Publicis has spent over $6 billion assembling three overlapping identity systems. Here is what each one does, where they conflict, and what the rest of the market should be thinking about.
Three acquisitions. Three different companies built for three different buyers. The table below puts them side by side as of May 2026.
| Epsilon | LiveRamp | Lotame | |
|---|---|---|---|
| Acquired by Publicis | 2019: $4.4B | 2026: $2.2B | 2025, price undisclosed |
| Core product | PeopleCloud platform covering identity, CRM, loyalty, email, retail media, clean room, and measurement | Data collaboration platform. Core products: RampID, Habu clean room, data onboarding, marketplace | DMP and data marketplace. Core products: Panorama ID, audience enrichment, global data exchange |
| Identity product | CORE ID: offline-anchored, deterministic, PII-validated. 200M+ privacy-protected US consumer IDs (Epsilon product page) | RampID: pseudonymized, open-web, cross-cloud. Interoperable across 25,000+ publisher domains | Panorama ID: probabilistic and deterministic, cookieless-native. 1.6B IDs across 109 countries |
| Primary customer | Brand advertisers, enterprise CRM, loyalty programs | Publishers, agencies, platforms, retail media networks | Publishers, international buy-side, open-web advertisers |
| Business model | Platform + managed services; deep agency integration via Publicis Sapient | SaaS + usage-based (shifting toward consumption pricing model) | Platform + data marketplace revenue; account management-heavy |
| Geographic strength | Predominantly US (97% of revenue at acquisition) | Predominantly US (94% of ~$800M annual revenue); 14 markets | Genuinely global. 109 countries. Strongest international footprint of the three |
| Services depth | High. Integrated into the Publicis agency model with a managed services heritage | Low historically. Tech-first, consultative sales model, not a services business | Moderate. Marketplace support and account management, but not full-service |
| Annual revenue | ~$2B at acquisition (2019); private post-acquisition | ~$800M (FY2026, public filings); 13% 5-year CAGR | Not disclosed; ~150 employees suggests smaller scale |
All three resolve identity. They do it differently, for different environments, with different anchor signals. That is the detail most coverage misses.
| Epsilon / CORE ID | LiveRamp / RampID | Lotame / Panorama ID | |
|---|---|---|---|
| Anchor signal | Offline PII: name and postal address required, validated against transaction data | Pseudonymized identifier derived from hashed email or authenticated signal, then tokenized | Multi-input: deterministic (customer IDs, email) plus probabilistic (behavioral, public web data) |
| Resolution type | Deterministic spine anchored to verified name and address. Epsilon markets it as “100% deterministic” but uses AI and probabilistic modeling to fill data gaps and extend reach. The anchor is deterministic. The extension layer is not. | Pseudonymized deterministic: real-person match with a privacy-safe transport layer | Primarily probabilistic in practice. Uses deterministic inputs (customer IDs, hashed email) as a foundation but extends reach via probabilistic graphing without requiring user authentication. Lotame has explicitly noted this makes it different from login-dependent deterministic solutions. |
| Cookie dependency | None. Built on offline identity, not cookie-based | None. Designed for post-cookie environments using authenticated traffic and hashed email | None. Explicitly cookieless-native, built specifically for the post-cookie open web |
| Operating environment | CRM, loyalty programs, email, CTV, on/offline integration. Strongest in closed-loop brand environments | Open web, publisher ecosystem, clean rooms, multi-cloud. Strongest in data collaboration workflows | Open web, mobile, CTV, global publisher network. Strongest in cookieless and international contexts |
| Behavioral attributes per ID | Deep transaction and intent signals from Abacus co-op and Conversant | Identity transport, not enrichment. RampID connects data but does not natively carry attributes | 200+ behavioral attributes per Panorama ID. Among the richest signal sets of the three |
| ID scale (claimed) | 200M+ US consumer IDs (Epsilon product page); 2.3B global profiles (Publicis figure, includes Epsilon + Lotame combined) | Not separately disclosed. Operates across 25,000+ publisher domains | 1.6B IDs across 109 countries |
| Unique strength | Most accurate for offline-to-online resolution; transaction-validated DETERMINISTIC | Most interoperable: works across all major clouds and a 1,000+ partner ecosystem NEUTRAL INFRA | Most globally scaled cookieless ID; richest behavioral signal per ID GLOBAL / OPEN WEB |
| Inherent limitation | US-centric; requires offline PII anchor limits open-web reach GEOGRAPHIC | Does not own the data. RampID is infrastructure, not an enrichment layer DATA DEPTH | Probabilistic components introduce uncertainty; smaller US enterprise footprint ACCURACY AT EDGES |
Each of these companies was built for a specific buyer. Epsilon served brands. LiveRamp served everyone, including Publicis competitors. That history matters now.
