Status: the data is in its final v1.0 build. Downloads open at release. The API and MCP server come later.
Verify

Nothing is hidden. Check your work.

Every release ships with a verification package: the rules the generator followed and the answers it expects. Build your model, then compare. If your ARR bridge disagrees with the answer key, one of you is wrong, and you can find out which.

What the package holds

Published with every release, in the verification/ folder and on this site.

Assumptions

Every number the generator uses, with its ID, value, reason, source and check. Where the data differs from the plan, both are shown.

Story bible

What happens to GameTide each year, why, and where each planted event shows in the data.

Data dictionary

Every file: what one row means, its keys, its columns, how it links to other files and what it ties to.

Policies

The accounting and metric rules: revenue recognition, ARR, churn, capitalization and more.

Answer key

Expected values of the headline metrics (ARR, revenue, retention, cash and more) for every month, with their definitions.

Monthly events and month-in-review

A short write-up of what happened each month, and the table of events the generator planted.

Test report

Every automated check and its result for the release: reconciliations, referential integrity, realism bounds, plan coherence.

Data map

Diagrams of how the systems connect: shared keys, business flows and tie-outs.

How to use it

  1. Pick a metric from the answer key, for example ARR at December 2025, and read its definition.
  2. Rebuild it from the raw files in your tool of choice: Excel, SQL, Python, Power BI.
  3. Compare. The warehouse snapshots show the expected result at every step, so a difference can be traced to one join or one rule.
  4. Read the story bible for the months where the numbers move. Every planted event lists the tables where it shows.

Known differences are documented

Where the generated data lands away from a planned value, the assumptions document keeps the plan and shows what the data does beside it. An honest list of what is not yet modeled is published too, so nobody spends an evening looking for an effect that is not there.