AX Field Notes

All casesNo. 060

Haidilao Automates Loyalty Journeys Across Markets

Visit, spend, and coupon data now trigger the next offer, not a regional team's campaign

Case no.
060
Primary source
Infobip customer story and press release
Source published
2026-09-02

1. Why this case

Haidilao is a hot pot restaurant chain with more than 1,000 restaurants across Asia, Europe, North America, and other markets, and a loyalty program with more than 200 million members. Its Infobip press release calls it “the world’s largest hot pot chain.” For a restaurant brand, the loyalty base is where repeat visits come from, and keeping members coming back is a communications problem: who gets which message, on which channel, and when.

The case is useful because it shows where the manual work sat before and where it moved. Haidilao already had the data. Its own CRM and marketing systems knew each member’s last visit, average spend, and coupons. What it lacked was a way to turn that data into timely messages in many countries without regional teams building a new campaign every time. With Infobip, Haidilao connected those signals to a customer data platform and let the platform’s AgentOS run journeys that start from member behavior. In the vendor’s words, “a completed purchase could trigger the next email and offer.”

Over six months, the vendor reports 2.72x higher email click-through, 2.14x higher open rate, and about 5% of targeted inactive members each month returning to dine. These are vendor-published figures from an Infobip customer story and press release, not independently audited results.


2. Problem

According to the Infobip customer story, Haidilao’s hospitality focus means it is “always looking for new ways to stay connected with members, encourage bookings, promote available offers, bring inactive members back, and give customers more reasons to return.”

International growth made this harder in two ways.

  1. Different markets, different channels. “As Haidilao expanded internationally, customers in different markets had different preferences for staying in touch with their favorite restaurant.” Email and WhatsApp matter more in the Americas and Europe. In Asia, LINE, Zalo, KakaoTalk, and SMS are considered depending on the market.
  2. Data was there, but turning it into messages took people’s time. Haidilao “already held valuable data in its internal systems and data warehouse.” The story names the gap directly: “The challenge was turning that information into timely, relevant communication without adding more manual work for its regional teams.”

The project scope was the overseas markets outside Greater China. The goal was to segment members by behavior and channel preference, build automated nurture journeys, and adapt communication to each market.

Confirmed facts: 1,000+ restaurants across Asia, Europe, North America, and other markets; loyalty program 200M+ members; scope = overseas markets outside Greater China; data already in internal systems and a data warehouse; constraint = no additional manual work for regional teams; channel preferences differ by region.


3. Solution

3.1 Member signals connected, internal systems kept as the record

Through APIs, Haidilao sent “selected profile, transactional, and engagement signals from its internally developed CRM and marketing systems” into Infobip’s customer data platform, “while keeping its internal systems as the systems of record.” The signals listed are:

  • the customer’s most recent visit
  • average spend
  • coupon availability and usage
  • purchase and membership activity
  • engagement with previous campaigns

With these fields in one place, the story says, regional teams “could move away from broad campaigns and engage each member according to where they were in their relationship with the brand.”

3.2 Journeys that start from a visit

Email covers the period before and after a restaurant visit: seasonal campaigns, event invitations, coupon reminders, booking prompts, and follow-ups. Mobile push is the global direct channel.

Campaigns are triggered by member activity. In Singapore, the team built a journey to encourage repeat visits: “Customers received a sequence of offers based on their purchase activity, with different incentives introduced as they progressed through the journey.” The story describes the change in the team’s work this way: “Infobip AgentOS automated the flow so a completed purchase could trigger the next email and offer. The team could manage the journey without manually creating a new campaign after every visit.”

3.3 One framework, local content and channels

The same data and journey framework is adapted per market. Content differs: in some regions, messages focus on loyalty benefits, available coupons, and repeat visits; in others, they introduce hot pot dining and tie birthdays and seasonal celebrations to a restaurant booking.

On channels, the two sources describe different stages. The customer story says that today “AgentOS supports Haidilao’s regional email journeys and global mobile push communication,” and that Haidilao is “exploring” WhatsApp, LINE, Zalo, KakaoTalk, and voice. The press release says email and mobile push “served as the performance baseline” and that “the brand is now scaling engagement via WhatsApp to all markets using the channel, as well as LINE in Thailand, Taiwan region and Japan, and Zalo in Vietnam,” using “permissioned member data and engagement signals, subject to local privacy requirements.”

3.4 Agentic AI on top of the centralized data

The press release places AgentOS as the newest layer: “To sustain this growth, Haidilao has recently adopted Infobip’s AgentOS platform. By layering agentic AI on top of its now centralized customer data, Haidilao is automating complex customer journeys.” Infobip’s Daniel Kou, Head of Customer Success APAC, calls Haidilao “an early adopter of Infobip AgentOS.”

Infobip also supported onboarding and implementation, helping Haidilao’s teams set up templates, campaign flows, and processes that could be reused across regions.

Confirmed facts: API integration of five signal types into Infobip CDP with Haidilao systems as systems of record; email for campaigns, invitations, coupon reminders, booking prompts, follow-ups; mobile push global; Singapore repeat-visit journey with staged offers; completed purchase triggers next email and offer; content varies by market; live channels = regional email + global push (customer story); WhatsApp/LINE/Zalo rollout described as scaling (press release) or exploring (customer story); AgentOS recently adopted (press release); Infobip onboarding support.