| Epsilon | LiveRamp | Lotame | |
|---|---|---|---|
| Primary buyer | Brand CMO; enterprise loyalty and CRM teams | Publisher data teams; agency trading desks; platform partnerships; retail media networks | Publisher monetization teams; international buy-side; open-web advertisers |
| Primary use case | Personalization at scale; CRM activation; closed-loop measurement | Data onboarding; clean room collaboration; identity connectivity across partners | Audience enrichment; cookieless targeting; global data marketplace access |
| Role in ecosystem | Buy-side data layer that powers brand targeting and measurement | Neutral infrastructure. Historically served both buy-side and sell-side equally | Sell-side heritage (DMP) transitioning to buy-side; serves publishers and advertisers |
| Post-acquisition conflict | Lower. Epsilon was already woven into the Publicis agency model and existing clients expected it | High. Publishers and competing holding companies (WPP, Omnicom, IPG) depend on LiveRamp infrastructure. Neutrality is now structurally in question HIGH RISK | Moderate. Lotame still serves 4,000 non-Publicis clients and brand independence is unclear over time MEDIUM RISK |
| Publicis commitment | Full integration into CoreAI and agency stack | LiveRamp to operate as separate company; CEO Scott Howe stays; pledged "trusted neutrality" and no data sharing beyond existing contracts | Folded into Epsilon; Andy Monfried stays as CEO; Lotame brand may persist in APAC |
Publicis has made contractual commitments to maintain LiveRamp's neutrality. No current or prospective customer will be restricted from using its services. That is the contractual answer. The structural question is different: whether publishers and competing agencies will choose to deepen their dependence on infrastructure now owned by the company also competing for their media budgets. Contracts and incentives are not the same thing.
This is where the redundancy is most obvious. All three have clean room products. They are not at the same level of maturity, and they are not solving the same problem.
| Epsilon Clean Room | LiveRamp / Habu | Lotame | |
|---|---|---|---|
| Product origin | Native: built within the PeopleCloud platform | Acquired: Habu purchased January 2024 for approximately $200M and integrated into the LiveRamp platform | Limited. Lotame's clean room functionality is less developed. Its strength is the marketplace, not collaboration infrastructure |
| IDC MarketScape 2025 | Leader, cited for “most robust third-party data sets and pre-configured integrations” LEADER | Leader, cited for extensive partner network, interoperability, and native activation LEADER | Not assessed in 2025 IDC clean room MarketScape |
| Multi-cloud support | Partial. PeopleCloud has cloud integrations but not zero-copy multi-cloud architecture | Full: AWS, Azure, GCP, Databricks, Snowflake. Zero-copy architecture. No data movement required STRONGEST | Limited. Not a cloud-native clean room product |
| Identity integration | Native CORE ID integration. Comes pre-loaded with third-party data attached to real people. Fastest time-to-value within the Epsilon ecosystem | RampID native but also supports alternative identifiers including hashed email, making it flexible for non-RampID users | Panorama ID-linked but not a full clean room infrastructure product |
| Activation destinations | Integrated activation across PeopleCloud channels; strong in email, CTV, retail media | 350+ activation destinations; native activation directly from clean room environment BROADEST | 40+ DSP/SSP connections via data marketplace; not clean room-based activation |
| Primary use cases | Audience creation, cross-channel attribution, CTV measurement, brand-to-retail matching | Cross-partner overlap analysis, media measurement, retail media attribution, second-party data deals, AI model training BROADEST USE CASES | Audience enrichment and cookieless targeting. Not primarily a clean room use case |
| Notable clients / examples | Publicis agency clients; DISH Media household matching | Hershey, Comcast, Microsoft, Uber Advertising (Uber Intelligence), Netflix, Roku, Snap | PHD EMEA, Teads, PepsiCo Mexico, Chamber of Commerce Bogota |
| Redundancy risk | Overlaps with LiveRamp on clean room; Epsilon's version is more data-rich but less technically interoperable OVERLAP | Market leader and least redundant inside the stack. This is what Publicis did not have before UNIQUE ADD | Minimal clean room overlap. Lotame adds data and ID, not clean room infrastructure DIFFERENT ROLE |
Publicis frames this as a deliberate architecture, not a collection of acquisitions. Each layer has a job. Here is how they stack.
"Sapient goes in first. Epsilon connects infrastructure to real people. LiveRamp adds the collaboration layer. Marcel sits on top, activating all of it across the client's business." The pitch is coherent. Whether clients can actually run the full stack in practice is the execution question nobody has answered yet.
