Fuzzy Matching
Fuzzy matching is a technique that recognizes two fingerprints as the same device even when some signal values differ, by measuring similarity rather than requiring an exact match. It tolerates the small, expected changes that occur as browsers update and configurations drift, reducing false negatives without collapsing distinct devices together.
How Fuzzy Matching works
Instead of comparing whole fingerprints for equality, fuzzy matching compares individual signals and computes a similarity or distance score across them. Signals are weighted by their stability and distinctiveness, so a change in a volatile, low-value attribute counts less than a change in a stable, high-value one.
The system sets thresholds that decide when accumulated similarity is high enough to declare a match. This allows a device whose browser version incremented, or which added a font, to still be recognized, while genuinely different devices remain separated because too many strong signals diverge.
Effective fuzzy matching depends on modeling how each signal tends to change over time. Signals that drift predictably are handled with tolerance, while abrupt changes across many independent signals are treated as evidence of a different device or an evasion attempt.
Why Fuzzy Matching matters for fraud prevention
Fuzzy matching is what makes device recognition durable in the real world, where software constantly updates and no fingerprint stays perfectly constant. Without it, routine changes would fragment one device into many identities, undermining detection of returning fraudsters. It directly improves the balance between missing real matches and wrongly merging distinct devices.
How TRACIO handles it
TRACIO relies on tolerant, similarity-based matching so that a returning visitor keeps a stable visitor ID even as their browser and environment evolve. By weighting signals according to observed stability, TRACIO minimizes both false negatives from drift and false merges of different devices. This approach is a key contributor to its identification accuracy on internal benchmarks.
Related terms
Explore further
Frequently asked questions
Identify every device with confidence
Start with a free plan of 2,500 API calls per month. No credit card required.