OUR PROOF

Identity Resolution Benchmark Results

In head-to-head testing against the top 15 identity resolution providers, DarkMath's vector-based architecture demonstrated significant advantages in accuracy, match rates, and false positive reduction. These results represent real-world performance on production-scale datasets, not synthetic benchmarks.

Talk to an Expert
HEADLINE RESULTS

Measured against 15 leading providers

In head-to-head testing on production-scale data, DarkMath's vector-based approach outperformed the top identity resolution providers on accuracy, match rates, reach, and duplicate detection.
86.44%
F1 accuracy — +19.82% vs. next best
+28.68%
higher match rate
+55%
more reach (+3.8M households)
47%
fewer false duplicates
ACCURACY & MATCH RATES

Head-to-head comparison

MetricDarkMathCompetitorAdvantage
F1 score (vs. Provider 1)86.44%72.14%+19.82%
Match rate (vs. Provider 1)78.96%61.36%+28.68%
F1 score (vs. Provider 2)83.42%68.86%+21.14%
Match rate (vs. Provider 2)72.22%54.91%+31.5%
REACH VS. PRECISION

Household resolution

MeasureDarkMathCompetitor
Households resolved10,838,3896,992,837
Additional reach+3.8M (+55%)baseline
Largest household size37 (realistic)994 (false positives)
INTERPRETATION

What these numbers mean

F1 score: the balance of precision and recall
F1 is the harmonic mean of precision (how many predicted matches are correct) and recall (how many true matches are found). DarkMath's 86.44% F1 means it finds most true matches while avoiding false merges — a 19.82% improvement over competitors who trade one for the other.
Match rate: finding connections others miss
A 78.96% match rate means DarkMath resolved nearly 79% of record pairs that should link. Competitors at 61.36% miss almost 40% of valid connections — fragmenting customer views and leaving revenue on the table.
Household size: the mega-cluster problem
When a competitor builds a "household" of 994 people, something is broken — overly permissive rules merging unrelated records. DarkMath's largest household of 37 reflects real-world limits, keeping analytics free of false aggregations.
Duplicate detection: the hidden problem
DarkMath surfaced millions of duplicate records competitors missed entirely. Those duplicates drive double-spend, inflated counts, and corrupted lifetime-value math — a ~47% reduction represents major operational savings.
METHODOLOGY

How we tested it

Benchmarks were run on production-scale datasets, with ground truth established through extensive manual validation and synthetic corruption testing. Identical input data was processed through each vendor's resolution pipeline. Competitor identities are anonymized to focus on capability, not vendor criticism. Results vary by data quality and use case, we recommend a proof-of-concept evaluation on your own data.

Validate these results on your own data

See how DarkMath performs on your records with a proof-of-concept evaluation.
Contact Us