KNSLKinsale Capital Group, Inc.

KNSL — Kinsale Capital Group, Inc. — dossier, 2026-09-22

Slot names-insurance-1. Model-drafted unattended by the cloud research routine (Opus 5), following .claude/skills/stock-sourcing-pipeline/SKILL.md. Nothing here is human-verified. Sourced facts carry the document they were read from; [background] marks background knowledge and [search-summary] marks anything that came from a search result or a summarising fetch. Private: this directory never reaches dist/.

Access this run is recorded once, in ACGL/dossier-2026-09-22.md §"Access this run", and applies identically here: EDGAR and 3spread reachable, Equibles unreachable (connector failed to connect — 0 of the 25-call budget spent because none could be), ROIC prices yes / calendar paid, fool.com reachable but carrying no Q2 2026 KNSL transcript.

§1 — The finding: the compounder's top line turned, and the moat paid for it

The cluster pass of the same date scored Kinsale ✅ 3/4 — the only pass among eight members — and wrote its falsifier as: "Kinsale has never underwritten through a soft E&S market. Its whole record (2020-2025) is a hardening one." That falsifier is no longer hypothetical. It is in the Q2 2026 10-Q, filed 2026-07-23, three weeks before the cluster pass was written and not read by it.

Gross written premiums fell.

This is the first decline in the company's public record, in a business whose earned premiums went 3.8x in five years. Management names the cause in one sentence and does not hedge it:

"The decrease in gross written premiums was primarily due to a 32.7% decline in the Commercial Property Division, one of the Company's larger divisions, driven by heightened competition." — 10-Q Q2 2026, Item 2 MD&A, Gross written premiums, three months ended June 30, 2026

Two qualifiers that cut in Kinsale's favour, both from the same paragraph: excluding Commercial Property, gross written premium grew 3.7% in the quarter and 4.8% over the half, on "continued strong submission flow across most divisions and an increase in bound accounts"; and the decline is price, not volume — the average premium per policy written fell from ~$14,300 to ~$12,300, −14%, "due to an increase in competition, primarily on larger accounts and in our Commercial Property Division". Kinsale's model has always been small-account; it is the large-account end that is being competed away, which is the end it cares least about.

The part the cluster pass would want to know about

The net line was held nearly flat while the gross line fell 5%. That did not happen by itself. Ceded premium dropped 22% and:

"The net retention ratio was 85.8% for the three months ended June 30, 2026 compared to 82.6% for the three months ended June 30, 2025 due to increased retention in our reinsurance treaties and change in the mix of business." — same MD&A, Net retention ratio

Retaining more has a price and the filing states it plainly: the expense ratio rose from 20.7% to 21.7%, "primarily due to lower ceding commissions as a result of increased retention on our reinsurance treaties". Direct commissions paid as a share of gross written premium barely moved (14.9% against 14.8%), so this is not distribution cost — it is the disappearance of the reinsurers' contribution to Kinsale's overheads.

That matters because the expense ratio is the single scoring point on which the cluster's S0 pass rests. The cluster file's fourth test, the only one Kinsale scored where the other seven members scored zero, reads: "The expense ratio is the moat, not the capital: an owned technology stack in a market where competitors run on brokered paper." One year of a softening market has moved it a full point in the wrong direction, and the mechanism is the company's own response to the softening. The moat is real; it is being spent.

§2 — What has NOT deteriorated, stated as plainly as what has

A soft-market narrative is easy to over-tell. Three series say the underwriting is still excellent:

The current accident year ex-cat loss ratio is the one series a reserve release cannot flatter, and it improved, 58.1% → 57.0%. Annualised ROE for the half was 28.9% and operating ROE 24.4%. Underwriting income was $105.4m in the quarter. Net investment income +19.9% to $55.7m as the float compounds. On the evidence in this filing, Kinsale is writing less business at better margins, which is the correct response to competition and exactly what the cluster pass praised Arch for doing.

One caution on the headline: GAAP EPS grew 34.0% and operating EPS grew 15.9%. The gap is mark-to-market on the equity portfolio — $56.2m of fair-value change in Q2 2026 against $28.6m in Q2 2025, plus $6.7m of realised gains against $0.1m. Any screen reading GAAP EPS growth here is reading the equity market, not the insurance company. Use the operating series.

§3 — The second finding: the first buyback in the company's history

Kinsale's entire investment case, as the cluster file put it, is that "premiums earned 3.8x in five years with the diluted share count flat at 23m throughout. Nobody else in this cluster grew 4x without issuing a share." That is still true of the past. It is no longer the current policy.

• December 2025: the Board authorised the first share repurchase program in the company's public life, up to $250.0m.
• 2026-06-30: $87.5m of capacity remained, i.e. $162.5m spent in roughly six months.
• July 2026: "In July 2026, our Board of Directors approved an additional share repurchase authorization of up to $250.0 million of our common stock, bringing the remaining capacity to $337.5 million under the share repurchase program."
• Share count: 23,145,751 (2025-12-31) → 22,778,425 (2026-06-30), −1.6%, after five years inside 23.0-23.4m.

