Nobody Said Take My Money — Part 6 of 7

There are a thousand posts about how to start a company and almost none about how to end one, which is strange, because most companies end. This is the post about how Mother Hen ended — deliberately, in writing, on evidence — and about the one piece of demand data that only becomes available after you pull the plug.

A decision record, not a mood

On June 18, 2026, I wrote the shutdown as a formal architecture decision record — the same document format I’d used all year for technical choices like radio architectures and sensor designs. Number in the sequence, status, context, decision, consequences. ADR-151: sunset the consumer business.

That sounds like bureaucratic theater for a company of one, and I want to defend it, because I think it’s the most transferable thing in this post. Writing the shutdown in the same format as every other engineering decision did three things.

First, it forced the grounds to be stated. The context section says why, in terms of evidence, not feeling: three independent demand tests — paid acquisition that wouldn’t convert, a warm room of exact buyers producing zero sales, a community already self-solving for free — converging on enthusiasm without purchase. The sentence at the center of it: nobody ever said take my money. Reading that section, you cannot mistake this for burnout or a bad week. And on the days afterward when it felt like a bad week and the reversal instinct stirred — I’d already reversed one shutdown twelve days earlier, remember — the document held the line the mood couldn’t.

Second, it separated the two verdicts that failure blurs together. The decision explicitly records that this is not a verdict on the engineering — the platform worked, flawlessly, to the end. The business failed; the system didn’t. Keeping those sentences apart isn’t self-soothing, it’s accuracy, and it’s what makes the lessons portable instead of just painful. If you let “nobody bought it” collapse into “it was no good,” you’ll learn the wrong lesson and be gun-shy about the wrong things.

Third, it defined what reopening would take: an explicit superseding decision. Not a wistful evening, not a promising conversation at a feed store — a written record that new evidence exists. The door isn’t welded shut; it has a lock with a stated key. That structure protects against relapse in both directions: no zombie revival on a hopeful whim, and no need to re-litigate the shutdown every time the idea twinges, because the terms of revival are already on file.

What “responsible” meant concretely

A shutdown with users, even free beta users, is a promise-keeping exercise. Mother Hen had two beta sites — one in Tennessee, one in Colorado — running production hardware, with real birds behind the alerts. They had given me a year of field data and patience. Here’s what winding down meant for them, decided in the same document:

Full support through December 31, 2026 — monitoring, alerts, bug fixes, uptime, all of it, free, exactly as it had been. Both sites told plainly, in words, that the service retires at year end — no quiet degradation, no letting the certificates lapse and calling it a sunset. And after that date, the alerts stop and the service comes down, cleanly.

The intellectual property took the opposite path from the service: preserved intact, not destroyed. The platform, the firmware, the two patent filings (19/439,421 and 63/953,780) — shelved as a working whole. A wound-down business with its assets kept whole is an option; one that was dismantled in anger is just a corpse. Several pieces of that platform are already earmarked for lives in other products, which is a kind of afterlife the bundle itself never earned.

The economics of the wind-down are almost comically small — the AWS bill for the whole platform runs one to seven dollars a month, so keeping two beta sites alive for six more months costs less than a pizza. If your infrastructure is similarly cheap, know that the temptation runs the other way: it costs so little to keep a dead thing running that you can avoid ever declaring death. The end date, in writing, is what makes it a sunset instead of a haunting.

The last demand test

Then came the part nobody warns you about, and the reason this post belongs in a series about demand rather than a series about grief.

I told people. The beta sites, the interested parties, the people who’d followed along — the service ends December 31.

And everyone said “okay.”

Politely. Kindly, even. With sympathy and good wishes. And with not one single person — not a user, not a prospect, not one of the hundreds of people who had been so enthusiastic across a year of conversations — pushing back. Nobody said “wait, don’t.” Nobody asked what it would take to keep it alive. Nobody offered, now, at the hour when it would have mattered most and cost least, to finally pay for the thing.

I’ve come to think of this as the back half of the sentence the series is named for. In life, nobody said take my money. At the death, nobody said don’t shut it down. Those are the same measurement taken at the two moments of maximum honesty — the moment before commitment and the moment of loss. Everything in between is noise, courtesy, and vibes.

There’s even a name for what I’d accidentally run: a take-away test. Marketers manufacture scarcity to force a purchase decision; I’d produced the real thing. If genuine demand had been hiding somewhere under all that politeness — suppressed by price, by timing, by trust — the removal of the product is when it would have surfaced. Loss aversion is the strongest force in a buyer’s psychology. Pulling the product away swings the hardest lever there is, and still nothing moved.

So here is the lesson, and it’s one you can use without shutting anything down: reaction to taking it away is a first-class demand test. Threaten a beloved feature with retirement and users will riot; that riot is data money can’t buy. If you can announce your product’s death and hear only warm condolences, you haven’t lost your market. You never had one. The funeral just made it official.

Killed well

The venture failed. The shutdown, I’ll claim, succeeded: on evidence, in writing, with promises kept, assets preserved, and terms of revival defined. Fourteen months in, that distinction was the only one still available to get right — and it’s worth more than it sounds, because how you end determines what you carry forward. I walked out with a working platform, two patents, two grateful beta sites, a clean conscience, and — the actual treasure — a documented, dated, evidence-backed record of exactly how a competent person builds the wrong thing.

The final post is that record, inverted: the playbook I’ll run next time, so the $500 lesson never again costs a year.

Part 7: The Playbook I’ll Run Next Time.