Amana is a solo-built iOS app that renders the sky outside your window — from your time, your location and the sun's position — and gets you to go outside and give a real sky a name. I shipped it on the App Store on 6 August 2026 for the RevenueCat Shipaton. This is the part of the story that is not about code: what I did to be seen, and what the numbers said afterwards.I gave the app two faces. An indie app is easier to remember when someone is standing in front of it, and I did not want that someone to be me. So I made the Amana sisters. Akane, the elder, belongs to dusk. Kahatare, the younger, belongs to the hour before dawn. They share the app's name as their surname, and they speak the way the app works: Akane asks what a sky should be called, Kahatare answers in two words. They introduce the app in a series of short videos, each in Japanese and English. Illustrated and animated with generative AI, and every upload says so.Their names come from the two ends of the day, but only one of them is a question. Kahatare is written 彼は誰 — "who is he?" — the pre-dawn half-light in which a face cannot be read (Digital Daijisen). Akane is not a question at all: it is 赤根, "red root", the madder plant whose root yields a red dye, and by extension 茜色, the color of a sunset sky (Digital Daijisen). The dusk word that does ask "who is that?" is 誰そ彼 — tasogare (Digital Daijisen). The elder is a color; the younger is a question. I had the pairing wrong in an earlier piece, and checking a dictionary made the sisters more interesting, not less.I took them to four channels under one name. X, Instagram, TikTok and YouTube, all as amana_sky, all with the same brush-lettered logo. Every video is built once in 9:16 from scripts checked into the repository, then posted to each platform in the format it prefers.On YouTube I ran both short and long. A Shorts series — the sisters' self-introductions, a 30-second tour of the app, a 10-second hello — and a full-length video for the theme song, with the lyrics and the App Store link in the description.I wrote the app a theme song. "Sora ni Namae o" (Give the Sky a Name) is a quiet Japanese ballad, generated with Suno and sung by two soft voices — the sisters. Shakuhachi, koto, a taiko heartbeat far in the back. The app now has a sound as well as a look. I kept it out of the submission demo on purpose: the app's rule is restraint, and the song belongs where people meet the app before they open it.What it actually did. A month after launch I stopped and counted. Three systems, three different questions, so I am naming the source of every number rather than blending them.WhatNumberSourceVideos published18 (17 Shorts, 1 full-length)YouTube, public countsTotal views3,590YouTube, public countsTotal likes27YouTube, public countsChannel subscribers3YouTube, public countsBest single video488 views — Kahatare's introduction, 19 AugYouTube, public countsThe theme song, full length9 viewsYouTube, public countsUnique app users40In-app analytics, 6 Aug – 9 SepUsers who opened it on a second day0In-app analytics, same windowReached the paywall11 of 40In-app analytics, same windowGross revenue$9.41, from one transactionRevenueCat, all timeApp Store ratings and reviews0App Store Connect API and the public lookup API, bothThree thousand five hundred and ninety views, three subscribers, forty people in the app, and nine dollars and forty-one cents.The most expensive thing reached the fewest people. The theme song took the longest to make — lyrics, a generated vocal, a mix, a full-length video — and it has nine views. The cheapest thing I made, a Short of the younger sister saying almost nothing, has 488. The format did not reward the effort; it rewarded the length.The first honest act was deleting myself. Before I could report any of this I had to take myself out of it. My raw analytics said 70 users and 2,013 events. After excluding simulators, TestFlight builds, Apple's review device, and the one person_id that had opened the app on sixteen consecutive days — me — it said 40 users and 520 events. Forty-three per cent of my users and seventy-four per cent of my events were my own hands.That is not a rounding error. It is the difference between "people are using this" and "I am using this." If you are launching alone, your analytics will flatter you by exactly the amount you test, and the flattery is largest in the first weeks, when you are testing most and strangers are fewest.One feature made the point sharply. The app posts a short line to its widget when it thinks the sky is worth looking at. In the raw data that fired eight times, which felt like a working feature. Every one of those eight was a simulator, a TestFlight build, or my own phone. Real users reached it zero times. It worked perfectly in every test and it has never once been seen.Recognition is not the same as being moved. I started with the assumption that a face makes a product memorable, and it did: the sisters are the most-watched thing I have made, by a wide margin over the app's own footage. But recognition did not become installs, and installs did not become second days. Forty people opened Amana. Eighteen of them went outside and took a photograph, which is the thing the app exists to cause. None of them came back the next day.So the honest finding is narrower than the assumption I started with, and more useful. The characters solved the problem I gave them — I did not have to be on camera, and the app got something a person can point at. They did not solve the problem I had not yet named, which was that a first session with no reason to return is a first session that ends. That is a product problem wearing a marketing problem's clothes, and no amount of distribution was ever going to fix it. I spent a month finding that out at the front door when the answer was inside the house.What I would do differently. Instrument the return before instrumenting the launch: I could measure how many people arrived long before I could measure whether any reason existed for them to come back. Exclude your own devices on day one, not in week five, because every week you delay is a week of numbers you will have to throw away. And put the effort where the format pays — seventeen Shorts outperformed the one thing I actually laboured over, and the data said so within a fortnight, if I had been counting.