Not everyone in the market carries the same exposure here. The risk depends on who you are and how deeply you rely on infrastructure Publicis now owns.
| Stakeholder | Data Access Risk | Neutrality Risk | Pricing Power Risk | Measurement Trust Risk |
|---|---|---|---|---|
| Brands using Publicis agencies | Low. Full stack access. Proprietary data advantage is the value proposition | Low. Data is theirs. Publicis contractually cannot use it beyond permitted scope | Medium. Bundled stack may reduce ability to negotiate individual components | Medium. Publicis both buys media and measures it. That conflict has existed since the Epsilon acquisition |
| Brands NOT using Publicis agencies | Medium. LiveRamp and Lotame remain open, but Publicis now controls infrastructure those brands rely on | High. If LiveRamp's neutrality erodes over time, alternative identity infrastructure is limited | Medium. No immediate pricing change pledged. Longer-term leverage is unclear | Low. Can still use independent measurement providers |
| Independent publishers | High. LiveRamp was the neutral onboarding and identity layer for the open web. That neutrality is now structurally in question | High. Publishers who monetize through non-Publicis demand may not want their data passing through Publicis-owned infrastructure | Medium. LiveRamp's 25,000 publisher relationships give Publicis significant leverage over time | High. Publicis controls identity and buys media against publisher inventory. Clear conflict |
| Competing holding companies (WPP, Omnicom, IPG, Dentsu) | High. All are LiveRamp clients. Their onboarding and identity workflows now run through a competitor's infrastructure | High. Contractual protections exist today. Strategic exposure is long-term | Medium. Switching costs for identity infrastructure are significant and migration is not simple | High. Measurement data flowing through LiveRamp is now adjacent to a competitor (contractually limited, but the proximity is real) |
| Adtech vendors / DSPs / SSPs | Medium. LiveRamp integrations are how most DSPs access RampID. Publicis now owns that connectivity | Medium. LiveRamp has pledged open interoperability. Watch for future partner program changes | High. Publicis could, over time, favor Epsilon's Conversant DSP over independent DSPs within the stack | Low. Adtech vendors are not primarily measurement consumers |
| Retail media networks | Low. LiveRamp is deeply embedded in retail media and existing RMN relationships should continue | Medium. RMNs that also have relationships with competing agencies may reconsider LiveRamp as their collaboration layer | Medium. Publicis's Citrus Ad retail media stack now sits adjacent to LiveRamp's RMN relationships | Medium. LiveRamp powers attribution for many RMNs. Publicis proximity creates perception issues even if the data is protected |
"The neutral infrastructure layer that publishers, brands, and competing agencies relied on has been absorbed into one of the three largest holding companies on earth. The neutral middle is gone at the infrastructure level." That is the cleanest summary of what changed this week.
All three run data marketplaces. The overlap in segment categories is real. The differences in data ownership, geography, and signal type are what actually matter for how Publicis manages this going forward.
PROPRIETARY FIRST-PARTY DATA + CURATED 3P
OPEN MARKETPLACE: 160+ THIRD-PARTY SELLERS
GLOBAL DATA EXCHANGE: 109 COUNTRIES
| Epsilon | LiveRamp | Lotame | |
|---|---|---|---|
| Data ownership model | Owns its core data (Abacus, transaction co-op, proprietary graph). Not just a marketplace host | Platform and connector. Does not own the data sold through the marketplace. It is the distribution infrastructure | Owns some proprietary data (Panorama ID behavioral graph); also hosts third-party seller data |
| Geographic coverage | US-only for 3P data (GDPR restricts EEA use) US ONLY | Global. Sellers span 14 markets with geo-tagged segments. Strongest in the US and Europe PARTIAL GLOBAL | 109 countries. Genuinely global and the strongest international third-party marketplace of the three GLOBAL LEADER |
| Data signal type | Transaction-led: known purchase behavior with high accuracy and lower probabilistic risk TRANSACTIONAL | Mixed. Platform aggregates behavioral, demographic, purchase, and B2B data from 160+ sellers | Behavioral-led: web, mobile, CTV signals with 200+ attributes per ID and probabilistic extension |
| Key differentiator | Pre-loaded into clean room; attached to deterministic IDs; fastest activation for Epsilon clients | Neutral distribution to 350+ destinations; AI-powered discovery; expanding into AI model training | Global scale no other marketplace matches; cookieless-native; open web publisher penetration |
| Redundancy with others | Overlaps with LiveRamp on segment categories; unique on transactional depth and deterministic attachment | Overlaps with both on segment types; unique as a neutral distribution platform and AI training data hub | Overlaps on segment categories; unique on global reach and cookieless ID enrichment LEAST REDUNDANT |