The results behind the announcement
Haidilao, the hot pot chain with more than 1,000 restaurants and over 200 million loyalty members globally, disclosed engagement results from a data-led personalization overhaul built on Infobip's cloud communications platform. Over a six month window, the company reported click-through rate up 2.72 times, click efficiency up 2.37 times, and open rate up 2.14 times against its prior baseline performance. Those are engagement metrics rather than direct revenue figures, but a more than doubled click-through rate across a loyalty base that large represents a genuinely meaningful shift in how many customers actually take action on a given message.
The mechanism behind those numbers was not a single new channel launch or a particularly clever campaign, it was consolidating customer data that had previously been fragmented across markets and channels into one consistent view the company could reliably act on. Seven Lin, Haidilao's product director, described the shift directly, saying that centralizing customer data allowed the company to move from reactive communication toward intelligent, data-backed engagement built on a foundation the whole organization could trust and query consistently across regions. That consistency is what let a single global loyalty team design messaging logic once and have it apply predictably everywhere, instead of every regional office rebuilding its own rules on its own local data.
A genuinely multi-market data problem
Haidilao operates across a set of markets with wildly different dominant messaging platforms, which is exactly the kind of structural fragmentation that breaks a naive, one-size-fits-all customer data strategy applied uniformly everywhere. The company used email and mobile push as consistent baseline channels across every market, then layered WhatsApp broadly across regions, added LINE specifically in Thailand, Taiwan, and Japan, and brought in Zalo for its Vietnam operations. That is a deliberately regional approach to channel selection built on top of a single, centralized customer profile shared across the whole business.
This pattern, one unified customer data layer feeding market-specific channel execution tailored to local habits, is the practical shape most global retail and hospitality brands genuinely need but rarely manage to build cleanly in practice. It requires resisting the tempting shortcut of letting each regional team run its own siloed customer database independently, which is usually the easier short-term choice operationally and precisely the decision that makes coherent personalization at global scale nearly impossible to achieve later on.
From campaigns to autonomous journeys
The data centralization work was the necessary foundation, but Haidilao has since moved a step further by adopting Infobip's AgentOS platform to automate more complex customer journeys using AI rather than relying purely on scheduled campaigns. Daniel Kou, Infobip's head of customer success for APAC, described this as Haidilao taking a bold step to leverage autonomous, AI-driven capabilities that transform diner interactions globally, framing the shift as a move from periodic scheduled campaigns toward always-on, behavior-triggered engagement running continuously in the background.
That sequencing, building clean centralized data first and layering autonomous agent orchestration on top second, is the order most successful enterprise AI rollouts actually follow in practice, even when marketing narratives around agentic AI often imply the model itself does most of the heavy lifting on its own. Haidilao's six-month results happened primarily because the underlying customer data was already consolidated enough to reliably trigger action on, not simply because an AI agent was layered onto messy, siloed data and asked to compensate for gaps it could not realistically fix.
Why this matters beyond restaurants
For retail and consumer brands running large loyalty programs spanning multiple regions, this case offers a useful data point on what a well-executed customer data platform migration can realistically produce within a relatively short window of time. Six months from initial data centralization to more than doubled click-through rate is an aggressive but far from implausible timeline, particularly when the primary blocker heading in was fragmented data infrastructure rather than a fundamentally broken engagement or messaging strategy that would need a longer rebuild.
It also serves as a useful reminder that personalization gains at this kind of scale come primarily from data architecture decisions made early on, not from creative or copywriting improvements layered on afterward. Brands chasing similarly large lifts should look first at whether their customer data is actually unified consistently across markets and channels before investing heavily in the AI layer that ultimately sits on top of it, since an agent orchestrating fragmented, inconsistent data will reliably underperform one working from a single coherent customer view.
The caveat worth keeping in view
These are vendor-reported figures drawn from a customer success story rather than an independently audited case study, and improvements in click-through and open rate do not automatically translate into proportional gains in revenue or long-term retention for the business. Enterprises evaluating a similar data centralization and AI orchestration project of their own should ask any prospective vendor for the equivalent downstream business metrics, repeat visit rate, average spend per visit, and churn, rather than settling for top-of-funnel engagement numbers alone as proof of value.
Even accounting for that caveat, the underlying architecture Haidilao describes here, centralized customer data feeding market-aware channel strategy that in turn feeds autonomous journey orchestration, is a genuinely sound blueprint worth studying closely. It is the sequence PE-backed retail and commerce operators currently evaluating their own martech stack consolidation should be benchmarking their own plans against, regardless of which specific vendor ultimately ends up supplying each individual piece of that stack.