Put the two findings in the order they happened and the sequence is the ordinary one: the reinvestment opportunity narrows (gross premium −5%, average premium −14%), and the capital comes back instead. The CEO's framing in the release is "We are generating significant operating cash flows resulting in excess capital and are pleased to report an additional share repurchase authorization of $250 million" — excess capital, not a narrower opportunity set. Both descriptions fit the same facts; only one of them is a change in the thesis.

Buying back at 4.01x book is a different act from Arch buying back at 1.41x. It is accretive to EPS and dilutive to book value per share, and for a company whose compounding has come from redeploying capital into premium at ~25% returns on equity, returning it at four times book is a statement about the redeployment rate.

§4 — S2 kill test

(XBRL companyconcept, CIK 1669162.)

Share count over five years: 22.85m (2020) → 23.26m (2025) → 22.79m outstanding at 2026-06-30. Cumulative dilution over the entire five-year 3.8x growth phase: under 2%, now reversing.

Survives or dies: survives outright, with no balance-sheet question to answer. $2.0bn of equity against $2.9bn of reserves, $6.5m of interest expense in a half, no maturity wall, an unlevered underwriter with a 75.5% combined ratio and 24.4% operating ROE. The kill test for this name has never been about survival; it is about whether the growth rate survives a market it has never seen. This pass's answer: the first evidence is in, it is a 5% decline concentrated in one division, the rest of the book still grew, and the margin held. That is a warning, not a verdict.

§5 — S3 gate: the three biggest risks management itself discloses

Kinsale's Item 1A is ordered and the first risk is the one this pass cares most about.

1. Reserve adequacy — listed first, ahead of everything else. "Our loss reserves may be inadequate to cover our actual losses, which could have a material adverse effect on our financial condition, results of operations and cash flows." The same risk factor supplies the figure that makes it concrete: "Approximately 84.2% of our net casualty loss reserves were associated with \"occurrence\" policies as of December 31, 2025." On an occurrence policy the claim can arrive years after the policy lapses — which is why the adverse development Kinsale is already taking on 2017-2019 construction liability is worth more attention than its size suggests.
2. Loss limitations and coverage issues. "The failure of any of the loss limitations or exclusions we employ, or changes in other claims or coverage issues, could have a material adverse effect on our financial condition or results of operations." (Item 1A, cross-referenced from the reserve risk.)
3. Claims inflation and litigation trends, disclosed inside the reserve risk as the named drivers management monitors: "claims inflation, which is the sustained increase in cost of raw materials, labor, medical services and other components of claims cost" and "litigation, judicial and regulatory trends (e.g., increased litigation, higher jury awards and third-party litigation funding, among others)". This is social inflation, named by the company, in the reserve risk, in a book that is 74.1% casualty by gross written premium in H1 2026.

Prior-year development, and why it rhymes with Arch. The Q2 2026 release of $19.4m is a net figure:

"During the three months ended June 30, 2026, prior accident years developed favorably by $19.4 million, of which $22.6 million was attributable to the 2021 through 2025 accident years due to lower emergence of reported losses than expected across most lines of business, particularly in our shorter-tail lines of business."

— offset by adverse development "primarily in the construction liability business in the 2017 through 2019 accident years". Recent short-tail years funding older long-tail casualty, which is the identical shape found in Arch's Note 5 in the same reporting season and documented in ACGL/dossier-2026-09-22.md §1. Two books with nothing in common but the calendar.

Auditor and red-flag sweep. KPMG, "We have served as the Company's auditor since 2009." — i.e. seven years before the IPO. Sole critical audit matter: "Reserves for unpaid losses and loss adjustment expenses". Filing index back to 2010: no NT 10-K or NT 10-Q ever, no 10-K/A, no 10-Q/A. SEC comment-letter correspondence exists — 7 UPLOAD and 11 CORRESP — but it clusters around the 2016 IPO and the most recent exchange closed 2023-10-30, three years ago. Nothing open.

§6 — S5 ownership

13F holders at 2026-06-30 (threespread.py holders KNSL):

The register is doing something interesting: Baron added 636k shares and Giverny added 263k into the quarter in which gross premium declined, while Turtle Creek and First Manhattan each sold more than 400k. That is a genuine disagreement between concentrated active holders, not index drift, and it is the most informative thing in this name's S5.

⚠️ Quarter incompletely ingested — 381 filers against 569 for 2026-03-31; both Vanguard entities, Capital World, Geode, Morgan Stanley, Davenport, JPMorgan, Kayne Anderson Rudnick and Dimensional have no filing in it. No exit may be read from an absence.

⚠️⚠️ The Sixth Street row (199,715 shares, $65.9m, byte-identical to Schwab's) is not a holding. It is a defective 13F filing, traced to the source and documented once in ACGL/dossier-2026-09-22.md §5; threespread.py holders now warns on it automatically.

