Decisions that made Songbai's people figures smaller and truer

Selecting agents instead of recruiting them

At the end of 2020 Songbai's agency had fallen to 164,000 from 231,000 a year earlier. Every January our branches ran a recruitment drive before the start-of-year sales campaign, paying new agents up to RMB 3,000 a month for three months if they met a low sales threshold. Only 9% of new agents were still selling thirteen months later. About 38% of their first-year business was written on their own lives or their families', and those policies lapsed at almost twice the rate of other agency business. The draft plan for 2022 assumed 120,000 recruits and an allowance budget of RMB 1.1 billion.

I proposed ending the allowance-led drive. Applicants would instead complete a four-week preparation course with a training stipend, pass an assessment that included a recorded needs-analysis conversation, and receive a written statement of what our second-year agents actually earned. Only teams whose managers kept more than 25% of recruits for thirteen months could recruit. Jiahui Rong, head of the individual agency channel, opposed the plan in writing: he expected reported headcount to collapse and feared losing team leaders whose incomes depended on team size. Our chief financial officer worried about new business value. Both papers went to the president with mine.

The president approved the change for every branch from 2022, with quarterly reports to the board. We gave up about RMB 1.1 billion of first-year premium against plan that year, and our agent count fell faster than most peers'. We redirected RMB 380 million of the allowance budget to stipends, trainers and the assessment.

By June 2026 the agency stood at 68,500. Thirteen-month retention reached 31% for the 2024 intake. Average monthly first-year premium per active agent rose from RMB 5,300 in 2021 to RMB 15,800 in 2025, and the share of new agents' business written on their own families fell to 7%. Thirteen-month persistency on agency business rose from 84% to 93%.

The board now receives agency reports by cohort and income band rather than headcount. What I would do differently is the treatment of team leaders. About 2,300 of 9,400 left within eighteen months, many of them capable coaches. We created a salaried training-lead track only in 2023; offered at the start, it would probably have kept several hundred.

Ending forced ranking for the people who check sales

In September 2021 our internal audit team sampled 6,000 recordings of return-visit calls from four provincial service centres. These are the calls an insurer makes to new policyholders during the cooling-off period to confirm that they understood what they bought, a control the regulator expects. In 412 calls marked as completed, the recording showed no conversation with the policyholder. Callers were ranked each year on completion rate and handling time, and the bottom 5% of the operations division received the lowest grade. Two years in that grade meant reassignment or a contract that was not renewed.

We re-contacted the 38,000 policyholders those centres had handled in the period. About 2,900 had never had a genuine call; 1,130 had questions, mostly about participating dividends presented as if they were deposit interest, and 214 chose to cancel. We refunded their premiums in full, RMB 61 million, and reported the findings to the regulator's local offices.

Haoyu Deng, our executive vice president for operations, and our chairman wanted to keep the forced curve and add monitoring, seeing the curve as the discipline of a large operation. I argued that the curve had produced the falsification and that monitoring would only teach people to falsify more carefully. The president agreed to remove forced ranking and bottom elimination for 4,700 staff in return-visit, underwriting, claims review, quality inspection, compliance and internal audit roles, keeping it elsewhere for now. Accuracy would be measured by independent sampling, and grades calibrated across teams without a quota.

We gave up a simple tool for removing weak performers and accepted RMB 41 million a year of extra cost, mainly 96 additional callers, because genuine calls take longer. Nineteen staff and four supervisors were disciplined.

By 2024 the genuine-contact rate in independent sampling had risen from 83% to 98%. Complaints about misrepresentation per 10,000 new policies fell 44%, and cancellations within the cooling-off period rose from 1.1% to 2.4%, which I count as a gain. Voluntary attrition in return-visit teams fell from 29% to 13%.

What I would do differently is the order of consultation. I took the change to the employee representative congress after the decision rather than before it. And the first calibration round produced grades so generous that managers stopped trusting them; publishing the distribution quarterly from the start would have prevented that.

Growing an actuarial bench instead of buying one

In early 2022 Songbai's actuarial function had 148 people, of whom 22 were qualified actuaries. The strengthened solvency rules had just taken effect, IFRS 17 would govern our Hong Kong reporting from 2023, and competitors were offering qualified actuaries rises of 50% or more. We had lost seven in 2021. Peiwen Luo, our chief financial officer, and Ruoxi Shen, our chief actuary, proposed hiring twenty qualified actuaries from the market at packages of RMB 1.5 million to RMB 2.4 million, about RMB 38 million a year.

I agreed that the transition needed experienced people, but argued that buying twenty would reset our pay structure and tell the market that our own people could be bought back. I proposed six senior hires for the hardest valuation and asset-liability roles and a programme to grow the rest: up to ten paid study days for each examination, fees reimbursed, a mentor for every student, rotation every eighteen months and a pay step on qualification. The chief financial officer objected that a graduate pipeline would do nothing for the 2023 interim results. He was right about that.

We settled on six hires and an external consultant to carry part of the transition workload, which cost RMB 52 million across 2022 and 2023. That was the price of not buying the whole team. We also opened a data track: thirty-four actuarial and finance analysts moved into a new data and analytics team of 88 people, so analysts could change direction without leaving the company.

By June 2026 Songbai had 39 qualified actuaries: the original 22, six hires and fifteen internal qualifiers, less four who left. Voluntary attrition in the function fell from 18% to 7%. Spending on external actuarial support fell from RMB 31 million in 2023 to RMB 6 million in 2025. The chief actuary now confirms readiness for each signing responsibility from a dated record of the work each person has led.

What I would do differently is the qualification pay step. I set it at 15% when the market gap was nearer 40%, and four of the first eleven internal qualifiers left within a year. We raised the step to 30% in 2025 and added a deferred retention award. I should have priced the step against the market, not against our own salary bands.

A six-year limit for provincial branch leaders

In 2023 eleven of Songbai's 27 provincial general managers had led the same branch for more than eight years. They knew their provinces, their banking partners and the regulator's local office, and several ran our strongest branches. The audits were less comfortable. Branches led by long-serving general managers averaged 7.6 expense findings per audit against 2.9 elsewhere, including agent training costs that had in fact paid bank-channel fees above the rates filed with our products, exactly what the regulator's insistence that actual expenses match filed assumptions was meant to stop.

I proposed a maximum of six years as general manager in one province, phased over three years, with deputies moved to a second province before promotion and a family package: housing, two return journeys a month, help with school places and support for a spouse's job search. Our head of bancassurance opposed it, expecting partner banks to treat each change of leader as a reason to renegotiate. Three of the general managers concerned said the policy treated them as suspects. The president was cautious and asked for evidence that rotation itself, rather than the extra audit attention, would change the findings.

We gave up some of our best local relationships. Four general managers chose to retire or leave instead of moving. In the first two quarters after each rotation, bancassurance premium in the rotated branches fell 7%, against 1% elsewhere. The family package costs RMB 19 million a year, and each appointment waited several weeks for the regulator's review of the new general manager's qualifications.

By mid-2026 ten general managers had moved. Expense findings in the rotated branches fell to 2.2 per audit, against 3.1 in branches that received extra audit attention without a change of leader, which answered the president's question. Bank-channel premium recovered within three quarters in eight of the ten branches. Deputies with experience in two provinces rose from nine to 28, and 22 branches now have a successor who has worked in more than one province.

What I would do differently is the calendar. The first round ignored the school year, and two strong general managers with children approaching the university entrance examination declined to move; one left Songbai. A one-year deferral for families in that position would have cost little and might have kept both. We adopted it for the second round.