Identity Resolution

The process of combining multiple customer identifiers — from different devices, sessions, channels, and data systems — into a single, unified customer profile. In ecommerce, identity resolution enables a retailer to connect an anonymous website visitor, a mobile app user, an email clicker, an in-store loyalty card scan, and an order history into one coherent view of a person, making Personalisation, Retail Media targeting, and cross-channel suppression possible.

---

How it works

Matching methods

Four core matching approaches, as described by CDP.com/Treasure AI (updated 2026-07-08):

• Deterministic matching — exact match on a known shared identifier such as email address, phone number, or loyalty card ID. Highest precision; limited to interactions where the customer has authenticated. The most reliable anchor for ecommerce: post-login and post-purchase events carry deterministic signals.
• Probabilistic matching — statistical inference from device fingerprints, IP address, browser characteristics, and behavioural patterns. Confidence thresholds typically set between 70–95% (as-of 2026). Extends reach to anonymous sessions but introduces false-positive risk (incorrectly merging two different people).
• Transitive matching — graph chaining: if Profile A = Profile B, and Profile B = Profile C, then A = C. Extends the identity graph but amplifies over-merge risk with each additional hop. (CDP.com/Treasure AI 2026-07-08)
• Adaptive matching — dynamically selects the method based on use case. More precise methods prioritised for high-stakes contexts (financial, medical), probabilistic tolerated for advertising reach. (CDP.com/Treasure AI 2026-07-08)

Architecture patterns

CDP-native identity: identity resolution and profile storage live inside the vendor's platform. Sub-second event stitching; PII handled within the vendor's infrastructure. Vendors: LiveRamp, Segment (Twilio), mParticle, Tealium, Adobe Real-Time CDP, Amperity, Salesforce Data Cloud.

Warehouse-native identity: customer data stays in the organisation's own cloud data warehouse (Snowflake, BigQuery, Databricks). Identity resolution applied against the warehouse rather than a vendor silo. Historically batch; streaming/CDC narrowing the latency gap (as-of 2026). Segment notes warehouse-native identity as an ongoing pain point rather than a solved problem (Segment, Aug 2025).

Embedded identity: LiveRamp offers identity resolution as a native app inside Snowflake, Amazon Data Exchange (AWS Entity Resolution partner), and Google BigQuery stored procedures — raw data never leaves the customer's cloud account. Achieved 200× performance improvement on Snowflake native app over multiple iterations (LiveRamp, Nov 2024).

The identity graph

The resolved output is an identity graph: a network of identifiers linked to a unified profile (or "Golden Record"). Key operational requirements (CDP.com/Treasure AI 2026-07-08):

• Merge lineage — auditable record of why two profiles were merged
• Confidence scoring — each probabilistic link carries a score
• Self-healing — automated detection and correction of over-merges and under-merges
• Consent-aware merging — GDPR/CCPA consent status propagates through the graph; a revoked consent must cascade to all linked identifiers
• Cross-border data residency controls — for multinational ecommerce; EU-based profiles must not be resolved in the US without legal basis

---

Signal environment (as-of 2026)

The identity resolution challenge is driven by fragmentation of tracking signals:

• 34.9% of US browsers block third-party cookies by default (Safari + Firefox), even after Google reversed its deprecation plan in April 2025 and scrapped Privacy Sandbox entirely in October 2025 (eMarketer, Jan 2026)
• 54% of all mobile impressions carry no user identifier of any kind, including alternative IDs (Comscore 2025 State of Programmatic, cited in Epsilon, Apr 2026)
• ~96% of iOS users declined app tracking under Apple's ATT framework (Epsilon, Apr 2026)
• 21 connected devices per US household on average (as-of 2026), creating fragmented journeys across mobile apps, websites, email, social, CTV, and in-store (LiveRamp, Mar 2026)
• US retail media growing 19.4% to $58.79 billion in 2025 — identity resolution is the infrastructure that powers audience matching for retail media (eMarketer forecast, Sep 2025 cited in Jan 2026)

