See It in One Demo
Identity & Data

What 'Match Rate' Actually Means: The Numbers Vendors Won't Show You

Demo match rates run 3-5× higher than production. Most vendors quote company-level rates and let you assume person-level. Here's how to actually evaluate identity resolution match rates.

7 min read

A vendor pitches you on identity resolution. They show a demo. The dashboard lights up — 80% match rate, the deck says. You sign. Two weeks later your dashboard is showing 22%. The vendor’s not lying. They just didn’t explain what those numbers mean.

Match rate is the most-quoted, least-understood metric in the identity resolution category. Five things determine whether the number you’re shown maps to the number you’ll actually get.

1. Person-level vs. company-level — the biggest deception

Most B2B identity tools quote company match rates: the percentage of anonymous visitors that resolve to a company name and IP block. Person-level identification — actually getting the visitor’s name, email, or postal address — is fundamentally harder.

Industry benchmarks land here:

  • Company-level identification: 30–65% match rates depending on the provider, with the leading B2B tools claiming 70–80%. Source: Customers.ai state-of-industry report.
  • Person-level identification: 5–20% is typical. Some leading consumer-focused providers reach 30–40% in the US with deterministic methods. Source: Warmly match rate guide.
  • The conflation: “80% match rate” usually means 80% of companies identified, not 80% of people. Most tools quietly inflate their numbers by mixing the two.

When a vendor quotes a number, the only correct follow-up is: that’s company-level, person-level, or both? If they hedge, they’re conflating.

2. Demo rates run 3-5× higher than production rates

Vendors run demos against pristine, high-quality test traffic. Your real site has bots, ad-blocked sessions, mobile traffic from carrier-rotated IPs, and visitors who arrived via privacy-protective referrers. Every one of those degrades match rate.

Across the category, demo match rates run 3–5× higher than what brands see in production. Source: Warmly. A vendor showing 70% in the demo is realistically delivering 14–23% on your traffic.

The right way to ask about match rate isn’t “what’s your match rate?” It’s “what’s your average match rate on US consumer e-commerce traffic with at least 50,000 monthly sessions, after de-duping bots and known-customer logins?” The number you get back will be lower, and it’ll be the one that actually matters.

3. Deterministic and probabilistic produce different numbers

Two different methods sit underneath the match rate banner.

Deterministic matching — comparing exact identifiers like email addresses, phone numbers, or hashed login tokens against an identity graph — yields 100% confidence on the records that match, but matches a smaller share of total traffic. Source: GrowthLoop.

Probabilistic matching — using device fingerprints, IP location, browser signals, and behavioral patterns to infer identity — extends reach significantly but lands at 70–80% accuracy on what it does match. Source: mParticle.

A vendor claiming 70% match rate via probabilistic methods is delivering ~50% accurately-resolved records (70% reach × 80% accuracy). A vendor delivering 35% via deterministic methods is delivering 35% near-certainty records. The deterministic number is lower but more useful — every one of those records is mailable, callable, emailable.

4. Geography compresses the range

Identity graphs are denser in some markets than others. The US is the densest market for consumer identity resolution because of the depth of consented data sources and the scale of the underlying postal database.

Typical match rates by region:

  • US consumer traffic: 40–60% globally for leading providers; up to 70–80% on US-only traffic with deterministic-plus-probabilistic blends. Source: Data-Mania 2026 review.
  • Canada / UK / Australia: 15–35% — significantly lower because identity graphs are smaller.
  • EU: Often single digits, and constrained further by GDPR consent requirements that block most identity resolution use cases without explicit user opt-in.

A vendor showing a 70% number against your global traffic mix is implicitly assuming most of it is US. If 40% of your traffic is international, the blended rate drops fast.

5. The traffic source matters more than the vendor

Resolution rates vary by where the visitor came from:

  • Direct + branded search traffic: Highest resolution. The visitor is committed enough to be identifiable.
  • Paid social (Meta, TikTok): High resolution because the platforms hand off rich signals on click.
  • Organic search: Mid-range.
  • Display ad traffic: Low resolution. Visitors are often serviced by privacy-protective intermediaries.
  • Referral and direct-to-product-page: Variable.

A campaign sending paid social traffic to identity resolution will see materially different numbers than a campaign sending organic search traffic to the same pixel on the same site. Smart vendors break down match rate by source so the marketer can see where the channel pays off.

What a realistic person-level match rate looks like

Strip away the marketing-speak and put a US consumer e-commerce site behind an identity resolution pixel. Mix of paid social, paid search, organic, and direct traffic. After bot filtering and de-duping logged-in customers:

If a vendor quotes 80% on person-level US consumer traffic without caveats, ask for the contract clause. Real numbers come with caveats. (For how DirectMail.io’s match rates compare directly to leading peers, see DirectMail.io vs Customers.ai, vs Warmly, and vs Tie.)

How to evaluate any vendor’s claim

Six questions that cut through the marketing:

  1. Person-level or company-level? Get them to commit on this in writing.
  2. Deterministic or probabilistic? If both, what’s the ratio?
  3. Demo number or production number? Ask for production data on a comparable site.
  4. What’s the resolution rate by traffic source? If they don’t track this, they don’t know their own product.
  5. What’s the ceiling on US-only consumer traffic, and what drops it? Honest answers will name the geography mix and the bot filtering.
  6. Can you run a 30-day pilot on real traffic with a money-back floor? If they can’t put a number on the floor, they don’t trust their own match rate.

The vendors who answer all six cleanly are the ones running real businesses. The vendors who hedge on three or more are the ones whose match rate slides down 60% the day after the contract starts.

Why this matters more than it sounds

Match rate is the variable that decides whether the program economics work. A 50% person-level rate at $0.40 per resolved identity on a 200,000-monthly-visit site produces 100,000 mailable identities at a data cost of $40,000/month. That math runs profitably for almost any consumer brand with a moderate AOV.

A 20% person-level rate at the same price produces 40,000 mailable identities at $16,000/month — still profitable on the right AOV, but materially different program math. A 5% rate produces 10,000 identities at $4,000 — and most of those won’t be high-intent enough to mail.

The match rate isn’t a vanity metric. It’s the input variable for every downstream calculation. Get it wrong by a factor of 3–5× — which is what happens if you accept the demo number at face value — and the whole acquisition program is mispriced.

DirectMail.io’s identity resolution solution reports per-source match rates in the dashboard from day one. The numbers are the production numbers. There’s no separate demo math.

Sources: