Chapter 2

The economics of the guarantee model

FinVolution's reported profit leans on an estimate. Guarantee income — 30% of 2025 net revenue — is the scheduled release of a liability the company books against its own forecast of loan losses, and it is nearly offset by the credit costs charged against the same loans. Reported profit converted to operating cash at about two-thirds over 2022–2025; the gap traces to a growing on-balance-sheet loan book and the working-capital intensity of the guarantee model, not to earnings that never arrive.

What "guarantee income" actually is

For most of the loans it facilitates, FinVolution does not simply match a borrower to a bank and step aside. It stands behind the credit. Under its quality-assurance commitment, when a borrower funded by an institutional partner defaults, a guarantee company — third-party or one inside the group — repays the partner, and FinVolution is then obligated to reimburse that guarantee company for the same amount [1].

That promise creates two entries at the moment a loan is written, both governed by US accounting standards. One is deferred guarantee income, the fair value of the stand-ready obligation, released into revenue over time as the risk runs off [2]. The other is the liability from quality-assurance commitment, set at the expected lifetime credit losses of the covered loans — a figure derived from historical default experience, current conditions, and macroeconomic forecasts [3]. Guarantee income is the release of the first; the cost of the second flows through the income statement as credit losses. Both are management's estimates before they are anything else.

The mechanics are visible in the rollforward. The line "Release of quality-assurance obligations upon repayment" in the deferred-income schedule matches the "Guarantee income" line on the income statement to the thousand — ¥4,124.9 million in 2025 [4]. Behind that net figure sit gross flows an order of magnitude larger: the company paid out ¥7.7 billion to make partners whole in 2025 and recovered ¥4.3 billion from delinquent borrowers [5].

Source: FY2025 20-F, Note 2(s) deferred guarantee income and quality-assurance commitment rollforward [6].

Two balances are shrinking. Deferred guarantee income fell from ¥1,882 million at the end of 2023 to ¥1,119 million at the end of 2025, and the quality-assurance liability from ¥3,306 million to ¥2,575 million [7]. Management attributes the decline to a "decreased proportion of loans bearing credit risk in China," partly offset by more risk-bearing volume overseas [8]. The company is, deliberately, carrying less of its own China credit risk on its books than it used to.

The guarantee book is a thin, credit-sensitive layer

Guarantee income looks large — ¥4.1 billion, roughly 30% of net revenue — but it is not 30% of the profit. Set the guarantee income against the credit losses booked on the same commitment, and the net contribution is modest and volatile.

Source: FY2025 20-F, Consolidated Statements of Comprehensive Income — guarantee income and credit losses for quality-assurance commitment [9].

In 2023 the guarantee book was essentially breakeven: ¥4,479.0 million of income against ¥4,422.8 million of credit losses, a net ¥56 million [10]. By 2025 it earned ¥663 million as losses fell faster than income — about a fifth of the ¥3,101 million pretax profit [11]. Widen the lens to all credit costs — the quality-assurance losses plus provisions for loans receivable and for accounts receivable — and total credit charges of ¥4,526 million in 2025 exceeded guarantee income of ¥4,125 million [12].

The profit engine, then, is fees, not the guarantee. Loan-facilitation and post-facilitation service fees together were ¥6,806 million in 2025, with net interest income of ¥1,336 million; the guarantee layer sits on top as a small, swing contribution whose sign depends on how well the loss estimate holds [13]. That is why the group's net margin compressed even as revenue grew — the same dynamic the Business and Discount chapter flagged: rising credit charges absorb an outsized share of each incremental revenue dollar.

Profit converts to cash at about two-thirds

For a lender that guarantees its book, the more important test is whether the estimated profit shows up as cash. Over the last four years it has, partially. Operating cash flow trailed net profit in three of four years, and equaled it in none but 2024.

Sources: net profit, FY2025 20-F Statements of Comprehensive Income [14] and FY2022 20-F [15]; operating cash flow, FY2025 20-F cash-flow summary [16] (2022 figure restated per note below).