Regulatory context

• 20 US states now enforcing comprehensive privacy laws; 8 new state laws took effect in 2025 (Delaware, Iowa, Nebraska, New Hampshire, Tennessee, Minnesota, Maryland, New Jersey); Indiana, Kentucky, Rhode Island followed January 1, 2026 (eMarketer, Jan 2026)
• Federal APRA (American Privacy Rights Act) stalled in June 2024; no federal standard as-of Jan 2026 (eMarketer, Jan 2026)
• Digital privacy laws in 20 US states contain provisions directly implicating AI profiling and behavioural advertising, even without explicit AI naming in statutory text (eMarketer, Jan 2026)
• GDPR in the EU raises the bar on data collection and use; identity resolution has become "the infrastructure on which compliant, effective digital retail marketing is built" (Epsilon, Apr 2026). Consent-aware identity graphs are required — revoked consent must propagate through all linked identifiers

---

Retail use cases

Loyalty programme enrichment

For a grocery retailer, resolving loyalty card scans + mobile browsing + sponsored listing clicks + desktop checkout requires a continuously updated identity graph. Stale graphs create false positives — wasted media spend on already-purchased audiences (Epsilon, Apr 2026).

UK grocery retailer case study: before name+address-based identity resolution, reached approximately 40% of active loyalty members through digital channels. After resolving in-store transaction history against verified name + address data, digital reach extended to over 75% of the same customer base. Suppression accuracy also improved — customers who purchased a promoted product stopped receiving that promotion within 24 hours rather than days (as-of Apr 2026; Epsilon, Apr 2026).

Retail media audience matching

Identity resolution is the infrastructure enabling retail media networks to match advertiser audiences to shopper data. Retail media fragmentation is a critical challenge: each retail media network operates its own identity namespace. In 2025, acceleration toward universal identity standards via interoperability frameworks was observed — not forced consolidation (BDEX, Dec 2025).

Indeed case study: used LiveRamp AWS Embedded Identity Resolution to unify fragmented first-party data without moving sensitive records. Results: 54% increase in retargeting audience size and 20% boost in ad response rates (as-of Oct 2025; LiveRamp, Oct 2025).

Whirlpool case study: identity resolution eliminated 200,000 duplicated audiences (reducing wasted ad spend) and achieved 5× greater addressable audience reach across 13 global brands (LiveRamp, cited 2026).

Tailored Brands case study: after building identity infrastructure on LiveRamp RampID, Men's Wearhouse saw 26% higher ROAS on web; Jos. A. Bank saw 42% increase in incremental revenue; customer match rates on Google nearly doubled (LiveRamp, Mar 2026).

Cross-device personalisation

Core use case: connecting an anonymous browsing session to a known customer profile after login or purchase. mParticle's IDSync offers five configurable identity strategies suited to different privacy/business requirements (mParticle, undated © 2026):

• Profile Conversion — anonymous → known on authentication event
• Default IDSync — standard merge on login
• Profile Link — tracks the anonymous-to-known transition event; useful for attribution
• Profile Isolation — optimises for regulations prohibiting anonymous data attribution to known users
• Best Match — probabilistic fallback when deterministic match is unavailable

Data Subject Request (DSR) compliance

Handling DSRs without automated identity systems costs organisations an average of $1,524 per request (Gartner, cited by mParticle, 2026). Automated identity resolution enables compliant deletion and portability requests to cascade across all linked identifiers.

SBS Australia case study

93% improvement in advertising cost efficiency by using Adobe Real-Time CDP to centralise first-party data (as-of Aug 2025; Adobe, Aug 2025).

---

Match rates and quality

Match rate improvement approaches:

• Name + address as primary anchor — more durable than email alone (multiple emails per person, shared family accounts, addresses that change). Epsilon's COREid and LiveRamp both use name + postal address as their core matching foundation (Epsilon, Apr 2026; LiveRamp, Mar 2026)
• Third-party enrichment — mParticle's Match Boost (Rokt, 2025) appends Experian-verified identifiers to existing customer records, raising advertising platform match rates by more than 25% (as-of 2025; mParticle, 2025). This signals that first-party identity alone is increasingly insufficient for paid media match rates

Performance benchmarks (as-of 2026)

System	Resolution latency	Source
Tealium real-time CDP	under 100ms for profile return	Tealium, Feb 2026
Bloomreach Loomi	5ms–2s ingestion to activation	Bloomreach, Jun 2026
LiveRamp Snowflake native app	200× performance improvement over multiple iterations	LiveRamp, Nov 2024

