Draft Started November 1, 2026.
Golf Sparks went from "testers only" to "sent for production review" on Google Play over one weekend, and the App Store submission followed two hours later. Most of that weekend was not coding. It was payments, store consoles, privacy pages and a handful of settings that are easy to miss until a store refuses to move. Here is what it actually took, in the order we hit it, so the next launch (yours or mine) goes faster.
I did all of this in one long session with Claude Code, which wrote the fixes, ran the tests, read the store consoles and clicked through the ones I asked it to.
The timeline
Saturday, Oct 31: Decided to launch with payments rather than give credits away and add payments later. Fixed the server so test purchases are credited.
Saturday evening - About 30 minutes in Google Play Console and Adapty to set up the Android side. First test purchases: Android at 7:40 PM, iPhone at 7:43 PM.
Sunday, Nov 1 Played a final validation round of golf (it is tough - but I am willing to struggle) with the new build, found bugs on the course, fixed them. Wrote the account-deletion page and updated the privacy policy.
Sunday, about 7:30 PM Google Play production release ready: release build tested with real purchases, all countries selected, sent to Google for review.
Sunday, 9:24 PM App Store submission in: version 1.0 (8) and the three credit packs, after an evening of App Store Connect forms.
Now Waiting for both reviews.
Payments are the part that breaks
The belief I started with: in-app purchases can't really be tested until the app is live. That's wrong, and it was costing me time. Both stores let you test before launch, and nobody gets charged:
- App Store: every purchase in a TestFlight build is a sandbox purchase.
- Google Play: add tester accounts under Settings → License testing, install the app from an internal or closed testing track (not sideloaded), and they pay with a test card.
I use Adapty ↗ between the stores and my server: the app asks Adapty for the credit packs, the store takes the payment, and Adapty calls my server (a webhook), which adds the credits. Getting that working on Android turned up four separate problems, each one invisible until the one before it was fixed:
1. The products didn't exist on Google Play. They existed on the App Store, so the iPhone showed prices and Android said NO_PRODUCT_IDS_FOUND. Having in-app purchases on one store does nothing for the other. Use the same product IDs on both (credit_pack_20, credit_pack_60, credit_pack_150). 2. Adapty didn't know the Google Play IDs. Each Adapty product needs both store IDs filled in, and its type set to Consumable. If it isn't Consumable, the Android SDK acknowledges the purchase instead of consuming it, and the golfer can buy that pack exactly once. 3. Adapty couldn't talk to Google Play. It needs a Google Cloud service account invited in Play Console (with permission to view financial data and manage orders), its key uploaded to Adapty, and Play's real-time notifications pointed at Adapty, or refunds never arrive. 4. My server was ignoring test purchases. Rejecting sandbox purchases looks like a safe default, but it means testers (and Apple's reviewer, who also buys in sandbox) pay and get nothing. Now sandbox purchases are credited, with a per-account cap so a test build can't become a free-credits machine.
The test that proves it works is not "buy a pack". It's buy the same pack twice on the same account. The second purchase only goes through if the first one was consumed. I briefly thought we'd passed that on Saturday, until the purchase log showed the two purchases were one Android and one iPhone. Check the store and the account on each purchase, not just the time.
One more: the production Android build uses R8, which shrinks the code and can strip classes the billing library needs. So the final check was on the actual release build, installed from Google Play: two purchases of the same pack on Sunday evening, both credited. That was the last thing between us and the production button.
The App Store side
The payments part of the App Store work was already done; it just needed checking:
- Agreements: the Paid Apps agreement has to be active, with banking and tax filled in. Apple still wants a couple of bank account details from me, which can wait; it doesn't block review.
- In-app purchases: all three were "Ready for Review". Adapty showed a scary orange "Action required" on them, which turned out to mean exactly that: Apple hasn't approved them yet.
- The first in-app purchases must be submitted together with an app version. You attach them on the version page. Turning purchases on from the server can't help if Apple never reviewed them.
- Review notes: say where the shop is (Settings → Credits → Credits & Get More) and whether a sign-in is needed. Golf Sparks gives every install an anonymous account, so the reviewer can buy without one.
The App Store's hidden checklist
The build was the easy part. Golf Sparks had been on TestFlight for weeks, so I assumed the App Store listing was nearly done. It wasn't: a first submission needs a stack of App Store Connect forms that TestFlight never asks for, and Apple only tells you about them a few at a time. Each submit attempt failed with a shorter list:
1. Pricing. There was none. Even a free app needs a price ($0.00) and its availability (all 175 countries or regions). 2. App information: a primary category (Sports) and a copyright line. 3. App Review contact: name, email and a phone number, which must start with "+" and the country code. The "Sign-in required" box was ticked; untick it if the reviewer doesn't need an account, or Apple expects demo credentials. 4. Content rights: whether the app shows third-party content (weather, courses other golfers publish) and that you have the rights to it. 5. Age rating questionnaire: about 25 questions, now including loot boxes and age assurance. For Golf Sparks everything is "none" except user-generated content. 6. Regulated medical device: a declaration that only appeared after everything above was fixed. 7. App Privacy: every type of data the app collects, what it's used for, whether it's linked to the user, and whether it's used for tracking. Two lessons here. Data tied to an account (name, email, user ID, purchases) is "linked to you" by definition; marking it otherwise contradicts your own privacy policy. And saved answers aren't enough: they must be published, and the submit kept failing until they were.
