Your data is about to be tested. Better by someone with nothing to sell.

No. We sell no software and no implementation, and we never fix what we find. Your team owns the estate and the fixes; we provide an independent read and a fix list it can work from.

Those watch the pipelines and tables you already run, one system at a time. We match records between systems, against a specific AI decision, and sign off a verdict on it. It is a service with a written memo, not a tool you install.

Six weighted pillars: Accuracy 22%, Completeness 18%, Governance 18%, Consistency 16%, Connectivity 14%, Lineage 12%. Accuracy, Completeness and Connectivity are critical: if the weakest of them is below 80, the overall score is capped at 0.55 × that pillar + 35, rounded. In our real sample, the weighted average is 75 and Connectivity is 44, so the cap is 59: Band C, significant work needed. A pillar with too little signal is shown as limited signal or not assessed, never guessed. The methodology sets out the detail.

The memo is about the data and the decision, not people. Findings come with record counts and sample records. A gap between two systems that were never built to agree is a design fact, not a failing, and the memo says so.

The Truth Score is rule-based: the same data, scored on the same date, gives the same score, and no AI model decides it. We do use a language model (Anthropic’s Claude, through its commercial API) to read the plan or proposal you send and to help draft the memo’s wording from our findings. A person checks every line before it reaches you. If you’d rather we didn’t use it on your engagement, say so and we do that part by hand.

It arrives as exports only, under NDA and our data processing agreement. No logins and nothing installed. The full list of sub-processors comes with our data processing agreement, before any data moves. Any transfer outside the EU is covered by Standard Contractual Clauses. Raw exports are deleted after the engagement; we keep quality metrics, not your records.