Accuracy methodology

This page is a structure, not a published result. Every bracketed field below has to be filled from a real measurement before this page goes live. Publishing an accuracy figure without the method behind it is the single fastest way to lose a technical buyer — and the claim on the rest of the site depends on this page existing. Until it does, no accuracy percentage appears anywhere on this website.

What we measure

Accuracy is reported as the proportion of records where the verdict LeadQC returned matched the ground truth established independently. It is not a bounce rate, a deliverability figure, or a match rate, and we do not report those three as if they were the same thing.

Sample

Sample size: [CONFIRM] records
Drawn from: [CONFIRM: which population, and how sampled]
Geographic split: [CONFIRM]
Date of the run: [CONFIRM]

Ground truth

[CONFIRM: how the true answer was established for each record — manual review, independent source, direct confirmation. State who did it and whether they could see the LeadQC verdict while doing it.]

Result

Overall accuracy: [CONFIRM]%
False removals — records we removed that were in fact valid: [CONFIRM]%
False keeps — records we kept that were not: [CONFIRM]%
Unverifiable, returned as flagged rather than judged: [CONFIRM]%

The last three matter more than the headline. A system can post a high accuracy figure by refusing to judge anything difficult, so the flagged rate is published alongside it.

How often it is re-run

Quarterly, on a fresh sample, with the previous result kept on this page rather than replaced — so you can see whether the number moves.

Ask us for the working

The record-level results behind the figure are available under NDA. If your team wants to run its own sample instead, that is what the free 5,000-record assessment is for.