We built a feature for 5% of our users. I checked that number after we shipped it.
Numbers first, because that's the whole post.
3,080 signups. Broken down by what people say they do:
Product planning — 1,561
Business / founder — 1,028
Other — 184
Design — 158
Development — 154
Designers are 5%.
Three weeks ago our team built a feature that reads an app's store listing and works out its screen structure how many screens, what sits under what, where the navigation goes. Paste a link, get a map.
I was proud of it. It even scores its own confidence and tells you which screens it guessed at.
Here's the thing I didn't do until this week: ask who it's for.
It's a designer's problem. You save a reference, you get pictures, you rebuild it by hand to understand it. That's a workflow I know intimately, which is exactly why I never questioned whether anyone else had it. The 84% who actually show up don't open Figma. They write the plan and hand it to someone else.
The distribution numbers say the same thing. I wrote about the feature four times across a month. Reach went 180, then 108, then 56, then 49. Four posts, one substantive reply.
I read that decline as "wrong platform" for weeks. It was more likely "wrong audience, correctly ignoring me."
What makes this worse: we sell a tool whose entire pitch is checking an idea against real data before you build it. I had the data. It was in our own dashboard. I just didn't point it at us until the feature was already done.
Where I'm stuck now is what to do with it. Two options I can see:
Ship it as a separate thing and hope designers find it, knowing they're 5% of a base I already struggle to reach.
Or reframe the output entirely for the 84% not "here are the layers" but "here's what a shipped competitor decided, and here's what your plan doesn't cover yet." Same engine, different artifact.
Option two is obviously more convenient for me, which is exactly why I don't trust my read on it.
So: has anyone here built something for a user segment that turned out to be a rounding error? Did you find a way to reframe it for the majority, or did you just eat it and move on? I'd rather hear "we ate it" than nothing.
the existing reply calling option two "cooking the data" is a little too harsh I think, there's a real, honest version of option two that isn't reframing to justify a sunk cost. the test would be: does "here's what a shipped competitor decided and what your plan doesn't cover" solve a problem the 84% actually described somewhere in your own onboarding or support messages, or are you inventing a use case for them because the engine already exists. if it's the former, that's legitimate repositioning. if it's the latter, that's the sunk-cost trap wearing a customer-insight costume
the four-post reach decline (180 → 108 → 56 → 49) is honestly the more damning data point to me than the 5% designer stat, that's not "wrong audience discovered a mismatch once," that's a signal repeating and getting worse each time, and it took four tries to admit it was the message, not the platform. worth remembering that pattern the next time engagement quietly declines on something, the decline itself is data, not noise to push through
to actually answer what you asked: I don't have a "we ate it" story yet, still too early, but this post is making me want to go check whether I've built anything for an assumed 95% that might actually be my own 5%
Good point. I will go check first. If someone already told us about this problem in onboarding or support, then option two is real. If nobody did, then I am just trying to save the work I already did.
And yes, the reach numbers hurt more. I never checked the 5% even once. But I saw 180, 108, 56, 49 with my own eyes and still blamed the platform every time.
Go check your own 95%. That is more useful than my story.
fair, taking that as an actual assignment rather than a rhetorical aside. will go check whether I've built anything on an assumed majority that's really my own quiet minority, and report back either way, same as you did here
good luck with the onboarding/support check, hope it comes back real rather than invented
Will do. I'll report back once I've gone through the messages, whichever way it lands.
And good luck with your check too. Curious what you find.
Honest update: I don't actually have usage data to check yet — StareBrain's pre-launch, so there's no "84% vs 5%" split sitting in real logs the way yours was. What I can check is the design-decision version of the same trap: which features in the current build are there because I decided they mattered, versus something an actual person told me they needed. Going through that now, and the early read isn't good — most of what's built so far is my own assumption about what a "confirm before it runs" flow needs, not something pulled from a support ticket or user complaint, because there aren't any yet.
So the honest report is: can't tell you whether I've built for a real 95% or an imagined one, because I don't have the users to know which one I'm looking at yet. What I can commit to is checking this for real once there's actual usage — same test you ran, applied the moment there's data to run it on instead of intentions.
The irony here cuts deep: you sell a tool to validate ideas against data, but you validated the feature using assumptions instead. You had 3,080 data points showing exactly what your users actually do, not what you assumed they need. Your dashboard had the answer the whole time.
This is the measurement trap - when you already believe something is useful (because it solves YOUR workflow), the data feels optional rather than mandatory. You skip the validation step because you're confident. Then you ship, and the distribution numbers spend a month silently telling you no one cares.
The 84% aren't ignoring you. They're correctly identifying that the feature solves a problem they don't have. That's the measurement system working perfectly - it just needs to be checked before the three weeks of building, not after.
Option two (reframing for the 84%) reads like you'd be cooking the data to fit the build, which you'd obviously catch if you asked: "would 84% of our users actually use this if we framed it differently?" The measurement is what it is. The real choice is whether designers are part of your market or not.
You are right. I sell a tool for checking, and I did not check. The data was in my own dashboard the whole time.
I kept thinking it was a marketing problem. It was not. The 84% just don't have this problem. No better wording can fix that.
Your last question is the hard one. If designers are my market, I need a way to reach them, and I have none. If they are not, this was just a side project and I should say so. I think it is the second one.
Thank you for being direct.