Imagine ordering a pizza on your laptop after seeing an advertisement on your phone. For you, this is one mildly hungry evening. For an advertiser, it can look like two strangers: somebody who saw an ad and somebody else who bought dinner. The order arrives. The explanation does not. Conversant made a business out of repairing that peculiar separation.
- Conversant connected customer identity with personalized advertising and measurement.
- The former ValueClick business became Epsilon’s digital media arm, within Publicis Groupe.
- Its useful lesson: agree how to measure extra sales before spending more on media.
The company sells to brands and agencies, helping them find prospects, retain customers and coordinate messages across devices. Its raw material is the gap between what a business knows about a customer and what its advertising systems can recognize. A retailer may have years of transactions while its media tools behave as though every browser visit is a first date.
First came the click. Then came the person.
Brian Coryat’s ValueClick business began inside Web-Ignite in 1997; the operating company started in May 1998. Its early proposition was wonderfully concrete: advertisers could pay when someone clicked. It went public in March 2000. Over time, acquisitions assembled a wider collection of capabilities, from affiliate marketing to mobile advertising and ad serving.
By 2014, the original display business was struggling. Contemporary earnings coverage described falling revenue in traditional insertion-order display advertising. The response was a narrower focus and a new organizing idea. In February, ValueClick became Conversant, combining the capabilities of ValueClick Media, Commission Junction, Greystripe, Dotomi and Mediaplex beneath a personalization proposition. Five specialist businesses were supposed to become one coherent conversation.
Different screens. A connected customer record.
This explains its place in the market. A publisher sells an audience. A buying platform purchases impressions. An identity provider connects records. Conversant’s pitch brought those jobs together with creative execution and measurement. It sought to reduce the handoffs where a customer’s identity could get lost.
In its training material, the company used Road Scholar, the educational travel organization, to explain why this mattered beyond retail. Repeat customers were valuable, and a purchase could follow advertising on another device. The marketing task was to maintain a recognizable relationship over time, rather than restart the introduction each time someone moved from a phone to a computer.
A pizza order with a £113,000 question
Domino’s UK supplies a useful example. Some channels were fragmented; television had become expensive. The company wanted to bring display advertising back into its mix and demonstrate an incremental return. Conversant helped activate its first-party data, matching customers to pseudonymous identifiers and connecting media delivery with orders.
In February 2020, the campaign announcement reported £1.13 million in incremental revenue over three months from £113,000 invested, reaching 3.1 million people. The reported 10:1 return exceeded a 4:1 target. These are the company’s published campaign results, rather than a forecast for the next buyer. The spend is a concrete example of campaign economics, not a universal entry price.
Investment → reported incremental revenue over three months
Incremental revenue per £1 spent, as reported in the campaign announcement.The less photogenic detail matters. Conversant described involving data security and legal teams before execution. Someone had to agree how the records could be used and how success would be demonstrated. For a marketing director trying to justify customer-data work to a board, the experiment made an abstract infrastructure project easier to discuss.
The bundle had to be built
Conversant also had an affiliate business. CJ connected advertisers with publishers who could earn commissions for referred transactions. In the September 2014 acquisition announcement, that network included more than 60,000 publishers and 3,000 advertisers. The arrangement gave merchants a distributed sales channel and publishers a way to earn from purchases, rather than simply renting space on a page.
The broader commercial model combined advertising technology with service: campaigns needed planning, implementation, optimization and reporting. Alliance Data saw a fit with Epsilon’s customer-data and relationship-marketing operations. Its approximately $2.3 billion acquisition of Conversant closed in December 2014. Publicis then bought Epsilon in 2019, bringing Conversant along. The announced $4.4 billion cash consideration covered Epsilon, a larger business.
“We offer a bundled solution for our clients.”
Bryan Kennedy, then Epsilon CEO, 2019
A bundle still has seams. Reporting in 2019 described a slow initial integration of Conversant’s components. The company’s CORE identity capability, launched in late 2016, helped connect data assets and digital media. The lesson for anyone assembling marketing technology is less glamorous than a product launch: buying compatible capabilities and making them work together are separate projects.
There is also a revealing souvenir in public view: the company’s GitHub organization, now named Epsilon. Its projects include a Java queue library called Disruptor, a multidimensional index and advertising development tools. This is the machinery behind the marketing language. The queue project even tells adopters to benchmark on their own hardware, a refreshingly specific reminder that performance depends on the environment.
The operational work had its own specialists. In 2019, Agatha Rymanowska described progressing from client integration engineering to leading U.S. operations, with more than 100 people spanning campaign operations, engineering and activation. The employer profile also reported 48 hours of paid annual professional development. Personalization required people who could translate between customer needs and technical systems.
Recognizing someone is only half the test
Conversant’s old website now leads to Epsilon. The successor offering, Epsilon Digital, describes a full-service programmatic platform covering display, connected television, online video and audio. Epsilon’s December 2025 Ocado Ads partnership applies COREid to offsite retail media. In March 2026, Epsilon announced in-store attribution reporting, connecting digital activity with physical-store transactions.
These developments extend the same problem into new places. But a matched purchase does not, by itself, establish that advertising caused it. Buyers should define incremental success and comparison groups before launch. They need usable customer records, permission to process them and enough outcomes to support a meaningful test. Sparse data or disconnected sales records weaken the proposition. The copyable move is to fix the customer record, involve the operational teams early and decide what evidence would justify more spending. The same discipline applies to a buyer choosing between an integrated service and separate tools: test the complete workflow, including the reporting, before expanding the budget.
That is why the hungry customer matters. Dinner was never in doubt. The difficult question was whether the advertisement changed the order.