China's real estate agents hoarded listings and poached each other's clients to survive, so Lianjia paid them to cooperate instead — splitting one commission ten ways rather than handing it all to whoever closed.

the problem

agents in a fragmented brokerage market have no reason to share listings or split the work

background

what everyone would do

The available levers for fixing agent hoarding and client poaching were procedural, better training, a code of conduct, firing agents caught hoarding, all of which addressed the symptom while leaving the underlying winner-take-all commission structure, where whoever closed a deal kept everything, completely untouched.

what they saw

Lianjia saw that agents weren't hoarding listings out of bad character, they were behaving rationally under a compensation structure where any cooperation on a stage of a sale, entering a listing, holding a key, doing a first showing, was a pure loss unless the cooperating agent also closed the deal themselves. The fix wasn't a rule against hoarding, it was unbundling the transaction into roughly ten distinct, separately tracked roles and paying every contributor for their recorded portion of the work, so cooperating with another agent stopped being a financial sacrifice and became a normal, compensated part of doing business.

the move

Lianjia built the Agent Cooperation Network (ACN) in 2014, then extended it platform-wide through the Beike marketplace from 2018 onward: a transaction is broken into roughly ten distinct, separately tracked roles split across the listing side (entering the listing, maintaining and showing it, holding the key, handling paperwork) and the buyer side (referring the client, doing the first showing, closing the deal, financial and transaction advising). The eventual commission is divided among everyone who touched the deal in proportion to their recorded role, turning one winner-take-all payout into several smaller, guaranteed ones.

why it works

Splitting one deal's commission among everyone who touched it, listing entry, maintenance, key-holding, paperwork, client referral, first showing, closing, financial advising, meant an agent no longer had to control an entire transaction end to end to earn anything from it, removing the exact incentive that had driven hoarding and mid-transaction client poaching in the first place. Because every micro-contribution across a deal was now separately guaranteed rather than contingent on personally closing, agents could specialize in and profit from partial roles, and the system converted a single-shot, extract-everything-before-quitting dynamic into what one analyst called a repeated game, agents building a track record of paid contributions across many deals rather than one all-or-nothing transaction. This structural realignment, not better training or enforcement, is what let the network scale from 170,000 agents at Beike's 2018 launch to more than 460,000 by mid-2020, with the majority coming from rival brokerages who chose to join a system paying for partial work rather than compete against it from outside.

the payoff

The network scaled from roughly 121 brands, 20,000 stores and 170,000 agents at Beike's 2018 platform launch to more than 250 brands, 40,000-plus stores and 460,000 agents by mid-2020, with agents from brokerages outside Lianjia itself making up the majority of participants — rival firms chose to join a system that paid their agents for partial contributions rather than compete against it from outside. KE Holdings, Beike's parent, listed on the NYSE in 2020.

where it breaks

The mechanism depends on being able to accurately track and attribute each distinct role in a transaction to the specific agent who performed it, since a system that can't reliably verify who actually entered a listing, held a key, or did a first showing would reopen the same incentive to claim credit for work not actually done, or to dispute contested contributions. It also depends on having enough transaction volume and platform trust to make many small, guaranteed payouts across many deals genuinely more attractive to agents than the occasional large winner-take-all payout the old system offered, a tradeoff that requires real scale to pay off reliably. And unbundling a transaction into paid sub-tasks only removes the incentive to hoard and poach within that specific fragmented market structure — it doesn't address other sources of low trust or quality variance in the underlying service, meaning agents could still cut corners on the quality of any individual sub-task even while being properly compensated for completing it.

what came after

ACN is credited in Chinese real-estate and platform-strategy analysis with converting a low-trust, high-turnover, single-transaction industry into what one Western analyst described as a 'repeated game' — agents build a track record of paid micro-contributions across many deals instead of extracting everything possible from one deal before quitting, a structural fix to industry churn no code of conduct had managed on its own.

references

1. [1]KE Holdings Inc. Form F-1 Registration Statement (ACN model description)U.S. Securities and Exchange Commission, 2020sec.gov
2. [2]Lianjia and Beike's Big Platform Play in Chinese Real EstateJeff Towson, Tech Strategy — Asia, 2020jefftowson.com

keep it

same kind of clever

• 2011 · Airbnb (Host Guarantee)Airbnb underwrote the nightmare so strangers would swap keys
• 1996 · eBayeBay let strangers grade each other and the market policed itself
• 1841 · Mercantile Agency (later Dun & Bradstreet)Lewis Tappan sold ledgers of who could be trusted, a century before databases