Want this dataset evaluated?

Your dataset is still only on your machine

Exporting created a zip file on your side and sent nothing. The click that brought you here sent nothing about the dataset either. TensorOpt sees data only if you choose to share it after scoping, under a written agreement (a data processing agreement if the queries may contain personal data).

What the call covers

• The decision the evaluation should inform.
• What the export holds, judged from manifest.json (event count, field coverage, dropped events) rather than the raw data.
• Whether a Readout is worth it, and what you can do yourself if it is not.

What an evaluation needs from you

• The export, or a redacted subset of it.
• The configurations to compare, and a staging instance to run them on.
• A few hours of a domain expert's time to label results.

What you get back

A Readout is an independent reading of what the numbers in your export mean for your decision: what to measure first, a labelling plan sized for your case, and a fixed quote for an Evaluation. It comes as a short written report, with CSV files of the top and failing queries. An Evaluation measures one decision with the labels your experts produce, and you keep the judgment set.

What you already hold

• events.jsonl: one search per line, with the query, filters and ranked results.
• manifest.json: coverage statistics and the settings in force.
• README.md: the schema and how to read it.

With default settings the plugin has the log ready after 30 days and 500 searches, keeps records for 90 days, and exports no raw user identifiers: each search carries only a salted hash that changes weekly. Query text is typed by your users and can still contain personal data. Check title and snippet coverage in the manifest before planning any labelling.