Twelve people told me the convenience store BOGO deals app was a great idea — and then 6 weeks later I had a 0.2% conversion rate and a theory about why I'd wasted a month feeling confident.
The Part Where I Felt Really Smart
Before I wrote a single line of code, I did the thing you're supposed to do. I validated.
I talked to people. I sent a survey. I even made a little Notion doc titled "User Research" which, in hindsight, was doing a lot of work with very little inside it.
The survey had 11 responses. The informal chats — friends, former coworkers, a couple people from a Discord I'm in — added up to about a dozen more. The vibe was warm. "Oh yeah, I hate missing those BOGO deals." "I'd totally use that." "Why doesn't this exist already?"
I wrote "strong signal" in my notes. I believed it. I moved into build mode feeling like I'd done my homework.
What I Was Actually Doing
Here's the thing I didn't want to look at until the numbers forced me to.
Every single person I talked to was someone who already knew me, or someone who knew I was building stuff and thought that was interesting. Not one of them was a random stranger who had no reason to be nice to me.
The survey — I'd shared it in a group where people are explicitly supportive of indie builders. The link was framed as "I'm working on something, would love your input." That framing is basically a kindness request dressed up as a research question.
I wasn't looking for signal. I was looking for permission to keep going.
The 12 positive responses weren't market validation. They were social responses. People were reacting to me, not to the product. There's a difference, and I papered right over it.
The Filter Bubble I Built by Hand
What's weird is I didn't stumble into a filter bubble by accident. I constructed one, methodically, under the label of "being thorough."
I asked people who were already interested in apps. I asked people who already think about deals and savings. I asked people who have a social incentive to be encouraging. And then I averaged those responses and called it data.
The one time I got actual pushback — a friend who asked "but won't people just check the store's own app?" — I wrote it off as one skeptic in a sea of enthusiasm. I filed it under "objection to address in onboarding." I did not file it under "maybe he's right."
He was right, by the way.
Validation feels like research but functions like a mirror when you get to pick who's holding it.
When Reality Showed Up
Six weeks post-launch. I'd spent about $40 on a small Meta ads test — 1,200 impressions, cold audience, no warm referrals, nobody who knew me. The kind of audience that has no reason to be polite.
Conversion rate to install: 0.2%. That's roughly 2 or 3 people out of 1,200 who saw the ad and thought it was worth the tap.
That number doesn't lie the way people do. Not because people are dishonest — they're just human. They didn't want to tell me my thing was boring. The algorithm had no such concern.
I also ran a small survey on a neutral subreddit (not one for indie builders — one actually populated by deal-hunters). Eight responses, unsolicited framing, just "hey do you use anything to track store deals?" Five of the eight said they just check the store's own loyalty app or the flyer in their inbox. Two said they hadn't thought about it. One said "there's an app for that already."
That $40 and those 8 responses told me more than the entire month of friendly conversations.
Cold audiences are expensive but they charge you in honesty, not dollars.
What the Survey Actually Measured
I've been turning this over since. The survey I sent to people I know — what was it actually measuring?
Best I can figure: it measured whether people could imagine themselves in a scenario where a product like this would be useful. That's not the same as whether they'd download it, use it twice a week, tell someone about it, or pay for it. Imagining use is almost free. Actual use costs attention and habit change, which are both very expensive.
"I'd totally use that" is cheap to say. It costs nothing. It's the app store install that costs — not money, but the micro-decision of: is this actually worth the space on my phone?
I had 12 imaginary users. I needed real ones.
What I'd Do Differently (What I'm Actually Doing Now)
I'm not saying interviews are useless. But I ended up running a $6/month landing page with a working email capture before I built the next thing. No friends, no Discord link, just a Reddit post in a relevant community and some light SEO.
After 3 weeks: 34 emails collected, 0 people I knew personally. Smaller number than my original "validation" group, but the composition was completely different. These people found me. That inversion matters.
One of them emailed back with a complaint before I'd even shipped anything. I was thrilled. Complaints from strangers are worth more than compliments from friends.
I also stopped framing early conversations as validation. Now I call them "hypothesis sharpening" in my notes — deliberately lower stakes, no box to check at the end. If I come out with a sharper question rather than a green light, that's a win.
The goal of early research isn't to feel confident. It's to get surprised.
The "Love It" Trap
Looking back, every "love it" I collected was a small drug hit. It felt like progress. It was actually just comfort. And comfort is the thing that lets you spend 4 weeks building something before a $40 ad spend whispers "hey, maybe not."
The brutal part is that the 12-person filter bubble wasn't laziness. I put real effort in. I made a doc. I scheduled calls. I analyzed the responses.
Effort without adversarial signal is just expensive daydreaming.
Twelve friends saying yes is not the same as twelve strangers not saying no.
TL;DR: I ran what I called "idea validation" but it was really a permission-slip collection tour — friends, supportive communities, and leading framing. A $40 cold ad test with a 0.2% conversion rate was more honest than all of it combined.
Next episode: I tried to fix the cold-traffic problem with SEO instead of ads — here's what three months of organic actually looks like in a niche this small.
Get the $31 Stack Prompt Pack — $19
The exact prompts, cost calculator, and n8n templates from this stack.
Top comments (0)