---

Vendor landscape (as-of 2026)

Market CDPs (identity + activation in one platform):

• Bloomreach (Loomi), Klaviyo — fast identity resolution and activation for DTC ecommerce; Klaviyo noted for same-day setup from Shopify connect (Bloomreach, Jun 2026)

Infrastructure CDPs (identity-as-data-layer, feeds downstream tools):

• Segment (Twilio), Tealium, mParticle (acquired by Rokt, Jan 2025), Adobe Real-Time CDP (Bloomreach, Jun 2026)

Specialist identity / network players:

• Amperity — recognised as "best for complex identity resolution" across fragmented, offline, and legacy data at scale (Bloomreach, Jun 2026)
• LiveRamp — acquisition by Publicis announced 2026; 600+ brand network; RampID as cross-platform spine; embedded resolution inside Snowflake/AWS/GCP (LiveRamp, 2026)

Universal ID solutions (cross-publisher):

• The Trade Desk UID2.0 (encrypted email), ID5 (hashed email + page URL + IP), LiveRamp RampID, Lotame Panorama ID (web + mobile + CTV) (eMarketer, Jan 2026)

CDP market size: projected to reach $37 billion by 2030 (analyst estimate via Bloomreach, Jun 2026)

---

Emerging: identity in the agentic era

Tealium (May 2026) argues identity resolution must now extend beyond web/app surfaces to resolve one person across: voice assistants, chat interfaces, Agentic Commerce purchasing agents, and human agents augmented by AI models — each breaks an assumption the CDP-era identity graph was built on. "The consequences of getting it wrong compound across an agent's chained actions in ways they never did across a marketing funnel" (Tealium, May 2026).

CDPs are described as the platform on which agentic architectures stand: "The CDP era is not over. Its infrastructure is the platform that everything after it stands on." (Tealium, May 2026)

CDP.com (Jul 2026) notes "agentic identity resolution" as emerging: AI agents beginning to auto-discover identifier relationships, recommend matching thresholds, detect over-merges, and optimise graph quality — reducing specialist data engineering overhead. Still early-stage in production as-of 2026.

---

Adoption and ROI (as-of 2026)

• 65% of marketers are working to unify their data; 75% say fragmented data makes customer engagement more difficult; 72% say it leads to conflicting messaging (Adobe 2025 AI and Digital Trends Report, cited in Adobe Aug 2025)
• 89% of CDP users report high satisfaction in meeting business objectives vs 60% of non-CDP users (Tealium citing vendor data, Feb 2026)
• 45% of CDP adopters achieve ROI within 3–6 months; 88% achieve ROI within 18 months (as-of 2026; Tealium, Feb 2026)
• 58% of retailers have fragmented or incomplete customer data; only 23% use AI in production for identity resolution; CDP ownership correlates with 2× higher AI adoption frequency (Amperity 2025 State of AI in Retail, Jul 2025)

---

Key terms

Term	Meaning
Identity graph	A network of linked identifiers (email, device ID, cookie, loyalty ID) resolved to a unified customer profile
Deterministic matching	Exact match on a shared identifier; highest precision
Probabilistic matching	Statistical inference from device/behaviour signals; extends reach but introduces false-positive risk
Transitive matching	Graph chaining: A=B, B=C → A=C; extends graph with over-merge risk
Golden Record	The single authoritative merged customer profile output by identity resolution
Match rate	Percentage of identifiers successfully linked to a known profile; varies enormously by methodology
Over-merge	Incorrectly combining two different people into one profile
Under-merge	Failure to link two identifiers belonging to the same person
DSR (Data Subject Request)	Consumer right under GDPR/CCPA to access, correct, or delete personal data — automated identity resolution required for efficient compliance
RampID	LiveRamp's cross-platform pseudonymous identifier, used as a spine for retail media audience matching
UID2.0	The Trade Desk's open-source alternative ID using encrypted email
Data clean room	Privacy-preserving environment for matching and analysing data across organisations without sharing raw PII — complement to identity resolution for retail media partnerships
Universal ID	A cross-publisher identifier standard (UID2.0, ID5, RampID) designed to replace third-party cookies for programmatic advertising