You know your pricing page is confusing.
So does everyone who leaves it.
They don't file a complaint. They don't send a note. They leave, and the analytics call it "drop-off," which is a soft word for a hard thing.
The median free-trial-to-paid conversion rate in B2B software is 8%. The median is polite. The spread is not. In a January 2026 analysis of 200 B2B software products by Userpilot with ProductLed and ChartMogul, 20% of products converted below 2.5%. 23% converted above 25% ([Userpilot, 2026](https://userpilot.com/blog/saas-average-conversion-rate/)).
A 10x gap.
Two companies buy the same traffic. Run the same spend. Build functionally similar pages. One converts ten times more than the other. The traffic didn't decide that. Something on the page did.
Behavioral science — the field that studies how people actually make decisions, not how they say they do — has spent fifty years documenting one uncomfortable fact: most decisions happen on autopilot.
Your brain has a fast mode and a slow mode. The fast mode reacts instantly, effortlessly, without you noticing it's working — it's the part that flinches at a loud noise and knows a face is angry before you can say why. The slow mode is the careful one: it reads, compares, calculates. And it's lazy. It only shows up when the fast mode can't handle things alone. This model earned a Nobel Prize in economics, and it has been tested in hundreds of experiments since.
Your visitor is not reading your page. Not carefully. The autopilot is scanning — reacting, not evaluating — and the careful reader gets called in only if the autopilot doesn't find a reason to leave first.
You built the page for the careful reader. The careful reader shows up late, if at all.
What follows: seven mechanisms, each grounded in a published study, each a way the snap judgment finishes before careful thinking starts. Then one more finding that changes what the seven add up to. It can wait. It's good at waiting.
Song and Schwarz, 2008, *Psychological Science*. Exercise instructions printed in a clean font or a hard-to-read script. Same words. The hard-font readers judged the routine longer and harder (Song & Schwarz, 2008).
The font did that. Not the exercise. The font.
If your page is hard to read — dense copy, cluttered layout, unclear next step — the product feels hard to use. The visitor hasn't used it. The feeling doesn't care.
Thorndike, 1920, *Journal of Applied Psychology*: officers rated 137 aviation cadets on supposedly independent traits. The ratings moved together — good at one thing, good at everything, and the reverse (Thorndike, 1920). He called it a constant error. It has stayed constant.
One typo rates your security. One screenshot rates your support team. Nothing on a page is judged alone.
Shah and Oppenheimer, 2008, *Psychological Bulletin*: people take mental shortcuts because thinking itself is a cost — searching, reading, weighing options all burn energy (Shah & Oppenheimer, 2008). People avoid effort the way they avoid pain. It is a mild pain.
Expedia had an optional field on its booking form: "Company Name." Users typed their bank's name and address into it. Address verification failed. Cards failed. Expedia removed the field and made $12 million more a year — the account comes from Joe Megibow, then VP of Global Analytics and Optimization; UX Movement).
One empty box. Twelve million.
A 1974 study in *Science*: a rigged wheel of fortune stops on 10 or 65. Then a question: what percentage of African countries are in the UN? Wheel said 10 — guesses averaged 25%. Wheel said 65 — guesses averaged 45% (Tversky & Kahneman, 1974). The wheel knew nothing. It moved the answer 20 points anyway.
Related, and verified: behavioral economist Dan Ariely tested *The Economist*'s subscription options on 100 MIT students. Web-only $59. Print-only $125. Print-plus-web $125. With all three shown, 84% took print-plus-web and nobody took print-only. Remove the useless middle option, and 68% of a new group took the cheap one (Ariely, *Predictably Irrational*, 2008). A number or option seen first, or alongside, changes what "the right choice" looks like — independent of its merits.
The first price on your pricing page is not a price. It's the ruler every other price gets measured with.
A 1979 paper in *Econometrica* proved what every gambler already suspected: people don't judge outcomes in absolutes. They judge them as gains or losses from where they stand — and losses weigh roughly twice as much (Kahneman & Tversky, 1979). The finding was central to a Nobel Prize in 2002.
"Start your free trial" is a gain. Gains are light. What a visitor loses by waiting weighs more — the asymmetry runs that way whether your copy uses it or not.
A 1973 study in *Cognitive Psychology*: people judge how common something is by how easily examples come to mind, not by actual frequency. Famous names made a list feel majority-male or majority-female regardless of the count (Tversky & Kahneman).
One named customer with one specific number beats ten generic claims. Not because it's representative. Because it's easy to remember, and easy feels true.
A 1981 study in *Science*: the same health program, described as lives saved or lives lost. Preferences reversed (Tversky & Kahneman). Identical facts. Different words. Different choice.
HubSpot, analyzing landing pages across 40,000+ customers: generic "Submit" buttons underperformed buttons that named the benefit — Download, Get, Claim. "Submit" describes what the form does to you. That may be the problem.
Not a new study — the shape of the seven above. In each, the snap judgment finishes before careful thinking starts, and the snap judgment is moved by something irrelevant to the decision: a font, a wheel, a famous name, a word on a button. None of these studies were run on SaaS websites. They were run on people. Your funnel is visited by people.
They leave anyway. You know where.
You've been reading this as seven leaks. Seven pages, seven fixes, seven line items.
Here is the finding that breaks that.
Redelmeier and Kahneman, 1996. 154 colonoscopy patients rated their discomfort at 60-second intervals during the procedure, then rated the whole thing afterward. The after-the-fact ratings tracked two moments — the worst one and the last one — not how long it took, not the average. How long the experience lasted barely registered in memory at all. In a related trial, patients whose procedure was deliberately extended with a gentler final phase remembered the longer procedure as *less* unpleasant, and came back for follow-ups more often (Redelmeier & Kahneman).
More total discomfort. Remembered as less.
Memory doesn't average an experience. It samples it twice: the worst moment and the last moment. Everything else is rounding.
So the visitor does not grade your homepage, then your pricing page, then your signup form — seven exams, seven scores. That's your spreadsheet, not their memory. They form one judgment of "that whole thing," and it is built from where the experience got worst and how it ended. Six clean pages and one confusing table: remembered as confusing. Friction early but a clear, confident final step: remembered as easy. It wasn't. Memory doesn't check.
Fixing seven problems across seven pages is real work. Finding the single worst moment and rebuilding the last one is where the work pays first.
None of these mechanisms were tested on a SaaS funnel. No study measured all of them acting together on a real one. What the research supports is narrower, and load-bearing: each is a documented, replicated way that irrelevant details move human judgment — and a funnel is built entirely of such details. Wording. Ordering. Visual weight. Sequence. And the details are not judged page by page. They are compressed into one memory with two authors: the worst moment and the last one.
Which leaves two questions worth asking before any others. Where does the friction spike hardest?
And the one that stings: what does your last screen feel like?
That's the one they remember.
You know your pricing page is confusing.
So does everyone who leaves it.
The difference, now, is that you know which two moments decide what they remember — the worst one, and the last one — and that both can be found.
Find your funnel's worst moment and its last moment — free, in about 10 minutes. [Get the Free Snapshot →]