# Subscription Finder Paste bank or card transactions to find recurring charges, forgotten subscriptions, next charge dates, and projected yearly cost. Free and private. ## Run it - **CLI:** `gizza tool subscription-finder "2026-01-15, Netflix, 15.99"` - **Web:** https://gizza.ai/tools/subscription-finder/ - **Agents:** machine-readable descriptor (parameters JSON Schema) at https://gizza.ai/tools/subscription-finder/tool.json ## Inputs - `transactions` — Transactions (one per line: date, description, amount) _(field)_ - `min_occurrences` — Minimum occurrences to flag as recurring (2–24) _(field)_ - `currency` — Currency symbol _(field)_ - `date_format` — Date format _(field)_ ## Output - Recurring charges (text) ## Query parameters Open the tool pre-filled and auto-run via URL: - `transactions` — Transactions (one per line: date, description, amount) - `min_occurrences` — Minimum occurrences to flag as recurring (2–24) - `currency` — Currency symbol - `date_format` — Date format Example: `https://gizza.ai/tools/subscription-finder/?transactions=2026-01-15%2C%20Netflix%2C%2015.99&min_occurrences=2¤cy=%24&date_format=auto` --- ## What this tool does Paste a list of your bank or card transactions and this tool finds the **recurring charges** hiding in them — the subscriptions and memberships that quietly renew every month or year. It groups repeat charges from the same merchant, works out how often each one bills, estimates the next charge date, and projects what each is costing you per year. Everything runs **locally in your browser** — nothing is uploaded, there's no bank linking, and no sign-up. Give it one transaction per line as `date, description, amount`: ``` 2026-01-15, Netflix, 15.99 ``` Most bank and card exports (CSV) already look like this, or can be pasted and lightly tidied to match. A header row or any line that doesn't parse is simply ignored. ## Worked example Paste this: ``` 2026-01-01, Netflix, 15.99 2026-02-01, Netflix, 15.99 2026-03-01, Netflix, 15.99 2026-01-05, Spotify, 9.99 2026-02-05, Spotify, 9.99 2026-01-10, Corner Cafe, 4.50 ``` and you get: ``` Found 2 recurring charges · $25.98/mo · $311.76/yr projected 1. Netflix — $15.99 monthly ×3 · next ~2026-03-31 · $191.88/yr 2. Spotify — $9.99 monthly ×2 · next ~2026-03-08 · $119.88/yr Total: $25.98/mo · $311.76/yr ``` Netflix and Spotify each repeat, so they're flagged; the one-off `Corner Cafe` charge is not. Charges are ranked by projected **annual** cost so the expensive habits sit at the top. ## How it reads each charge | Column | What it does | | --- | --- | | **Merchant** | The first spelling seen for the group. Repeat charges are matched on a normalized merchant name, so `NETFLIX #1234` and `Netflix` group together. | | **Amount** | The typical (median) charge. Small variations group too — `$15.99` and `$16.20` from one merchant count as the same subscription. | | **Cadence** | The median gap between charges → `weekly`, `biweekly`, `monthly`, `quarterly`, `semiannual`, `annual`, or `every ~N days` for anything irregular. | | **×N** | How many matching charges were found. | | **Next ~date** | An estimate: the last charge date plus the median interval. | | **Per year** | The amount × charges per year — what it projects to annually. | ## Options - **Minimum occurrences (2–24)** — how many times a charge must repeat before it's called recurring. The default is `2`. Raise it if you have lots of history and only want well-established subscriptions. - **Currency symbol** — the symbol shown in front of amounts (`$`, `£`, `€`, …). It's cosmetic; the tool never converts between currencies. - **Date format** — `auto` handles ISO (`2026-01-15`) and slash dates, guessing US vs EU from the day numbers. Force `iso`, `us` (MM/DD/YYYY) or `eu` (DD/MM/YYYY) if your dates are ambiguous (e.g. every day is ≤ 12). ## Limits and edge cases - It reads only the columns you paste — **date, description, amount**. It can't open a PDF statement, link to your bank, or auto-sync; export or copy your transactions as text first. - Detection needs at least the **minimum-occurrences** number of matching charges, so a brand-new or annual-only subscription with a single visible charge won't be flagged until it repeats. - Amounts are read as magnitudes — a `-9.99` debit and a `(9.99)` are both a `9.99` charge. It doesn't tell debits from credits, so a repeated refund could show up. - Cadence and the next-charge date are **estimates** from the spacing of past charges, not a guarantee of when a merchant will actually bill. - It never cancels anything or contacts a merchant — it just shows you what's recurring so you can decide. ## FAQ
Is my statement data private? Yes. The whole tool runs in your browser — your transactions are never uploaded to a server, there's no bank linking, and nothing is stored. You can even run it offline once the page has loaded.
What format should each line be in? One transaction per line as `date, description, amount`, for example `2026-01-15, Netflix, 15.99`. Only the **first** field is treated as the date and the **last** as the amount, so a description containing commas still works. Amounts can include a currency symbol or grouping commas (`$1,299.00`). Header rows and any line that doesn't parse are skipped.
Why isn't one of my subscriptions showing up? A charge only counts as recurring once it appears at least the **minimum-occurrences** number of times (default `2`). If a subscription bills annually, or you only pasted one month, there may be just a single charge to see — paste more history or lower the minimum. Charges whose amounts differ a lot between months may also land in separate groups.
How is the yearly cost worked out? From the detected cadence: a monthly charge is multiplied by 12, weekly by 52, annual by 1, and so on. The amount used is the **median** of the matching charges, so an occasional price blip doesn't skew the projection. The next-charge date is the last charge plus the median gap between charges.
My dates are day/month — how do I read them correctly? Set **Date format** to `eu` for `DD/MM/YYYY` (or `us` for `MM/DD/YYYY`). `auto` guesses from the numbers — it can tell `13/01/2026` is day-first — but if every date has a day of `12` or lower it's genuinely ambiguous, so pick the format explicitly.
## Related tools - [Extract Action Items from Meeting Notes](https://gizza.ai/tools/action-item-extractor/): Extract action items, owners, and decisions from meeting notes or daily notes with deterministic rules. Markdown checklist or JSON, private in-browser. - [Base Decoder](https://gizza.ai/tools/base-decoder/): Auto-detect and decode Base16, Base32, Base45, Base58, Base64, and Base85 text, including nested layers, locally in your browser. - [cURL Command Parser](https://gizza.ai/tools/curl-command-parser/): Paste a curl command to see its method, URL, query params, headers, body, auth, cookies and flags — or rebuild clean curl. Browser-only. - [Data Format Converter](https://gizza.ai/tools/data-format-converter/): Convert data between CSV, TSV, JSON, and NDJSON (JSONL) in any direction — auto-detects the source, infers types, unions keys. Free, in-browser, no upload. - [Dotenv Manager](https://gizza.ai/tools/dotenv-manager/): Parse, validate, merge and secret-mask .env files in your browser: flag duplicate and missing keys, lint names, and export .env.example or JSON. No upload.