g adds the disclosure rule (show constraints before the first attempt, not as error messages) and has argued since 2009 that masking with no reveal option hurts usability: “the only feedback they get is a row of bullets.” Mask by default, add a show-password toggle. ⚡ Quick win · ~1 hr.
WCAG 3.3.8 · Accessible authentication (AA)Step 3: the email verification wall
The user gave you an email and a password, and instead of the product they get “check your inbox.” Here is the honest version of this section: no rigorous public study quantifies how many users die at this wall. The 10 to 40% figures floating around come from email-verification vendors, who are selling the fix.
What is measured is the failure mode. The reset-email number above is Baymard documenting what happens when a flow stalls until an email arrives: delays, spam folders, the wrong inbox, a typo nobody caught (which is why Baymard suggests confirming the email, not the password). A verification wall inherits every one of those risks, at the exact moment NN/g’s login-wall research says users are most defensive: before you have shown them anything.
Step 4: the qualification quiz
The survivors now meet the survey: role, team size, use case, how did you hear about us, and a phone field for the sales team. Old but still directionally useful data from HubSpot’s analysis of 40,000+ landing pages (Dan Zarrella, circa 2010) found that which fields you ask matters more than how many: age, phone, and location questions measurably lowered conversion, with phone among the worst.
You met this user 40 seconds ago and you are already asking for their phone number. In any other context that gets you escorted out.
The quiz usually re-asks what the form already knows: company name when you have the work email domain, name fields collected twice. Since October 2023, WCAG 2.2 makes asking for the same information twice in a session a Level A failure. Auto-populate it or make it selectable. Every quiz question should either change the user’s first screen or be deleted. ⚡ Quick win · ~1 hr.
WCAG 3.3.7 · Redundant entry (A)Step 5: the empty first screen
The user survives all 4 steps and lands on a blank dashboard with a mute “Create your first project” button. This is where the win evaporates. In Lenny Rachitsky and Yuriy Timen’s 2022 survey of 500+ products, average activation was 34% and the median 25%. Self-reported, and every company defines activation its own way, but the shape is damning: most people who finish signup never reach value.
The empty state deserves its own audit (our fictional Driftly opens with “instantiate a workstream artifact,” dissected in the full heuristic evaluation walkthrough). The short rule: the first screen should already contain something, a template, sample data, one obvious action. A gauntlet that ends in an empty room retroactively makes every question you asked feel like a scam.
Score your own gauntlet
The usual interactive widget here would multiply survey percentages from unrelated studies and hand you a fake “you are losing 43% of signups” number. We are not doing that, because stacking a 2025 checkout survey on a 2009 eye-tracking study is astrology with extra decimals. Instead: toggle the frictions your signup actually has, and each one shows its real evidence, including the honest tag “magnitude unmeasured” where that is the truth.
Whatever you toggled, the output you want is not a score, it is tickets: severity, evidence, fix, effort, like the 3 above. The format is in how to turn UX findings into tickets devs actually fix, and the wider sweep beyond signup is the founder’s UX audit checklist.
One more honest note: you cannot grade your own gauntlet from memory, because you have not experienced it since you built it. An agent that actually attempts your signup catches these because it hits them: the wall, the reset email, the empty room. That is what ClapBack does with nothing but your URL.