Nobody Said Take My Money — Part 2 of 7

Mother Hen started the way a lot of doomed products start: with a real person and a real problem.

A friend of mine keeps chickens. Like every chicken keeper, he had a door problem — the coop door has to be closed after the birds go in at dusk and opened again in the morning, and the cost of getting that wrong even once is dead birds. He had automated parts of it, but automation without monitoring just moves the worry: now instead of wondering whether you closed the door, you wonder whether the machine did. He wanted to know, from the house, at any hour, that the coop was actually secure.

That is a real problem. I watched him live with it. And I could solve it — I’ve spent decades building large-scale software and cloud systems, and the pieces (cellular connectivity, cloud alerting, small embedded devices) were all things I knew how to put together. So I built it. And then I kept building it, for a market I had never spoken to.

Here’s the detail that should have been a flashing warning light, and instead became part of the origin story I told people: I don’t keep chickens myself. Every ounce of my contact with the problem came through other people. I wasn’t scratching my own itch; I was scratching my friend’s itch and assuming an entire market itched the same way.

What the friend actually proved

My friend proved that the problem exists. One real person, with real birds and real predators, genuinely wanted to know his coop was secure without walking out to it every night.

What he did not prove — what one person can never prove — is that the problem generalizes: that enough strangers feel it at the same intensity, and would pay money to fix it, to support a business.

The gap between those two claims is where the entire venture died. And the brutal part is that I never even noticed the gap, because I never wrote the second claim down. “There is a market for coop monitoring” was never stated as a hypothesis, so it was never exposed to a test. It rode silently along with the first claim, which was visibly, undeniably true every time I visited my friend’s place.

The leap from “this problem exists” to “this problem is a market” is a claim like any other. Claims get tests. Mine never did — I skipped straight to a bill of materials.

Why a friend is the worst possible sample

It took me until the after-action review to understand this clearly: a friend isn’t just an n of 1. A friend is a doubly biased n of 1.

First, a friend is loyal. He wants you to succeed. He answers “would you use this?” with the generosity you extend to someone you care about, and he keeps encouraging you past the point where a stranger would have shrugged and walked off. Loyalty is a wonderful thing in a friend and a poison in a data point.

Second, a friend who inspired a product is, almost by definition, an outlier. He’s the person who felt the problem intensely enough to complain about it in your hearing. The population of chicken keepers is mostly people who feel the problem less than he does — as I would find out much later, standing in a festival booth while person after person nodded, agreed the fear was real, and bought nothing.

“My friend loves it” is close to the weakest evidence a company can be built on, and it feels like the strongest, because it comes from someone whose judgment you trust, attached to a problem you’ve seen with your own eyes.

The fix is not “don’t build for one person”

I want to be careful here, because the obvious lesson is wrong. Starting from one real person with one real problem is a strong origin. It’s how you avoid building abstractions nobody wants. The error was never that Mother Hen began with my friend.

The error was treating the extrapolation from him to a market as free. It isn’t free. It’s the single most expensive assumption in the whole venture, and it can be tested for almost nothing. The fix is to make the extrapolation step explicit and give it a test it can fail:

Before building for a market, find five to ten strangers with the same problem at the same intensity, and get a binding commitment from each of them. A deposit. A paid pre-order. A signed-up paid beta. Something that costs them a real act, not a kind word.

Strangers, because strangers owe you nothing — their yes is clean. The same intensity, because a mild version of the problem produces enthusiasm without purchase, and enthusiasm without purchase is precisely the trap this series is named for. And a binding commitment, because it is the only reliable truth serum I know of. People will agree to nearly anything to be nice. They will not put a card on file to be nice.

The friend is the hypothesis. Ten committed strangers are the test.

What the test would have cost

Run the numbers on the version of history where I did this.

Finding ten chicken keepers to talk to is not hard — forums, feed stores, festival crowds, my friend’s own network. Call it two weeks of conversations, maybe a month if I was slow about it. If five to ten of them had pre-ordered, I’d have built Mother Hen with actual confidence instead of borrowed conviction — same product, radically different footing.

And if I couldn’t find them — which is what the following year of evidence says would have happened — I’d have learned that the problem doesn’t generalize, for the cost of ten conversations. Instead I learned it for the cost of more than a year of engineering: roughly 24,000 lines of firmware, 34 cloud functions, two patent filings, a certified booth at a festival, and a shutdown decision record.

Ten conversations versus fourteen months. That’s the exchange rate on an untested assumption.

The honest coda

Would past-me have run this test? Probably not, and it’s worth saying why: I didn’t want the answer. Building was fun. The problem was real, my friend was happy, the engineering was going well, and a test that could kill the project felt like an interruption. That’s the psychology the next post is about — how competence itself becomes the hiding place.

But if I had run it, I would have failed it in week one. Ten strangers with binding commitments did not exist for this product at this price in this market. Every demand test I eventually ran — a paid campaign, a booth in front of the exact target audience, a community deep-read — returned that same answer. I just ran them a year too late, after the answer had gotten expensive.

The cheapest failure of my life was available for the price of ten conversations, and I paid retail instead.

Part 3: The Year of Building — What Competence Costs You.