Marketing on a budget

I would rather build for a week than market for an afternoon. I know this about myself, and I know what it costs. A product nobody hears about is a hobby with a Firebase bill.

So I treat marketing the way I treat the rest of the stack. Small experiments with a cost and a number attached, keep what works, and let the model do the tedious parts so the whole thing fits into the hours I have. What follows is what survived a year of that.

Start where people are already looking

The cheapest customer is the one already searching. For an app that means the store search box, and the listing is where most of my marketing hours go. Keywords, title, subtitle, and the first three screenshots. This summer I rewrote Pool Pilot's screenshot captions so the first one says the app works with any brand of test strip, because that is the question a stranger has before any other. The listing is also carried in five languages, which sounds like a project and was an afternoon, because the model translates and I check.

The same logic works off the store, with less success so far. People who own pools buy test strips, so at the end of July I emailed two test strip makers and offered a free 90 day Pool Pilot trial as a QR code on their packaging. Zero replies. The experiment cost an afternoon, and most small experiments end this way, which is why they have to be small.

Talk to the people you already have

The second cheapest customer is the one who installed the app and drifted off. In August I sent a short reactivation email to twenty people who had run tests, gone quiet for a month, and never subscribed. Plain text, one link, honest about what had changed. Three clicked. Fifteen percent on a cold email is a good day, and it cost one draft and one review. The review is the rule. The model writes a fine first draft and would happily send it. Every message waits until I have read it and typed the word send, because the email carries my name.

The rating prompt taught me more. Pool Pilot used to ask "How is your experience?" before it asked for a rating, a polite gate with three buttons. About 95 people saw the gate. One rated, two sent feedback, and 92 tapped "later". The gate was the leak. It is gone, and the app now goes straight to the native rating sheet after a second good test.

Buy a little, with the kill rule written down

Paid ads are the channel everyone warns solo developers away from, and the warning is right if you spend like a company. I spend like a hobbyist with a spreadsheet. ParrotPadMD, my dictation app for physicians, runs a Google search campaign at three dollars a day. Over one month this summer it bought about a hundred clicks and four signups for under fifteen dollars. That is a cheap education in what people actually type, and every term goes back into the store listing for free.

The most useful thing a campaign told me cost nothing. A Spanish campaign for the same app, with a real budget and approved ads, has served zero impressions. The search terms barely exist. That is a market answer.

The rule I hold myself to is to write the number down before the first dollar goes out. Cost per signup above that number when the budget is gone means the campaign dies, with no "let it learn". The model drafts the ad variants and pulls the reports. The number is mine.

Fix the front door before buying traffic

Marketing done well can make a leaky product worse. When I pulled Pool Pilot's user records in August, 81 percent of accounts had never run a single test. Most people minted an identity and left before the one action that makes the app useful. Sending more people to that door would have been the most expensive thing I could do, so half of what I call marketing is the first five minutes of the product. The empty state, the first tap, and manual entry sitting right next to the scan for people with no interest in the camera.

Write it down, and keep it small

This essay is marketing, and so is the interactive explainer on this site that has nothing to do with my products. Writing is the slowest channel and the one with the longest tail, and the people it sends already know how you think. I resisted it for years because it felt like a fourth job. The model drafts, I edit, and a post that would have taken a weekend takes an evening, most of it spent deleting.

All of this works only if it leaves the building time alone. Marketing gets one fixed slot a week and a running list of small experiments, each with a cost and a success number. When the slot ends I stop, even mid experiment. I still find it a chore. The chore is now small and measured, and a model under supervision does most of it.