Decisions that gave up premium to keep promises affordable

Withdrawing a 3.5 per cent savings product before the last-chance rush

In early 2023 our increasing whole-life product, priced at 3.5 per cent, was the most successful savings policy Jiangheng had ever sold. It had produced 44 per cent of our first-year premium in 2022, mostly as single or three-year payments through agents and banks. By the spring, ten-year government bonds yielded less than 3 per cent and thirty-year bonds only a little above 3.2 per cent. The industry expected products priced at 3.5 per cent to be withdrawn within months, and our sales leaders planned a campaign around the withdrawal date worth about RMB 5.8 billion of first-year premium between May and July.

Each yuan would carry that rate for forty years or more. Our asset-liability projections showed the campaign adding a negative spread in most rate scenarios after 2030, and a capital need in the worst of them.

I recommended stopping sales on 30 April, about three months early. The head of individual business argued that agents would join competitors still selling the product; the head of bancassurance warned that banks would give our shelf to rivals; a shareholder director asked why we should forgo business others would write anyway. All three were right. The president accepted my recommendation after the chief actuary supported it.

We launched a 3.0 per cent replacement and a participating annuity in May. Between May and July we sold RMB 1.9 billion of first-year premium instead of the planned RMB 5.8 billion. Full-year first-year premium fell 17 per cent, we dropped seven places in the industry table, and about 2,100 agents left in the second half of 2023, many for insurers running the campaign we had declined.

The benefit arrived slowly. When ten-year yields fell below 2 per cent late in 2024, the policies we had not sold were liabilities we did not carry. The asset-liability committee now approves every savings product's pricing rate against yields actually available that quarter, and our sales plans no longer include withdrawal campaigns.

I would change two things. I would have stopped the short-pay versions in late 2022, when the yield trend was already plain. And I would have explained our reasoning to agents before the decision leaked rather than after; some of those who left would have stayed for an honest account.

Ending twenty-three bank partnerships to earn more from thirteen

At the end of 2023 Jiangheng sold through thirty-six banks: provincial branches of two large national banks, twelve city commercial banks and twenty-two rural commercial banks. Bank-channel first-year premium was RMB 12.1 billion, the highest in our history, and 81 per cent of it was single premium. On our partner-level analysis, only eleven of the thirty-six relationships earned more than their cost of capital once commission, persistency and the capital consumed by single-premium savings were counted. The rest were buying us a place in the premium table.

With rates falling, single-premium savings sold through banks were also the business most exposed to the risk we were trying to reduce. The regulator's rule that commissions paid must match those filed with each product had narrowed the differences between partners, but not the underlying arithmetic.

I proposed keeping thirteen partners, the eleven that earned their cost and two city commercial banks where regular-premium sales were growing, and ending the others at their renewal dates. Our head of bancassurance argued that a smaller footprint would cost us the attention of the large banks' provincial branches, which rank insurers by volume, and that some rural banks were our only presence in counties without agents. He was right on both counts. The executive committee approved the plan in March 2024 on condition that I reported the channel's results to the board every quarter.

Bank-channel first-year premium fell to RMB 7.0 billion in 2025. Regular premium rose from 19 to 58 per cent of the channel's first-year premium, thirteenth-month persistency from 88 to 94 per cent, and the channel's new business value from RMB 210 million to RMB 560 million. One large bank's provincial branch moved our products down its shelf. In nine counties of rural Hubei and Hunan we no longer have any way to sell.

The channel is now managed by value per partner, and bank managers' bonuses follow persistency and regular premium rather than volume. I would do one thing differently. I would have kept those rural banks on a protection-only agreement. Ending the relationships saved little, and it removed us from families who needed medical cover far more than another savings product.

Capping residence rights at the apartments we could provide

When we began selling residence rights for Songhe Court in 2020, two years before it opened, the arrangement seemed simple. A customer who bought an annuity or whole-life policy with total premiums of at least RMB 2 million received a letter giving priority to live in the community after reaching the qualifying age. Letters became the most effective tool our agents had with wealthier customers. By June 2023 we had issued 4,950 of them, and letter-linked policies had produced RMB 6.4 billion of premium in 2022.

Songhe Court has 780 apartments. My team modelled when each letter holder was likely to apply, using age, health and the experience of older communities, against expected turnover. The model showed applications exceeding available places from about 2034, with waits of several years by 2040 if sales continued at their pace.

I recommended capping Songhe letters at 6,200, the number at which projected waiting stayed within a year in our central case, and selling further letters only for partner communities where apartments were contracted. The head of individual business and the general manager of our elderly-care subsidiary both disagreed. They argued that many holders would never apply, that a second community would follow and that the cap would send wealthy customers to competitors. The board's risk management committee supported the cap in September 2023 on condition that it reviewed a residence-right register every quarter.

The Songhe cap was reached in February 2025. Letter-linked premium fell to RMB 3.8 billion in 2025, about RMB 3.1 billion below plan across 2024 and 2025, and some agents moved their clients to insurers less concerned about capacity. We contracted 1,350 reserved apartments with eleven partner operators in eight cities and wrote to every letter holder in 2024 with projected waiting times. Songhe's occupancy reached 91 per cent in June 2026, and complaints from letter holders have stayed in single figures each year.

The register is now a standing risk report, reviewed beside the capital plan. What I would do differently is timing. I would have set the cap in 2020, before the first letter was issued, rather than after nearly 5,000 were in customers' hands; the hardest promises to manage are those written before anyone counted.

Buying stop-loss cover for a crop book before a flood year

In late 2008 the crop book at Lanbei Property and Casualty had grown to RMB 1.1 billion of premium, including government subsidies, across ninety-two counties in Heilongjiang, Jilin and Inner Mongolia. Subsidies had made cover affordable for farmers, and the provincial agriculture department wanted us in more counties. Our reinsurance covered single catastrophic events but not an accumulation of losses over a season, and the book was now large enough for one bad summer to absorb a year of the company's profit.

As deputy head of planning I made two proposals for the 2009 plan. The first was to buy aggregate stop-loss reinsurance covering crop losses between 90 and 150 per cent of premium, at about RMB 96 million a year. The second was to decline seventeen western counties the department had offered us, where drought losses had been severe and programme rates did not reflect them. Our chief financial officer opposed the reinsurance as expensive in an ordinary year. The general manager of agricultural business opposed declining the counties, because the department's goodwill decided which counties we were offered next.

The president accepted the reinsurance at once and, after two further meetings, the decision on the counties. We gave up about RMB 160 million of annual premium. The 2009 season was benign: the stop-loss cost RMB 96 million and paid nothing, and I spent much of that year explaining why.

In the summer of 2010 heavy rain and flooding damaged crops across much of our territory. The crop loss ratio reached 137 per cent on premium of RMB 1.3 billion, and the stop-loss recovered RMB 611 million. Lanbei reported a small loss for the year instead of the capital shortfall our projections had shown without the cover. The board then adopted a written catastrophe appetite and an annual review of accumulation by river basin.

I would handle the western counties differently. Instead of simply declining them, I would have proposed a co-insurance pool with other insurers at a rate reflecting drought history. The farmers there were insured by another company at the same inadequate rate, and that arrangement collapsed after a dry year, leaving them without cover for a season.