#155 2018 · Haidilao (Zhang Yong) · Restaurants
Haidilao paid store managers more for their protégés' stores than for their own
the problem
Trusted managers had every reason to hold back a deputy good enough to leave
background
A restaurant chain that grows by promoting its best store managers into regional roles faces a quiet contradiction: the manager most capable of training a great replacement is also the person with the most to lose if that replacement leaves to open a competing or cannibalizing store nearby. The usual tools — non-competes, one-time bonuses, loyalty appeals — treat the symptom without changing the underlying math: one manager, one store, one paycheck.
Haidilao changed the math directly. A manager who opted into the mentor-track pay formula earned less from their own store but gained a permanent royalty stream from every store their trained successors opened, and a smaller one from their successors' own successors — a multi-generational override that made building a lineage of trained managers more lucrative than protecting one's own storefront.
what everyone would do
Lock capable deputies in with non-competes, or reward the training manager with a one-time bonus when a protégé is promoted — the standard retention toolkit for a business that grows by cloning its best people. Both leave the underlying math untouched: the manager still earns more by keeping a strong deputy in place than by pushing them out the door to open a new, possibly competing store, so the incentive to quietly hold back top talent survives every policy layered on top of it.
what they saw
Haidilao saw that the problem wasn't a loyalty gap that needed patching with bonuses or contracts, it was that a manager's pay was tied to exactly one store, which meant a deputy's success was structurally a threat rather than an asset. Making the mentor's income a running royalty on every store their trainees opened, and their trainees' trainees opened, turned succession from a risk the manager had to be compensated for tolerating into the single most profitable thing a manager could actively pursue.
the move
Instead of relying on non-competes or retention bonuses to stop capable deputies from leaving — or from being quietly held back by managers who feared training their own replacement — Haidilao let each store manager choose a second pay formula: a smaller cut of their own store's profit (0.4%), plus a share of every store later opened by a manager they had personally trained (3.1%), plus a smaller share of those proteges' own proteges' stores (1.5%). Training the most successful successors, not running the best single restaurant, became the highest-paying job in the company.
why it works
Because the multi-generational override paid more than the manager's own store profit share, the rational move for any manager stopped being 'protect my current position' and became 'train as many capable successors as possible, as fast as possible,' since each one who went on to train further successors kept paying out indefinitely. This aligned the company's growth bottleneck, the supply of trained managers capable of running a new store well, with the one thing every manager was already positioned and motivated to produce, replacing a policy that had to be enforced against self-interest with one that ran on self-interest directly, which is why store openings accelerated to roughly 41-50%+ annual growth without Haidilao needing to build out a separate training or retention bureaucracy.
the payoff
Store openings accelerated sharply, with Haidilao's outlet count growing at roughly 41% annually from 2017 and its opening pace exceeding 50% in 2018 around its Hong Kong IPO. The same incentive later became a documented liability: at a 2021 shareholder meeting, founder Zhang Yong publicly blamed the multiplying royalty structure for driving reckless over-expansion, and the company closed around 300 underperforming stores that year — an honest coda to a mechanism that worked exactly as designed, just for longer than it should have.
where it breaks
An incentive built to accelerate a specific behavior keeps accelerating it even after the optimal rate has been passed, and multi-generational royalties compound that risk because every successor a manager trains also starts training their own successors, so the growth the formula rewards can outrun what the market or the company's operational capacity can actually sustain. That is exactly what happened by 2021: Zhang Yong publicly blamed the same royalty structure for driving reckless over-expansion, and roughly 300 underperforming stores had to close. The approach needs a countervailing check, a cap, a quality gate, a demand signal, that can throttle the incentive once growth stops being the constraint, or the mechanism that solved the succession problem will keep pulling on the expansion lever long after expansion stops being the right answer.
what came after
The 师徒制 (mentor-apprentice) system is credited with the aggressive store-count growth that carried Haidilao to its 2018 Hong Kong IPO, and equally documented as the mechanism Zhang Yong himself later blamed for the over-expansion that forced roughly 300 store closures in 2021 — a case where the same lever that built the company also, unchecked, nearly broke it.
references
- [1]关店300家，海底捞承认"师徒制"翻车，曾被认为是最大的商业机密界面新闻, 2021jiemian.com
- [2]海底捞张勇："穷人思维"丢了首富澎湃新闻, 2021thepaper.cn