Every platform, CRM and ad server assigns its own identifiers. Identity resolution decides which email, cookie, mobile advertising ID, IP address or postal record belongs to the same person or household, then ties them under one persistent key.
That key quietly controls targeting, frequency capping and attribution. New independent tests suggest marketers should stop taking it for granted.
The messy math behind one customer
Resolution usually runs through four stages: normalisation, matching, clustering and persistence. Normalisation trims emails, lowercases them, strips Gmail’s periods and plus-sign suffixes, then hashes values with SHA-256 so companies can compare without exchanging readable data.
Matching is either deterministic, based on an exact observed key such as a login, or probabilistic, inferred from weaker signals such as IP address, device model, location and timing. Probabilistic thresholds trace to the 1969 Fellegi-Sunter model: every field gets weighted, and pairs above a threshold become matches.
- IP-to-postal links were correct 13% of the time in a Truthset-led study across nearly one billion IP records from six providers.
- IP-to-email links were correct 16% of the time, and providers agreed on the same postal link just 6.4% of the time.
- Residential IP targeting in connected TV reached the intended target 23% of the time, according to Adstra and InterMedia Advertising research.
Errors run both ways: over-linking merges different people into one profile, while under-linking splits one person across several.
Where it shows up in your numbers
The measurement effect is direct. A customer who is not linked can be counted twice and hit with double the intended frequency; a wrongly linked one receives ads meant for someone else, with conversions credited to the wrong exposure.
A July 2026 simulation by LiveRamp and the Marketing + Media Alliance, built on 1.9 million real exposures across 147,941 users, found a campaign truly returning $1.50 per dollar measured at $0.43 when identity precision fell to 50%. If an exposure cannot be connected, it cannot be credited.
Still, match rate is not accuracy. A 70% match rate says nothing about whether those 70% are right. Do not confuse resolution with an identity graph: the graph is the stored map, while resolution is the process that builds and queries it.
What to change now
Buyers should demand visibility into method and thresholds. “Demand more visibility in how matching is done, because for most partners, there isn’t any today,” said Andy Johnson of Adstra.
Meanwhile, matching is moving into sealed environments. Google made Confidential Matching the default for Customer Match in September 2024, the IAB Tech Lab launched PAIR as an open standard the same month, and OpenRTB 2.6 added provenance fields for inserter, matcher and match method.
Privacy pressure adds urgency. The FTC has said hashing does not make data anonymous, and the UK Information Commissioner’s Office called Google’s move to allow IP address and device fingerprinting from February 2025 “irresponsible”.
The market is consolidating around identity capability: Publicis agreed to buy LiveRamp for $38.50 a share, and LiveRamp shareholders approved the deal on August 17 with 99.88% of votes cast.
Source: PPC Land
