Usage Doesn't Add Up? Split Billing by Date and by Endpoint
"Why was this month so expensive?" is unanswerable if the invoice is a single number. The billing page now breaks usage down by date and by endpoint, and exports to a spreadsheet.
Here's when each view earns its keep.
By date: find the anomalous day
Spikes in a usage curve almost always have a specific cause: a new job shipped that day, someone re-ran a historical backfill, a loop didn't terminate. Split by date and you land on the day, then go check your deploy log for it.
"August 12th was six times normal" is a far more useful sentence than "this month is up 30%" — the first one you can investigate, the second you can only guess at.
By endpoint: find the expensive call pattern
Endpoints are priced individually, so the most-called endpoint is not necessarily the most expensive one. Splitting by endpoint typically surfaces:
- A detail endpoint called in a loop, one item at a time, where one list call would have done.
- Retry logic that keeps retrying after success.
- A scheduled job for a feature that was decommissioned months ago and still runs.
The third is the most common and the easiest win — just turn it off.
Export and reconcile
The export carries timestamp, endpoint and amount. Drop it into a pivot table, or diff it against your own server-side call log. Discrepancies almost always come from two places: failed requests aren't billed (they're in your log, not on the invoice), and retries (your log records one business operation, billing records several calls).
Explain the gap with those two first; bring us whatever is left.
Extra views for enterprise accounts
Enterprise accounts also get a period-summarised bill with a downloadable PDF receipt carrying company name and address, ready for expense claims. On postpaid terms billing runs twice a month, and the bill page shows both the next billing date and the usage accrued so far.
Three checks that pay for themselves
- Open the by-endpoint view and look at number one. Is it worth that?
- Open the by-date view and find the top three days. What happened then?
- Export last month, sum by endpoint, and compare against how you think you use the API.
Ten minutes, and step three nearly always turns up a job somebody thought was switched off.
Get started
Log in and open Billing to view and export. Prices are on the pricing page, endpoints in the docs. To control usage at the source, see managing API keys.
Related articles
- Pugongying Content Square and Keyword Analytics: Decide What to Say Before Who Says It Content Square for category-level distribution, keyword/daily to bet on rising rather than high, related for long-tail expansion, and why the live pool is a different pool.
- Build a RedNote Competitor Dashboard from Scratch: One Table, One Cron Job Questions before fields; a three-layer collect/store/present design; snapshots instead of current values; three metrics that earn their place; and what 50 calls a day costs.
- Batch Crawls That Don't Die Halfway: Error Classes, Backoff and Resumability Split failures into retryable, not-retryable and stop-the-job; back off with a cap; persist progress and flush it; then patch with a second pass. Copyable Python included.
- Reading Pugongying Audience Profiles: From Follower Counts to Real Due Diligence City concentration, age overlap, gender extremes and growth-curve shape — how to read the four Pugongying audience endpoints, plus a screening order that cuts call volume.
About the Rnote Xiaohongshu API
Skip building a RedNote scraper: one API key gets you notes, comments, users, search, products and Pugongying creator data. Pay per call, charged only for successful requests.