Cumulatively, FinVolution generated ¥6.36 billion of operating cash on ¥9.60 billion of net profit across 2022–2025 — a 66% conversion rate. The single weakest year, 2022, converted just 10% [17]. Three structural gaps explain most of the shortfall, and none of them is a sign that the profit is fictional.

First, a large slice of revenue is non-cash at the operating line. "Net interest income" earned on loans FinVolution originates and holds itself — through consolidated trusts and its overseas books — is recognized as a "net gain from investment in loans" and then removed from operating cash flow, because the actual cash for those loans is classified as investing activity. That reversal was ¥1,336 million in 2025 [18]. The company deployed ¥18.4 billion into these loans and collected ¥16.8 billion back during the year — real cash generation that never touches the operating-cash line [19]. As this on-balance-sheet book grows, reported operating cash understates the group's economic cash flow.

Second, the guarantee model is working-capital intensive. Growth requires posting security deposits with funding partners and building quality-assurance receivables, both of which consume cash as volume expands. In 2022 that drag was acute: a ¥914 million build in quality-assurance receivable and a ¥1,153 million increase in prepaid expenses and other assets — chiefly security deposits — pulled operating cash down toward zero even as profit held near ¥2.3 billion [20]. When that cycle reverses, as it did in 2024, operating cash can exceed profit — a ¥1.7 billion release from prepaid assets drove the 121% conversion that year [21].

Third, the timing of the guarantee accruals themselves — releases of deferred guarantee income and changes in the quality-assurance liability — and rising deferred-tax assets move the operating line year to year without changing the underlying economics [22].

One point on comparability: in the fourth quarter of 2024 FinVolution reclassified certain customer-funds flows between financing and investing activities and restated prior years, which is why 2022 operating cash appears as ¥236.9 million in the latest filing versus ¥268.8 million as originally reported [23]. The change did not touch the profit-to-operating-cash relationship, and the operating figures shown above are on the restated basis.

That the profit is cash-taxed is a useful check against the worry that it is an accounting mirage: FinVolution paid ¥963 million of income tax in cash in 2022, more than double its book tax expense that year, as earlier deferrals unwound [24].

Where the estimate gets tested

The durability of this profit rests on whether the loss estimate embedded in the guarantee liability proves right. That estimate is now being pressed. The 90-day-plus delinquency rate on FinVolution's China book rose from 1.98% at the end of 2023 to 2.13% a year later and 2.85% at the end of 2025 — a 44% increase in the delinquency ratio over two years [25].

Source: FY2025 20-F, Loan Performance Data — 90 Day+ Delinquency Rate [26].

Because guarantee income is fixed at the loan's inception while credit losses accrue against actual performance, worsening delinquency squeezes the guarantee book's thin margin first and the group's reported profit next — the mechanism behind the multi-year margin compression. This is the point on which the durable-versus-trap question turns from the cash-quality side: if realized losses run above the reserved estimate, provisions rise and profit falls; if the estimate proves conservative, the reverse.

Two facts weigh against reading the rising delinquency as an unmanaged deterioration. Management has publicly tightened underwriting — cutting exposure to lower-quality acquisition channels, trimming credit limits for higher-debt borrowers, and deploying earlier-stage AI collection — and has built a reserve buffer: it disclosed a provision-coverage ratio of 543% in the second quarter of 2025, up from 465% a quarter earlier [27]. And the ¥4.3 billion of gross recoveries collected in 2025 shows the collection machinery behind the estimate is real, not notional [28].

The read that fits the evidence: FinVolution's profit is real and cash-taxed, but it is estimate-dependent and credit-sensitive in a way a fee-only platform's is not, and it converts to operating cash at roughly two-thirds because the model funds its own growth in working capital and on-balance-sheet loans. What would change the read is the delinquency trend against the reserve: a coverage ratio that holds or rises while losses stay inside the reserved estimate would confirm the profit is durable; a coverage ratio that falls as delinquency climbs would mark the estimate — and the profit built on it — as too optimistic.