System regression & upgrade assurance

Prove nothing broke. Every version. Every time.

When software changes version, behaviour is supposed to stay the same. Tallystone proves it did — comparing the old release against the new, separating expected changes from genuine regressions, and tracking every difference to a verified fix.

The problem

An upgrade shouldn't be a leap of faith.

Every version bump — a patch, a release, a major migration — carries the risk that something that worked yesterday behaves differently today. The differences hide in millions of rows and dozens of screens. Finding them by hand is slow; missing them is expensive. Tallystone makes equivalence something you can measure, not hope for.

01 — CAPTURE

Two versions, side by side

You provide data and screens from the current version and the target version. We take an exact structural snapshot of each — no access to your live systems required.

02 — COMPARE

Deterministic diff, AI triage

We compute an exact difference — every table, column, row and pixel. Then AI classifies each one: expected upgrade change, or likely regression. Version stamps and sequence noise never reach your reviewers.

03 — RESOLVE

Tracked to verified

Your team accepts valid changes and confirms real regressions, following each through fix and verification — with a full audit trail on every decision.

What we build

One instrument, focused offerings.

Tallystone is a verification engine. We point it at the systems where getting an upgrade wrong hurts most.

Available now Insurance / claims

CRIMS Guard

Upgrade-regression assurance for CRIMS. We specialise in proving a CRIMS instance behaves the same after an upgrade as it did before — release by release, screen by screen.

Explore CRIMS Guard →
Roadmap Your systems

Regression verification, generally

The same engine applies wherever versioned software must stay equivalent across releases — line-of-business platforms, data pipelines, regulated systems. Tell us what you upgrade and we'll scope it.

Talk to us →
CRIMS releases a year — every one a chance to drift
100%of rows and screens compared, not a sample
0regressions you should be finding in production

Going through an upgrade?

Tell us which CRIMS instance you're moving and which versions you're moving between, and we'll show you what equivalence testing looks like.

Talk to us