Match Rule Validation
Match Rule Validation: Go-Live Certainty
Last updated: 2026-08-10
Match Rule Validation is the offer where we build the labeled ground-truth dataset and truth key for your domain, you run your match rules against it, and we report the result. You get a defensible precision, recall, and F1 number for the gate review, plus a ranked list of whatever the scoring surfaces, each item with a recommended fix, and a report you can forward to your steering committee. Price is Pricing unavailable for the first design partners, locked for renewal.
This is for MDM practice leads at SIs and governance leads mid-implementation who need a real number before go-live, not a gut check. Match Rule Validation replaces "we think the rules are pretty good" with a scored, defensible answer you can put in front of a steering committee. You get a defensible precision, recall, and F1 number without hand-labeling record pairs yourself and without adopting another generic test-data tool.
We build a labeled, fuzzy-duplicate ground-truth dataset shaped to your domain, you run your match rules against it, and we score what they caught, then hand back failure analysis and a client-ready report. Platform access is included, so you're not starting from zero the next time you need to regenerate a truth set.
Six components, each named for what it actually does:
- Labeled Ground-Truth Dataset: domain-shaped master data with known fuzzy duplicates, built to match how your entities actually drift, typos, nicknames, address variants.
- Go-Live Precision Scorecard: precision, recall, and F1 scored against that known truth.
- Silent Failure Finder: a ranking of what the scoring turns up, the duplicate pairs your rules missed and the non-duplicates they flagged, each with a recommended fix.
- Post-Tune Re-Validation Pass: a second scoring pass after you've applied fixes, so you can see whether the tuning worked before go-live.
- Gate-Review Certainty Report: the artifact you forward to your client or steering committee.
- Match Rule Truth Engine (12 months): platform access so you can regenerate truth sets for later phases without booking another engagement.
Together these six components turn "we think our rules work" into a number you can defend at the gate review, with the evidence attached. Every component is built and scored against your domain.
We build the Labeled Ground-Truth Dataset for your domain: fuzzy duplicates shaped to how your entities actually drift, typos, nicknames, address variants. You run your current match rules against it, or let us run them alongside you. We score the result and hand back the Go-Live Precision Scorecard, the Silent Failure Finder's ranking of what the scoring surfaced, and the Gate-Review Certainty Report.
Once you've applied fixes, we run the Post-Tune Re-Validation Pass, a re-score against the same truth set, ahead of go-live. Match Rule Truth Engine access is included, so you can regenerate truth sets for later phases without booking another engagement. We generate the truth and write the report; you run your own rules against it. That's done-with-you by design, not a black box that scores your production system without you in the loop.
83%
Precision
75%
Recall
79%
F1
Say your truth set contains 200 known duplicate pairs. You run your current match rules against it and your rules flag 180 pairs as duplicates. Of those 180, 150 are pairs the truth set confirms as real duplicates (true positives) and 30 are pairs the truth set says are not duplicates (false positives). Your rules also missed 50 of the 200 confirmed duplicate pairs entirely (false negatives). Precision = 150 / (150 + 30) = 83%. Recall = 150 / (150 + 50) = 75%. F1, the harmonic mean of precision and recall, = 79%.
In plain terms: precision tells you how much of what your rules flagged was actually correct. Recall tells you how much of what should have been flagged your rules actually caught. F1 balances the two into one number. These figures are illustrative assumptions, not a result from any engagement. Substitute your own truth set and rule output and the same three formulas apply.
Match Rule Validation fits teams heading into a real gate review or go-live decision on a specific domain, with a live implementation and match rules already built. If you haven't seen your domain's duplicates yet, start with the free Match-Ready Sample. Match Rule Validation includes the full six-component build: the scoring output ranked by the Silent Failure Finder, a Post-Tune Re-Validation Pass once you've applied fixes, the Gate-Review Certainty Report, and 12 months of Match Rule Truth Engine access. The Match Precision Audit is a dataset plus a one-page scorecard call. If your rules haven't been built yet, or you're still deciding which platform to run them on, that's a different problem than this page solves.
Frequently asked questions
What counts as a match-rule failure?
A failure is a duplicate pair your rules miss, or a non-duplicate pair your rules flag as a match. Both surface when the scoring step compares your rule output against the labeled truth set, one record pair at a time. The Silent Failure Finder ranks them and pairs each with a recommended fix.
What do we need to provide for Match Rule Validation to run?
Domain knowledge and participation. You walk us through your entities and how duplicates typically show up in your data, and through the engagement you join working sessions to review results. We build the dataset, run the scoring, and write the report.
Is Match Rule Validation self-serve or do we talk to someone first?
Human-closed. Match Rule Validation starts with an inquiry, not a checkout button. You tell us about your domain, we scope the engagement together, and it moves to kickoff once we're aligned. If you want something self-serve first, the free Match-Ready Sample and the Self-Serve Ground Truth License don't require a conversation.
How is this different from the Self-Serve Ground Truth License?
The License is software only: annual access to the Match Rule Truth Engine so you can regenerate truth sets yourself, with no scoring call, no failure analysis, and no report. Match Rule Validation is the full done-with-you outcome, of which 12 months of that same platform access is one component. Choose Match Rule Validation when you need the analysis and the report, not just the software.