The Problem

Impact is easy to see. Scope never is.

After a change, most teams can eventually list which systems were touched. Far fewer can say, with any real confidence, how much testing that actually requires. So testing defaults to everything, or to a hunch — and either way, nobody can defend the decision afterward.

The Approach

From change to test scope, in four steps.

Every change moves through the same four steps. A tester can always override the result, but only on the record.

01 Map dependencies 02 Score the risk 03 Derive the test depth 04 Review & document override, with reason
Every change is scored and resolved to a test depth automatically; a human can still override the result, but the reason is kept with the record.
01

Map dependencies

Trace which systems a change actually touches, using whatever architecture data already exists.

02

Score the risk

Combine criticality, change frequency, defect history and dependents into one explainable score.

03

Derive the test depth

Translate the score into a concrete recommendation — not a list of “maybe affected”.

04

Review & document

Any recommendation can be overridden, but only with a stated reason, kept for the audit trail.

How Risk Spreads

Risk comes from reach, not just change size.

A small change to a widely used system can matter more than a large change to an isolated one. Mapping dependencies means tracing both the systems a change touches directly, and the ones it reaches only through those.

CHANGE Auth-Gateway Kundenportal Zahlungs- abwicklung Partner- Onboarding Vertrags- verwaltung Filial- verwaltung Reporting direct dependent indirect dependent
Direct dependents weigh most heavily in the score; indirect ones still count, just less — so a well-connected system never looks “unaffected” by mistake.
Why It’s Different

Not just impact. Test depth.

Impact analysis tools can already tell you what might be affected. This approach commits to an answer: how deep this specific change needs to be tested. The score is transparent enough to explain to an auditor, and it works with whatever architecture data a team already has.

Risk score — low high Smoke test Focused regression Full regression
The same score that measures risk also decides how deep testing needs to go — one scale, one rule, no separate judgment call.
Where It Fits

Built around what you already have.

The approach doesn’t assume one source of truth. It’s designed to draw dependency and architecture data from wherever it already lives, without replacing the tools already in place.

CMDB Middleware / Integration Service Mesh & APM Manual / Federated
Talk to us about your test strategy