Proxy (DEF 14A 2026-04-09, as of 2026-03-27): "All executive officers and directors as a group (14 persons) (16)" hold 1,283,428 shares = 5.6% of 23,057,896 outstanding. CEO Michael P. Kehoe alone holds 916,362 = 4.0% (291,734 direct, 16,314 restricted, 22,576 exercisable options, 585,738 through M.P. Kehoe, LLC). The proxy's 5% table, on older 13G/A data, shows BlackRock 9.1% and Baron Capital Group 7.6%.

Shelf: an S-3ASR automatic shelf was filed 2025-08-27 and is the most recent registration on the index. Unlimited by construction, so it hangs over the name on paper; in behaviour the company is repurchasing and has never issued equity for growth. Both halves belong in the answer.

§7 — S8 valuation and sizing (proposals only — ⏳ OPEN (user))

Price $358.19 at 2026-09-21 (ROIC NYSE:KNSL). Shares 22.778m → market cap $8.16bn. Equity $2,035m → BVPS $89.34, P/B 4.01 — by a distance the most expensive name in the cluster and in this slot. ADV $69.6m over the 21 sessions to 2026-09-21. LTM GAAP EPS $24.64 (FY2025 $21.65 less H1 2025 $9.59 plus H1 2026 $12.58) → LTM P/E 14.5x.

The model runs on operating EPS, because GAAP EPS here is one-third equity-market beta. Assumptions: gross written premium −3% in 2026 then flat, earned premium therefore decelerating from +10% to ~+3% by 2028 as the written decline earns through; current accident year ex-cat loss ratio drifting 57% → 60% with competition; expense ratio 21.7% → 23.0% as retention stays high; favourable PYD fading 4.5 → 2.0 points; net investment income compounding ~15% a year on a growing float; tax 20%; buyback at the current pace.

At $358.19 that is 16.3x 2026E and 15.1x 2028E operating EPS. Note the shape: operating earnings are roughly flat 2026-2028 in this model, and the EPS growth that remains comes from the buyback. That is a very different security from the one the five-year record describes.

Base is +6% over ~2.3 years. The distribution is wide and skewed — bear is −42%, bull is +56% — which is the honest description of a high-multiple compounder meeting its first soft market.

Proposed buy price and size: ⏳ OPEN (user). The sizing input the owner needs: ADV $69.6m/day on 22.8m shares is the thinnest of the three names in this slot other than nothing, and the name is a 4x-book compounder, so a position that takes more than a day or two of volume to exit is taking liquidity risk on top of multiple risk.

§8 — S9 monitoring

• Next earnings ~2026-10-22, DERIVED from filing cadence. Q3 2025 results went out 2025-10-23 and Q2 2026 on 2026-07-23; Kinsale reports on the Thursday of the fourth week after quarter end, consistently, with the 10-Q filed the same day. Not a scheduled date — ROIC.ai's calendar needs a paid plan and Equibles did not connect.
• Filing feed: Tier 0 EDGAR sweep covers CIK 1669162.
• Leading indicator between quarters: E&S property rate filings and the wholesale brokers' quarterly submission-flow commentary. Kinsale's own "submission flow remains strong while average premium falls 14%" is a rate story, and rate is observable before Kinsale's next print.
• The specific trigger that converts this to a buy: gross written premium returns to growth including Commercial Property while the current accident year ex-cat loss ratio stays below 60% and the expense ratio stops rising — i.e. the franchise reprices through the soft patch without giving up margin — with the shares below ~3.0x book (≈$280 on today's book, less as book compounds). Today it is 4.01x book with the top line falling; the trigger is a long way off in both dimensions.

§9 — Evidence that is thin, named

• No earnings-call transcript. Equibles did not connect; fool.com has no Q2 2026 KNSL transcript. The Q1 2026 transcript exists there and was read, but WebFetch returns a model summary rather than a locatable span, so it is [search-summary] and quoted nowhere — the one thing it adds, that management described Commercial Property competition as "an abundance of competition and falling rates" on larger layered accounts a quarter earlier, is consistent with the 10-Q and is recorded here as summary only.
• No short interest. Equibles was the route; it is down. The cluster pass did not carry one for this name either.
• No analyst-coverage count. Standing gap across every name slot so far.
• FY2025 operating EPS is estimated, not read. The non-GAAP reconciliation in the FY2025 10-K was not extracted; the ~$18.3 figure is derived from the H1 2025 GAAP/operating gap and is marked "(est.)" in the table.
• Proxy compensation structure not extracted — ownership table only. Fifth consecutive name slot to leave it.
• Division-level premium table not transcribed. The 10-Q publishes gross written premium by division; this pass read the Commercial Property line and the ex-CP aggregate from the MD&A narrative rather than reproducing the table.

Process support, not investment advice.