Upload a usage export and we run five checks over your actual rows: caching, batching, right-sizing, retries, off-hours. You see the real total and one finding in full before paying anything. The full report is £99.
The upload needs no account and is never added to one. The findings live against a random token for an hour and are deleted within two, whether you buy or not.
Upload a usage export and we compute the findings from your actual rows. No account, no card.
Your file is analysed and discarded. We store only the computed findings, against a random token, for 60 minutes. Nothing is written to an account. How this is handled.
Connect a provider with a read-only key and we read the usage over the API instead: no size limit, and it keeps working after the first run.
Connect a provider insteadA CSV export from any provider console, or a read-only admin key for OpenAI, Anthropic, OpenRouter and 12 more. No account for the upload.
Your real total, the number of findings, and one of them opened in full: evidence table, formula and fix. Computed from your rows, not a sample.
Every finding with its evidence and formula, the PDF, and the data becomes your workspace's first import so the next run is one click.
This is the demo workspace, not your data: a worked example you can open yourself at any time. Your own figures come from your upload.
Sold for business use. Your report is generated and delivered immediately after payment. Because the full deliverable is provided on delivery, purchases are non-refundable once the report is delivered. If we fail to deliver your report, we refund in full.
Headline sums HIGH and MED findings only. LOW sits under “possible, verify first”.
The formula, plainly: saving = unclaimed_input_cost × net_cache_discount × 0.6 adoption. The evidence table its arithmetic recomputes from. The confidence grade: HIGH means the money is there; MED states its adoption assumption; LOW is listed but never counted. And the fix, as numbered steps. The full worked example → Or see a finished report with sample data →