Apple rejected our finance app under Guideline 2.1 Information Needed: what they asked and what we sent
Apple did not find a bug in Purser. It asked five questions and a video. Here is each question, the plain answer we sent back, and the one resubmission trap that cost us an afternoon.
On 4 October Apple rejected build 27 of Purser. I went looking for the bug. There was no bug. The reason was Guideline 2.1, Information Needed, which means a reviewer opened the app, could not tell what it was or how to get in, and asked us to explain.
A crash has a fix. A request for information has a pile of small honest answers you now have to write down. This post is the pile. The reply went back to Apple on 6 November, and as I write this it is waiting for review. I do not know yet whether the answers worked, so this is a record of what we sent, not a victory lap.
What Apple asked
Apple asked for a screen recording from a physical iPhone, and answers to five groups of questions:
- What the app is for, and who it is for.
- How to get in. Setup, and whatever access a reviewer needs.
- Which outside services the app uses.
- Whether anything differs by region.
- Whether the app is in a regulated industry. Purser handles bank data, so this one got real attention.
Each is a fair question. None was written down before.
The hard one: how does a reviewer get in?
Purser is in private preview. Sign in with Apple only opens for invited members of a household, and anyone else sees a “not available” screen and no data. That is right for a money app in its first months. It is also exactly what a reviewer cannot get past.
We could not invite Apple’s reviewers into a real household, because a household holds real accounts. So we built a door for them. On the first screen, under the Sign in with Apple button, there is a link that says “Have a review code?”. Enter the code and you land in a Demo household: a made-up couple, nine made-up accounts and about a hundred days of made-up transactions, with “Demo household” written under the title of every screen.
Three rules came with it:
- The demo cannot touch the real household. The demo session is refused by every route that reads or writes real data, and a test checks that.
- The demo never calls the AI model. Penny answers with a fixed sample reply.
- The code never goes in the repository. It lives in a secrets store, goes into the review notes by hand, and gets rotated after review ends.
One practical trap: the demo’s dates are relative to the day it was loaded. A reviewer opening it three weeks later would see stale bills and empty days. So the demo is reloaded the same day we submit. On 6 November it was.
The answers, in plain words
We wrote each answer in plain words, with no marketing in it.
Purpose. Purser shows where a household’s money goes: spending, bills due before payday, accounts, budgets, goals, credit score history and a safe to spend number. Penny is an AI helper that answers questions about the household’s own numbers. During the preview it is limited to invited households, starting with ours.
Outside services. We listed every one, because an incomplete list is worse than a long one:
- Sign in with Apple, for identity.
- SimpleFIN Bridge, which lets an account holder share read-only bank and card data. The server reads from it, never the app.
- Amazon Web Services in the US East region, which hosts the server and stores the data.
- Anthropic’s Claude, which writes Penny’s answers from a rounded summary of the household’s numbers, and which reads credit report files the person chooses to import. The file is read in memory for one request and not stored.
- Apple’s Vision framework, which reads text from a screenshot the person shares, on the device.
- Apple’s FinanceKit, for the Apple Card.
Then the sentence I most wanted to be able to write: no payments, no in-app purchases, no advertising and no third-party analytics inside the app.
Region. The app works the same everywhere, in English and US dollars. The bank feed comes from SimpleFIN, which covers US banks, so a real household’s feed is US only. The demo behaves the same for every reviewer in every region.
Regulated industry. Purser is a budgeting tool, not a bank, lender, broker or payment service. It cannot move money. There is no transfer, payment or ACH anywhere in it, and it holds no funds. Penny may suggest a budgeting change, but nothing changes until the person taps confirm. We hold no financial license and do not claim one, and the app gives no investment, tax or credit advice. When Apple asks for documentation, “we have none, here is why none applies” is a complete answer.
What we built before we could reply
The reply took a month, and the delay was not paperwork. Writing the answers showed us things the app did not yet do.
The big one was account deletion. Apple requires an app that signs people in to let them delete their account inside the app. Ours did not. So we built Settings, then Delete account. It asks Apple to confirm who you are, ends Purser’s Sign in with Apple access, and deletes your data from our server. If you are the last member of a household, that includes the whole household and the bank link. In the demo household the same button ends the demo session and clears it from the device, so the next reviewer is not locked out.
The second was FinanceKit. Apple approved our request for it on 5 October, the day after the rejection. In the first draft of the reply, the FinanceKit line said it was enabled but not active in the build. By the time we resubmitted, build 31 had it working, and the line changed to say so. A reply is only as true as the build you attach to it, so we rewrote it last.
The video
Apple wants a recording from a real iPhone, not a simulator. We followed a script of about three minutes: launch from the Home Screen, enter the review code, Today, Spending, Recurring, Upcoming, Penny, Accounts, Settings, and finally Delete account. We trimmed the end, which showed Control Center, and removed the audio. It came out at 38 MB and went in as the one allowed attachment.
The trap after the reply
Everything was ready, and resubmitting still failed. The old rejected submission answered every attempt with a 409 error saying the version was not ready to be submitted, and it kept saying so.
The fix was not to keep trying. We canceled the old submission, created a new one, added version 1.0 to it, and submitted that. It went straight to Waiting for Review. If Apple rejects your app and the old submission will not budge, make a new one.
What I would do differently
Write the reviewer’s path first. We built the app, then discovered that the person who has to judge it cannot get in. Next time the demo door and the answers get written before the first submission, because they are the app’s description in the one form a stranger can check.
And keep the answers. They are now a document in the repository that says, in plain words, what Purser reads, where it goes and what it will never do. It is the same text our privacy page and Trust page should say, and writing it for Apple made us go and check that they did.
Drafted with AI assistance.