Where do your expenses end up when you log them? Two trackers with identical interfaces can store data on your phone, on the developer’s servers, or in a git repository. That shapes more than privacy: it decides what happens to your history three years out — and when the service shuts down.
Three storage models: local, cloud, both
- Local. The database lives in a file on your device; the network matters about as much as it does to a calculator. GnuCash, released in 1998, keeps its books entirely in local files; hledger stores transactions in a text file you can put under git. The developer can’t sell your data, lose it, or hand it over on request.
- Cloud. Your device is just an interface, and every entry goes to the company’s servers. CountAbout states openly that data is stored in the developer’s cloud. Upside: access from any device. Downside: dependence on someone else’s infrastructure and decisions.
- Both. The database is local by default, sync is opt-in. Actual Budget keeps the budget on your device and encrypts sync end-to-end; Monefy works offline and syncs through your own Google Drive or Dropbox, not the developer’s servers.
A two-minute check
Turn on airplane mode: if the app keeps working, some data is definitely local. Then look for export: if you can’t dump a CSV or a database file, the history belongs to the service, not you.
The risks of the cloud
- Leaks. Your spending history is a detailed map of your life: where you get treatment, who you send money to, what you do on Fridays. It interests more people than just attackers, and cloud security is entirely its owner’s problem.
- Monetization. A free service has to earn money somehow, and the model can change: ads, “analytics” partnerships, sharing statistics. The terms may already allow it — reread the third-party data clause.
- Shutdown. The cautionary tale is Mint: the largest personal finance service in the US, owned by Intuit, shut down in 2024, and millions of users had to migrate. Some managed to export their history; others didn’t.
- New owners or new terms. The startup gets acquired, pricing gets reworked, features get cut for a subscription. Nobody to appeal to: moving without an export means starting over.
The risks of local storage: no backup, no data
The flip side of control is responsibility. A local file can’t be “restored with a call to support”: if your phone drowns, the history goes with it. The minimum worth doing:
- automatic export: dump a CSV or database file to your own cloud drive weekly;
- for text formats (hledger, Beancount), a git repository: change history plus a free copy on GitHub or GitLab;
- the 3-2-1 rule: three copies, two types of media, one outside your home;
- disk encryption (FileVault, BitLocker): a file with your complete financial history on a stolen laptop is worse than any cloud breach.
Five minutes a month against three lost years — the trade is obvious.
End-to-end encryption: when the developer really can’t read your data
Ordinary encryption protects the channel and the disks, but the service holds a key: in a breach, through an error, or under a court order, the contents can be read. With end-to-end encryption (E2EE), the key exists only on your side: the server holds ciphertext useless without your passphrase. Honest E2EE looks like this:
- the passphrase is required on sign-in from a new device and can’t be recovered “via SMS” or email — otherwise there’s a back door;
- losing the passphrase means losing the data: inconvenient, but that’s what no escape hatch looks like;
- the mechanics are documented in a whitepaper or verifiable in the source code.
Among verifiable examples: Actual Budget encrypts budget sync end-to-end, and onbalance stores data on the device and syncs it with end-to-end encryption — per the developer, no one can read it, including their team. A service promising both “complete privacy” and “recovery via SMS” is making at most one true statement.
Self-hosted and open source: total control, at the price of maintenance
The logical endpoint of local storage is your own server. Firefly III (AGPL-3.0, in development since 2014) installs via Docker on a home NAS or VPS and keeps everything in your own database; LedgerSMB covers small-business accounting. Actual Budget can go on your own sync server, or managed hosting at roughly $1.5 a month.
What you get: independence from any provider, direct access to the database, and — thanks to the open-source license — survivability: if the company folds, the code goes into a fork. What you pay: security updates, backups, HTTPS, external access. A server untouched for two years isn’t “full control” — it’s a hole holding your finances.
Hybrid setups: a local-first app with optional sync
The practical compromise is the local-first pattern: the database on the device, full offline operation, sync as an option.
- Buckets stores budgets locally and syncs through a self-hostable relay server; one-time $64 license, no subscription.
- Actual Budget: your data, your server, bank import optional.
- onbalance: offline-first cash-flow planning with no bank connections, and end-to-end encrypted sync.
The short version: decide what matters more — zero maintenance (a cloud with honest E2EE and export) or full control (local-first or self-hosted, with backup discipline). Exactly one option is bad: a cloud with no export, where years of history depend on someone else’s business plan.