4. Impact

Figures from the Infobip customer story and the 2026-09-02 press release. The baseline for the multiples is not stated.

MetricFigureBasis
Email click-through rate2.72x higherSix-month period
Click efficiency2.37x higherSix-month period
Email open rate2.14x higherSix-month period
Email delivery rate95%Customer story
Email volume~5 million per monthCustomer story
Inactive members who returned~5% of those targeted each month“Haidilao reported”

The press release credits the email results to “a personalized approach to engagement with data-driven segmentation.” The customer story credits them to “connecting its customer data with automated, personalized journeys.”

On revenue, the only statement is qualitative. Seven Lin, Product Director at Haidilao, said: “Centralizing our customer data has allowed us to move from reactive communication to intelligent, data-backed engagement. By focusing on personalization and prioritizing the specific channels our members use in different regions, we have significantly boosted both our engagement efficiency and revenue from our loyalty base. Infobip’s AgentOS provides the AI communication orchestration we need to ensure this hospitality remains seamless as we expand globally.”

On the regional teams, the customer story says they “now have a more structured way to plan and automate email journeys.”


5. Insight

The work that moved was campaign building. Before, a regional team had to decide each campaign’s audience and send it. After, the team designs a journey once, and member behavior decides who gets the next message and when. A completed purchase starts the next step. The people’s job shifts from producing campaigns to designing journeys and market content.

The data was the precondition, not the product. Haidilao already recorded last visit, spend, and coupon usage in its own systems. The project did not replace those systems. It copied selected signals into a platform that could act on them and left the internal systems as the record. A restaurant chain that cannot yet tie a visit to a member ID would not get the triggers this case depends on.

Churn work became a monthly routine. The inactive-member figure is reported per month: about 5% of those targeted return to dine. That framing implies an ongoing segment that is re-targeted every month, not a one-off win-back campaign.

Channels follow the market, and consent follows the channel. One framework runs email and push everywhere, with WhatsApp, LINE, and Zalo added market by market. The press release ties the regional rollout to “permissioned member data” and “local privacy requirements,” so the consent record has to be part of the member data, not an afterthought.

Limits. All figures come from the vendor. The multiples have no stated baseline, and the two sources credit them slightly differently: data-driven segmentation in the press release, automated journeys in the customer story. The press release says AgentOS was adopted recently to “sustain this growth,” so the six-month results should not be read as the result of agentic AI alone. No revenue, visit-frequency, or redemption figures are published.

One-line takeaway: For a restaurant loyalty program, the step that scales across markets is not writing better campaigns but letting visit, spend, and coupon data start the next message on its own, with regional teams designing the journeys instead of sending each campaign.


6. Implementation hypotheses (not in the source)

6.1 Event sync from the internal CRM to the customer data platform

Basis: The customer story says Haidilao integrated “selected profile, transactional, and engagement signals” through APIs while “keeping its internal systems as the systems of record,” and that “a completed purchase could trigger the next email and offer.”

For a purchase to start the next step of a journey, the visit has to reach the platform as an event soon after the bill is settled, not only in a nightly batch. The likely shape is two feeds. A near-real-time event feed would send a purchase or visit event with member ID, restaurant, spend, and coupon redeemed. A periodic profile sync would update slower fields such as membership tier, average spend, and coupon balance. Because the internal CRM stays the system of record, coupons are probably issued and validated there, and the platform only reads coupon status to decide the message. That split keeps a coupon from being created twice if a journey step fires again.

6.2 Staged repeat-visit journey with exit rules

Basis: The Singapore journey gives “a sequence of offers based on their purchase activity, with different incentives introduced as they progressed through the journey.”

A staged journey like this is typically built as wait-and-branch steps. After a first visit, the member waits a set number of days. If a new purchase event arrives, they move to the next stage and its offer. If not, they get a reminder of the current offer, and later a different incentive. A member who visits drops out of the waiting branch, so no one gets a return reminder the day after dinner. Stage, last offer, and redemption would need to be stored per member so the regional team can see where members drop off and change the incentive for one stage without rebuilding the whole journey.

Basis: About 5% of “the inactive members targeted each month” returned to dine. The press release says regional channels use “permissioned member data and engagement signals, subject to local privacy requirements.”

The monthly figure suggests an inactive segment recomputed each month, for example members with no visit within a set window, then targeted and measured by whether a visit event follows within a fixed period. Channel choice per member would likely follow a priority order by market. In Vietnam, for example, that might be Zalo, then push, then email, then SMS. A member would be eligible for a channel only if the matching marketing consent is stored with the profile, so the consent check runs before the channel choice rather than after a send fails.


7. Sources

  1. Infobip — Haidilao: Orchestrating personalized journeys with AgentOS
    https://www.infobip.com/customer/haidilao
    Publication date: not shown on page (accessed 2026-11-16)
  2. Infobip via PR Newswire — Haidilao scales global loyalty with Infobip’s data-led personalization
    https://kyodonewsprwire.jp/release/202609025124
    2026-09-02 (accessed 2026-11-16)