Figma's Own Trial Finds Figma Make Mostly Helps PMs, Barely Designers

A product manager wants one more item in the settings menu, and asks for a mockup by this afternoon. Every product designer knows that request, and until recently, answering it meant opening a file, matching the existing pattern, and sending back a frame. Figma has now published the paper it commissioned to test what happens once the PM can skip that step: a randomized controlled trial of its own prompt-to-design tool, Figma Make , run with the research firm MeasuringU and posted to arXiv on 22 September 2026 . Fifty product designers and fifty PMs did the same three tasks, half of each group with the tool, half without. The tool Figma sells to designers moved their numbers only marginally, and mostly on one hard task. It saved the PMs more than a third of their time, robustly and across the board. That is the tension sitting inside the company’s own data: a design tool whose measured beneficiary is the person who used to have to ask a designer for help.

The stopwatch never asked who decided the design

The trial’s mechanics explain why. Each of the three tasks — a dark-mode conversion, adding a menu item, and a comment flyout — had its correct answer fixed before the participant ever opened the tool. As the paper’s moderation guidelines put it:

Moderators started the timer at the participant’s instruction and stopped the timer after verifying, according to the moderation guidelines, that the participant’s design matched the reference image.

A task counted as finished only when the output matched that reference, and only completed tasks were timed; quality itself was never scored, only self-rated by participants afterward. What the trial measured, in other words, was how quickly someone can reproduce a design someone else already approved. Read across to what “design work” means for the people using the tool, that setup implies the PMs’ gains are gains at execution, not at the judgment calls — what to build, how it should look, what tradeoff it makes — that designers are actually paid for.

Where the thirty-five percent actually landed, and where the seventeen percent gets slippery

The Task 2 breakdown makes the split concrete: adding a “Help & Support” option to a dropdown settings menu, the exact small request designers field every week. PMs with Figma Make finished it 4 minutes 34 seconds faster, a 39% saving; designers saved a non-significant 1 minute 10 seconds. Across all three tasks, PMs saved 22 minutes 46 seconds cumulatively — a 35% reduction, comfortably significant. Designers’ cumulative number is where the paper gets slippery with itself: its Results section reports a 17% overall time reduction for designers (AME=7:30, SE=3:49, p=0.049), just inside the conventional 0.05 threshold, while its Discussion section states: “We did not find a statistically significant reduction in cumulative completion time among designers.” Both readings are in the same paper — designers’ overall gain is real but fragile in a way PMs’ 35% never is. Figma’s blog post from August put the figures at 23% for PMs and 20% overall for designers, from a different aggregation; the fuller arXiv paper is the one to trust. For a working designer, the shift isn’t abstract: the PM who used to file a ticket now builds the change themselves, and the designer’s job moves from producing that screen to judging the one someone else produced — a judgment call the trial never measured, and one that a recent survey of 900-plus designers found teams aren’t yet equipped to evaluate or reward.

The honest complications the paper doesn’t hide

The study names its own limits, and they cut against a tidy verdict. Random rendering failures inflated treatment-group times, meaning the true gap may be larger than reported. Designers posted a significant win on exactly one of the three tasks — the hardest, a comment-flyout interaction prototype — saving 26%, or 7 minutes 41 seconds; the other two tasks showed no significant gain, consistent with why the cumulative figure sits right on the edge of significance. That pattern suggests the tool may help designers most where a design has to be worked out rather than copied, which fits Figma’s framing, echoed across the prompt-to-design tools racing toward production : PMs absorbing trivial changes frees designers for harder problems. The counter-argument has nothing to do with what this trial measured. Writing in Smashing Magazine, Andy Budd argues that “plausible design is dangerous. It looks coherent in a product review… nobody in the meeting feels strongly enough to object. So it ships.” Nielsen Norman Group’s Anna Kaley and Raluca Budiu make a related claim : “production has become cheaper than UX evaluation.” Neither claim is measured by this trial; they describe what cheap, plausible output does to organizations that already treat design as decoration — and the tension in Figma Make’s own numbers, where reproducing beats deciding and PMs hold the tool that’s fastest at it, is exactly the soil those arguments grow in. It also complicates the properties-panel controls Figma Make shipped back in August , which route every “direct” edit back through the same prompt.

Figma’s own study didn’t set out to settle that argument, and to its credit it says so. But it did settle something smaller and more immediate: the mockup a PM used to wait for a designer to deliver, they can now build themselves, timed and verified against a picture someone else already drew.