The Service Excellence Puzzle

The Empirical Anomaly: When Service Logic Inverts

A service organization that serves more than 200 million customers per year, employs roughly 110,000 front-line staff, sustains an annual employee turnover rate above 30 percent, and operates more than 1,300 outlets across four continents is, by every convention of operations management, an organization that should be incapable of delivering distinctive service. Conventional theory is unambiguous on the point: as labor intensity, scale, and dispersion increase, the variance of service outcomes widens, employee-customer interactions drift from any specified standard, and the cost of maintaining a coherent service identity rises faster than revenue can absorb. Yet Haidilao's recorded customer satisfaction scores—a composite drawn from third-party platforms, mystery-shopper programs, and the firm's own post-transaction surveys—consistently exceed category averages in the casual-dining segment by 15 to 25 percentage points, and the differential has not narrowed in any of the last seven fiscal years despite a quadrupling of store count. The gap is not noise; it is structural, persistent, and unexplained by the variables that the service-management literature would predict should govern it.

To state the anomaly in its sharpest form, four variables are simultaneously elevated when only one of them should be:

| Variable | Haidilao Reading | Casual-Dining Median | Theoretical Implication | |---|---|---|---| | Annual customer volume | ~200 million | ~3–8 million per chain | Higher contact frequency increases variance | | Outlet count | ~1,300 | ~50–500 | Scale dilutes founder imprint | | Front-line turnover | ~30%+ | ~25–60% | Selection effects erode quickly | | Mean service duration | ~75–90 minutes | ~45–55 minutes | Longer exposure raises failure probability | | Net Promoter / Satisfaction | +15 to +25 pp above median | Baseline | Inconsistent with above |

Read column by column, each line should point in a different direction. Read across rows, the variables point in the same direction—toward predicted quality erosion—yet the dependent variable refuses to follow. This is the precise configuration that Lakatos termed a "research programme in crisis": a stable empirical regularity that the prevailing theoretical apparatus cannot derive without ad hoc auxiliary hypotheses.

The anomaly has acquired practical urgency. Haidilao's model is now studied by operators of hospitals, banks, hotels, and contact-center platforms, all of whom confront variants of the same question: if this firm can sustain above-category satisfaction under conditions that should defeat satisfaction, what is the firm actually doing that the rest of the industry is not? Trade-press answers have proliferated—each one convenient, each one wrong in a specific and instructive way. The first task of this chapter is to dispose of those answers with the precision they deserve, because until they are disposed of, the analytical work of the book cannot begin.

It is worth pausing on what is *not* anomalous about the firm. Haidilao is not anomalously profitable; its net margins are respectable but not industry-leading, particularly after adjusting for the cost of its waiting-area amenities and discretionary giveaways. It is not anomalously innovative in menu engineering; the hot-pot category evolves slowly and Haidilao's SKU structure is conventional. It is not anomalously advantaged by location; many of its outlets sit in mid-tier retail real estate that would be classed as B-grade by property analysts. What is anomalous, exclusively, is the firm-specific level of front-line service behavior relative to the firm's structural disadvantages. The book treats that anomaly as the dependent variable that requires explanation.

A second clarification is required before the analytical work proceeds. The book's central claim is not that Haidilao is unique in offering "good service." Many firms in the casual-dining segment offer good service at the level of the individual interaction, and a non-trivial subset do so reliably. The claim is narrower and stronger: Haidilao produces above-category service quality *reproducibly*, across thousands of geographically dispersed front-line units, despite a labor pool and turnover profile that are statistically indistinguishable from its peers. The reproducibility claim is what distinguishes the puzzle from a generic "good service is good" tautology and what makes the firm a tractable object of study.

The Three Standard Explanations and Their Limits

Three explanations circulate in management literature, trade press, and consulting decks. Each is empirically testable, each has been tested, and each fails under conditions that are diagnostic rather than marginal. The purpose of this section is not to ridicule these explanations—each captures a partial truth—but to specify precisely the conditions under which each breaks down, because those break-down conditions are precisely the phenomena the book must explain.

The Selection Hypothesis

The first explanation holds that Haidilao's service advantage derives from its ability to hire a qualitatively superior front-line employee. The mechanism is plausible: better hires mean better interactions, better interactions mean better satisfaction, and the firm protects its hiring advantage through a combination of employer branding, referral pipelines, and rigorous screening. The hypothesis is widely held because it is widely held in the broader service literature; Pfeffer's *Competitive Advantage Through People* and the entire "war for talent" tradition lend it intellectual support.

The hypothesis fails on three counts. First, demographic data on Haidilao's front-line workforce is publicly available through headcount disclosures and labor-force surveys: education levels, regional origins, prior experience, and age distributions are statistically indistinguishable from peer chains operating in the same labor markets. The firm does not draw from a privileged labor pool; it draws from the same provincial and metropolitan labor markets that feed its competitors. Second, the turnover rate of 30 percent or higher is incompatible with a stable selection advantage. Even if every hire were initially excellent, the continuous loss of experienced staff to competitors, return-to-hometown migration, and life-cycle events means that the selection effect is being constantly refreshed at the front line; the workforce is never the workforce that was hired. Third, Haidilao's hiring volume—on the order of 50,000 net new front-line staff per year to keep pace with outlet growth—exceeds the supply of plausibly "elite" service workers in any plausible definition, and certainly exceeds the supply that could be drawn from a single firm's referral pipeline.

If the selection hypothesis were the operative explanation, we should observe high variance in satisfaction across outlets and a steady decline in satisfaction at any outlet as tenure of the front-line workforce ages. Neither pattern is observed. Satisfaction scores are tightly clustered around the firm mean, and the score at any given outlet is better predicted by store-manager tenure than by aggregate front-line tenure. The selection hypothesis therefore explains none of the variance in the dependent variable and is rejected.

What the rejection tells us is positive: whatever mechanism produces the service advantage, it must be one that is *robust to workforce composition*. The advantage cannot depend on who is on the floor; it must depend on what they encounter when they get there. This is the first clue that the explanation lies in structure rather than in people.

The Founder Charisma Hypothesis

The second explanation locates the source of service excellence in the imprint of a charismatic founder. Zhang Yong, the co-founder of Haidilao and chairman until 2022, is a recognizable figure in Chinese business culture, and his per