Two smaller surprises on the way:
- My release tool had a bug. Publisher looked the build up by
version: "8", which in Apple's API means the marketing version (1.0), not the build number. It had never submitted a second build of the same version before. One-line fix. - The build took 45 minutes instead of 10. The Mac (8 GB) had Gradle allowed 4 GB, the Kotlin compiler another 3 GB, and a day's worth of browser tabs. It was swapping 15 GB to disk. Stopping the Gradle daemon (
./gradlew --stop) and quitting a few apps brought the retry back to normal.
The things that aren't code
These took longer to find than to fix:
- The account-deletion link. Google Play's Data safety form asks for a web page where people can ask to delete their account. Mine returned a 404. Worse, my privacy policy described a deletion form at that address that had never been deployed. I published a plain page explaining how to delete in the app or by email, and corrected the policy (with the old version archived).
- Button names in the policy. The policy said "Delete Account"; the app says "Delete Account & All Data". Small, but a reviewer or a user following the policy should find exactly what it names.
- Countries. The production release sat as a draft with 0 countries selected, which is why there was nothing to send. Select them under Production → Countries / regions.
- The first production release can't be sent from the API. My release tool, Publisher, can upload and create the production release, but before an app's first approval Google only accepts a draft, so the last step is pressing Send for review in Play Console. Doing that also restarted the closed-testing review that was already running.
- Managed publishing off means the app goes live the moment Google approves it. Turn it on if you want to pick the moment.
Release hygiene
- Same version on both stores. Android said 1.0.0 (5) and iOS said 1.0 (7). Both are now 1.0 (8), which makes "which build are you on?" a one-word answer.
- Store-neutral release notes. My first draft said "credit packs on Google Play and the App Store" and "smaller videos on Android". Each store reads the notes as written for it, so they now never name a store or platform, and fixes are described by what changed for the golfer. That rule now lives in the project's agent instructions.
- A pre-publish check before every Play release: R8 on, signed with the release key, no stray advertising ID, no secrets in the repo, no new permissions without a declaration.
- Keep the process, not the state. The payments steps became a reusable skill (adapty-store-payments ↗). Each app's own progress lives in its own repo, so the skill stays a template anyone can use.
What playing a round found
The best test of a golf app is a round of golf. Sunday morning's round turned up real bugs before any reviewer did:
- Some course videos had no sound, and the swing detector relies on the sound of the strike to find impact.
- On Android, the detector wasn't decoding the audio at all, so it never heard the strike, even in clips that had sound. With the golfer far from the phone, it picked the moment they walked back to the camera instead.
- Android recorded video at about three times the iPhone's bitrate, so an 8-second trimmed swing was too large to send to the Coach.
To fix the detector without guessing, we built a small replay harness: it pulls clips off the phones, decodes them exactly the way the app does, runs the detector on a Mac and scores it against hand-checked impact times. It went from missing most clips without sound to 12 of 13 within a tenth of a second.
Then we wait
Both stores are reviewing now, and both releases go live automatically when they're approved. If you're about to do the same, here's the short list I wish I'd had on Friday:
- Create the products on both stores, with the same IDs.
- Fill in both store IDs on each Adapty product, and make it Consumable.
- Connect Adapty to Google Play: service account, key, real-time notifications.
- Credit sandbox purchases on your server, with a cap.
- Add license testers; buy the same pack twice on the release build, on each store.
- Publish a working account-deletion page, and make your privacy policy match the app.
- Select countries, use store-neutral notes, and attach the in-app purchases to the iOS version.
- Fill in every App Store Connect form before the first submit: pricing, category, copyright, review contact, content rights, age rating, medical device and privacy.
- Send for review, and go play a round while you wait.
LATE BREAKING NEWS: Google Play ↗ approved the app silently - I did not know this, until I went to check on the status. Unlike App Store Connect, that sends you alerts when your review is started (when you get rejected 😢) and when you get approved 🎉, Google Play just flips status to Published. Good to know!
LATER BREAKING NEWS: Golf Sparks is live on the App Store ↗. I did get an initial rejection the following day (aka - "There's an issue with your Golf Sparks (iOS) submission" and requiring "[a] screen recording captured on a physical device, running the latest operating system, demonstrating the app's functionality. The recording must begin with launching the app and show the typical user flow"). Once I recorded the video (again - had to play another round of golf!) the app was approved a couple of days later. Efficient process I must say! Here is the video that I shared.





