Why it held
The reason it held is architectural. ReactorCX stands up a new platform in parallel with live production and replays real traffic against it until the two match transaction for transaction before any cutover. Encore replaced a program Kevin Meiners described as firmly anchored in points-equal-discount-on-product, by unifying four iconic brands into a single membership built around access to fashion and entertainment rather than transactional discounting. The technically interesting part was not that Encore shipped. It was the sequence Gap chose.
1. Relevance comes before revenue, and the order is not negotiable
Kevin framed Gap's strategy around a single sequence: if you are relevant, you drive revenue, and you have to get it done in that order. A large share of Gap's member base already shops more than one of the four brands. What changed with Encore was the decision to talk to those members as people with interests, not as four separate transaction histories. For a CMO, this is the reframe: the loyalty program stops being a discount mechanism and becomes the place where the brand earns the right to be relevant.
2. The Encore Market moved the program off "points equal discount"
Gap's old program was, in Kevin's words, super transactional. Encore did not soften that model; it replaced it. The Encore Market is a marketplace where points buy things a discount cannot: designer collaborations, atelier visits, festival experiences, tastings, opening-day tickets, family nights at the theatre. Gap launched with partners including Disney+, Live Nation, NBCUniversal, AMC Theatres, Playbill and Ticketmaster because those relationships were real before the program needed them. The platform's job is to make redemption against that supply real-time and reliable at the moment a member acts.
3. Speed of culture is an operating constraint, not a slogan
Kevin's proof point was almost offhand: his team built two activation sizzle reels in two months, because he looked at the first one and called it tired. It was eight weeks old. If a brand needs an engineering ticket and a multi-week release cycle to put an offer in market, the cultural moment is gone before the offer is live. Encore was built so that marketers and brand teams can structure and launch offers themselves. ReactorCX addresses it with a single polymorphic activity model and configurable promotion logic that lets program teams build tiers, challenges, streaks and segmented offers without code, while still exposing deep mechanics for advanced users.
4. The hard part was doing two things at once
This was not a like-for-like replacement. Gap re-platformed the loyalty infrastructure and redesigned the program at the same time, and put new developments live on day one, at full scale, with no pilot: a brand-new marketplace launched cold and a shift from static earn-and-burn rules toward non-purchase behaviours, gamification and segmented offers. The standard discipline, including Loyalty Methods' own sequence on prior programs, is to move the platform first and evolve the program after, because you cannot safely observe the blast radius of moving two variables. Gap's case broke that assumption on purpose. The new program logic had never run in production anywhere, so there was no legacy output to validate against for the new mechanics. What made it attemptable was the mechanism underneath and a second validation layer the team added on top.
5. Zero downtime is a process, not a promise
SafeSwitch™ is how Gap migrated without members feeling a thing. The mechanics are specific and worth stating plainly, because zero downtime is easy to claim and hard to demonstrate at this scale. ReactorCX is stood up and run in parallel with live production; real traffic is replayed against it until the two match transaction for transaction; cutover happens only when they do, with rollback armed. The program launched in February 2026 without service interruption: 100M+ Gap Inc. member records migrated, 700M+ Gap Inc. historical transactions migrated.
What to take from it
- Relevance is a design goal with an order of operations, and the platform has to let marketers act on it at the speed of the calendar.
- Migration and program redesign can happen together when correctness is proven in parallel, not assumed at cutover.
- "Zero downtime" should always come with the mechanism that produced it.