Financial Shenanigans

Figures converted from KZT at historical FX rates — see data/company.json.fx_rates for the rate table (NBK official daily rates, period-end). Ratios, margins, and multiples are unitless and unchanged.

Financial Shenanigans

Forensic risk score is 52 / 100 — Elevated. The reported numbers are broadly defensible, but three things keep this above a "Watch" grade: a previously disclosed material weakness in internal controls over the expected-credit-loss (ECL) model that drives lender profitability; a collapsing NPL coverage ratio that fell from 99% (FY2023) to 80% (FY2025) while the gross NPL pile grew 91% over the same window; and a transformative 2025 acquisition of Hepsiburada in hyperinflationary Türkiye that injected $827M of fresh goodwill, $64M of non-cash IAS 29 monetary gains into "other gains", and a $173M segment net loss into a Marketplace P&L that management still presents alongside a steady headline. The single data point that would most change this grade is the FY2026 disclosure that ICFR remains "effective" and that NPL coverage rebuilds above 90% — both confirmed by the FY2026 20-F.

Shenanigans scorecard

The 13-category review is dominated by sector-specific lender risks (ECL, NPL coverage, fintech yield) and acquisition-driven distortions (Hepsiburada). Revenue-recognition and CFO classification tests are largely clean.

Breeding Ground

The breeding-ground profile is mixed: the audit committee is genuinely qualified, the auditor is Big Four, and ICFR has been reported as effective for FY2025 — but founder/insider dominance, an extensive web of related-party transactions, and a high-profile activist short report keep the structural risk above benign.

The audit committee composition is materially stronger than median for a Kazakhstani issuer — three CPAs across the three independent directors, with deep CIS-region E&Y backgrounds. That partially offsets the founder concentration: Kim is also chair of Magnum, and Kolesa is consolidated via a trust deed under CEO Lomtadze's economic interest. The remediated ICFR weakness was in ECL — exactly the area where allowance shortfalls would flow straight to fintech-segment net income.

Earnings Quality

Earnings quality is the area where the most "yellow" turns to "red". Three issues drive concern: NPL coverage collapse, the widening gap between Adjusted EBITDA and IFRS net income, and the Hepsiburada P&L geography.

Revenue grew 60% in 2025 but net income grew only 1% (in fact it fell 6% in USD terms on FX). Roughly $2,014M of the $2,418M revenue increase came from one acquisition (Hepsiburada); cost of goods rose 288% on the same acquisition; provision expense rose 42%; and interest expense rose 47% as the deposit base reprices in a rising-rate environment. The marketplace segment swung to a 20% net-income decline as Hepsi contributed a $173M segment loss. Operating income grew 7% in USD (18% in tenge), well below revenue, which is the right behavior under hyperinflation accounting consolidation.

NPL coverage and Cost of Risk

This is the most important chart in the report. Gross NPLs grew from $535M to $899M in two years — a 68% increase in dollars (91% in tenge) — while the allowance only grew 35% in dollars (53% in tenge). Coverage dropped from 99% to 80%. At FY2025's NPL level, returning coverage to FY2023's 99% would require an incremental ~$170M of provision expense, equivalent to roughly 9 percentage points of pretax margin. Management is not flagging this as a deterioration; instead the disclosure language attributes it to product-mix shifts (BNPL growth) and a slightly higher Cost of Risk (2.0% → 2.2%). Both can be true — but the trend should be priced.

Adjusted EBITDA vs IFRS net income

The reconciliation strips out all interest revenue and interest expense from "other operations" (i.e. payments, marketplace, and parent-level), share-based compensation ($30M), other gains (which in 2025 mostly capture IAS 29 net monetary gain), income tax, and D&A. In 2025 the add-back of "non-fintech" interest expense alone is $711M — driven in part by the new Eurobond financing for Hepsiburada. Calling the cost of acquisition funding "other operations" understates the cost of running the broader group. The gap between Adjusted EBITDA and net income widened from 20% in 2023 to 46% in 2025, even as net income essentially stalled.

