Skip to content
Outrich Data

Data type

License your analytics and BI work for AI training

The analytics layer of a company, meaning its SQL queries, metric definitions, dashboards and written analyses, can often be licensed for AI training. Outrich Data, a direct buyer, licenses the logic and the question-to-query pairs with underlying figures masked. Client data, vendor-licensed datasets and unreleased financial results are kept out, and you approve each package in writing.

What counts as analytics and bi data

Analytics and BI data is the work your analysts built on top of the raw tables. It includes saved SQL queries, data models and transformation code, metric definitions, dashboard specifications, scheduled reports and the notebooks behind one-off studies. It also includes the request that started each piece of work and the written conclusion that ended it.

This layer is mostly logic and explanation. A query that defines active customer, with a comment explaining why trial accounts are excluded, says a great deal about the business without containing a single customer row. Dashboards you build for clients as an agency or consultancy, and reporting on data a client owns, are excluded.

Why AI labs want it

Turning a business question into a correct query is a core task for data assistants, and public examples are toy-sized. Real warehouses have hundreds of tables, inconsistent naming, slowly changing dimensions and competing definitions of revenue. AI labs want question and query pairs written against schemas of that size, with the result an analyst accepted.

The reasoning around the query matters too. An analyst's write-up states what was asked, which data was trusted, what caveat applies and what the number means. Chart specifications show which view a professional chose for which question. Labs use these to train and evaluate models on analysis, not only on syntax.

Removed before licensing

What comes out first.

  • Literal values in queries, such as customer names, emails and IDs used in filters
  • Result sets, extracts and cached data with row-level records
  • Actual revenue, margin and forecast figures in reports, scaled or removed
  • Analyst, requester and dashboard owner names
  • Credentials, connection strings and warehouse hostnames in code and config
  • Tables and fields sourced from licensed third-party datasets
  • Board packs and investor reporting, excluded in full

Data title search

What gets checked.

  • Underlying data ownership. Logic written over a client's or customer's data may reveal that data's structure or content. Check DPAs and NDAs for the sources each model touches.
  • Third-party data licenses. Market, credit, panel and benchmark data from providers such as Nielsen or Bloomberg is typically licensed with redistribution limits, which can extend to derived fields.
  • Securities law. For public companies, internal reporting can contain material non-public information. Timing and masking should be reviewed with counsel.
  • BI vendor terms. Workbook and model files are your work product, but check the subscription terms for limits on sharing vendor-specific formats.
  • Authorship. Models and dashboards built by a consultancy or contractor need an assignment or license that allows this use.

Who signsThe head of data or analytics sponsors it and the CFO usually weighs in because of financial reporting. The general counsel approves the license and the data platform team runs the export.

Value

What moves the price of analytics and bi data.

FactorEffect on value
Question and query pairsQueries stored with the request they answered are generally worth more than query text alone.
Schema scaleLogic written against a large, real warehouse is scarcer than examples on a handful of tidy tables.
Tests and descriptionsColumn descriptions, metric definitions and tests let a lab verify that a model's answer is right.
Written analysisNarrative conclusions and caveats show interpretation, which raw SQL does not.
Verified resultsQueries known to have produced an accepted report give a trustworthy ground truth.

Questions

Licensing analytics and bi data: what companies ask.

Is this the data in our warehouse or the queries on top of it?
The queries, models, definitions and write-ups. Row-level data in the warehouse is assessed under its own type, such as CRM, ERP or product usage data, with its own rights checks. The analytics layer can be licensed without any rows.
A lab needs to run the queries to check them. How does that work without our data?
Options include supplying the schema with synthetic rows, a small masked sample, or only the expected shape of the result. Which one applies is agreed per package and approved by you.
Could our metric definitions reveal our strategy?
They can reveal what you measure. Many definitions are common practice in an industry, but some encode pricing or targeting logic you would rather keep. Those models can be excluded by folder, tag or owner.
Our query history is huge and mostly junk. Is that a problem?
No. Failed, duplicate and exploratory queries are filtered, and scheduled or dashboard-backed queries are prioritized. Some failed attempts followed by a corrected version are kept on purpose, since they show how errors get fixed.

Find out if your data qualifies.

Eleven questions, about five minutes, no call required. You get an answer on screen and a written follow-up.

esc