Other gains and the Türkiye monetary effect

Reported "other gains" in FY2025 of $35M masks a $64M IAS 29 net monetary gain attributable to Hepsiburada (Türkiye hyperinflation) — net of a $54M financial-asset-and-liability loss. The IAS 29 gain is non-cash and reverses only if Türkiye exits hyperinflationary accounting. It is fully disclosed in the segment commentary, but a casual P&L read overstates underlying "other gains" durability.

Capex, D&A and goodwill

Capex nearly doubled in 2025 ($203M → $351M). D&A almost tripled ($61M → $151M), largely Hepsiburada PP&E rolling into the consolidated base. Goodwill jumped from $33M (legacy Kolesa, Magnum E-commerce, Portmone) to $890M — Hepsiburada accounts for roughly $855M of the $890M. Goodwill is now 4.0% of total assets, up from 0.2%. This is a single-bet impairment risk: if Türkiye operations underperform their purchase-price assumptions, the FY2026–2028 income statement carries the writedown.

Cash Flow Quality

Cash-flow classification is clean — no securitization, no factoring, no inventory swap masquerading as investing. The concerns are about the durability of CFO and the negative free cash flow after acquisitions in 2025.

CFO/NI collapsed from 1.31× (FY2023) to 0.56× (FY2024) and partly recovered to 0.63× (FY2025). The FY2024 dip reflects a roughly $1.0B working-capital absorption as loans to customers grew 36% while customer deposits funded only part of the gap. In FY2025 the addition of Hepsiburada brought in working-capital inflows that helped CFO grow even as net income stalled. The 3-year cumulative CFO/NI ratio (FY2023–25) is 0.80 — sub-1 for a bank-fintech, and a yellow flag if it persists.

Free cash flow after acquisitions

For the first time in five years, FCF after acquisitions is negative — Hepsiburada cost $1,065M net of cash acquired, against $946M of FCF. Management funded the gap by issuing $628M of Eurobonds and skipping the annual dividend ($1,371M paid in 2024, zero in 2025). This is an explicit management trade-off, disclosed in MD&A, but a recurring acquisition cadence of this size would force a permanent dividend cut or balance-sheet leverage step-up.

Accrual ratio

The accrual ratio flipped from −4.4% (FY2023, where CFO exceeded NI) to +6.0% (FY2024, where NI exceeded CFO by ~$971M on a growing asset base). FY2025 is moderating to 4.1% on the Hepsiburada-aided CFO bounce. Anything sustained above 5% would warrant red — for now this remains yellow because the asset-side growth (loan book) is the proximate driver, not aggressive revenue recognition.

Metric Hygiene

Two metric-hygiene issues are worth underwriting. First, "Banking service fees" — historically a meaningful component of Fintech fee revenue — were removed from new contracts mid-2024 and continue to roll off old contracts. This caused Fintech fee revenue to drop 44% in 2025 ($608M → $353M in dollars; same percentage in tenge) and is presented as a one-time effect. Second, the TFV-to-Net-Loan-Portfolio conversion rate has fallen from 2.2 (FY2023) to 1.8 (FY2025), meaning each unit of loan book now generates fewer issuances per period — a deterioration in portfolio velocity that management does not call out.

What to Underwrite Next

The forensic profile does not break the thesis, but it should temper position sizing and force three explicit diligence checks before adding.

Position-sizing implication. A buy thesis on Kaspi at current valuation is defensible — the business generates real cash, the disclosures are extensive by Kazakhstani standards, and management has remediated the ICFR weakness on its own timetable. But the forensic profile means three things. First, an investor should not size to a "high-conviction" level without sight of the first independent Section 404(b) ICFR attestation. Second, an investor should reserve provisioning conservatism — model FY2026 provision expense ~15% higher than guidance to rebuild NPL coverage to 90%, and discount Adjusted EBITDA back to IFRS pretax income for valuation. Third, the Hepsiburada acquisition concentrates downside: a Türkiye macro shock could trigger a goodwill impairment that, while non-cash, would re-anchor multiples lower. The accounting risk here is a position-sizing limiter and a valuation haircut, not a thesis breaker — but if NPL coverage falls below 75% or any class action survives motion-to-dismiss with prejudice, the grade moves to "High".