The Progressive PGR

Reference table

1. What the business is, and how the disclosure works

Progressive describes itself as an insurance holding company writing "personal and commercial insurance throughout the United States." Chief executive Susan Patricia Griffith; approximately 70,053 employees per the vendor profile; Mayfield Village, Ohio; NYSE-listed.

Two reportable underwriting segments as at the June 2026 10-Q — and the vendor's seg_prod block is out of date on the structure.

• Personal Lines — personal vehicles (auto and "special lines": motorcycles, RVs, watercraft) plus personal property (homeowners, renters, umbrella, and flood through the National Flood Insurance Program's Write Your Own arrangement). 88% of companywide net premiums written in the June quarter. Personal auto is approximately 90% of Personal Lines net premiums written.
• Commercial Lines — commercial auto liability and physical damage, business general liability and commercial property for small businesses, and workers' compensation for the transportation industry. Core commercial auto is roughly 80% of the segment.

seg_prod shows a THREE-segment structure through FY2023 and TWO from FY2024 — Personal Lines, Commercial Lines and a separate Property Segment ($2,551.4M in FY2023, $2,270.0M in FY2022) that disappears. This is not a data error: Progressive folded the Property segment into Personal Lines, and the 10-Q's segment note confirms the current two-segment structure. But the vendor supplies no note saying so, and a reader comparing FY2023's three lines to FY2025's two would conclude a segment vanished — the MTZ Power Delivery failure mode. The FY2021 row is a genuine defect: it shows "Agency Channel" $16,881.0M and "Direct Channel" $18,492.3M, which is a distribution-channel split, not a segment split, and totals $35,373M against FY2021 revenue of $47,676M. That row is discarded.

seg_prod FY2025 sums to $82,855M (Personal Lines $71,838M + Commercial Lines $11,017M) against reported total revenue of $87,637M — a $4,782M gap of 5.5%. That gap is investment income, net realised gains, fees and service revenue, which sit outside the underwriting segments. It is not a missing segment; it is the correct structure of an insurer's disclosure, and it is stated here so no reader treats the shortfall as an omission.

seg_geo is EMPTY. Progressive writes only in the United States, so there is nothing to report — this is the CVS/PDD empty-block pattern occurring for a legitimate reason, and it is recorded as legitimate rather than as a defect.

Policies in force — the unit disclosure, from the 8-K of 2026-07-15

Progressive crossed 40 million policies in force during the June quarter, adding 0.5 million in the quarter and 2.8 million year on year. Direct auto at 16,721 thousand is now 1.49x agency auto at 11,211 thousand and growing faster — 10% against 8% — which is the structural expense-ratio advantage compounding. The one weak line is property, up 1%.

2. The combined ratio — the number that decides this dive

From the 10-Q filed 2026-08-03:

And the monthly series, from the three 8-Ks in the archive — a disclosure cadence no other name in this batch offers:

Read the June row carefully, because it is the freshest hard datum in this file and it is not good. Net income fell 31% year on year in a month when the combined ratio deteriorated 3.4 points and net realised gains swung from +$179M to −$13M. May was excellent and April and June were poor, which is the normal noise of catastrophe timing — but the two-out-of-three direction and the quarterly aggregate both point the same way.

Why 87.3 against a 96 goal is a problem rather than an achievement. The 10-K states the objective explicitly: "our long-standing goal of growing as fast as we can, at or below a companywide 96 combined ratio on a calendar-year basis", and adds that "our focus on achieving our underwriting profitability goal takes precedence over growth." Progressive is currently 8.7 points inside its own target. In an industry where prices require state regulatory approval and competitors observe each other's published results monthly, that surplus is temporary by construction. The 10-Q documents all three channels of erosion in its own words:

> "We also reported steady year-over-year growth in both premiums and policies in force despite increased competition in the marketplace."

> Commercial Lines growth was "primarily driven by volume growth due to rate decreases in targeted state and BMT combinations, and increased advertising and agent incentive spend."

> "During the second quarter 2026, advertising expense was $1.4 billion, or 16% higher than the second quarter last year."

And the growth deceleration is the confirming evidence. Personal Lines net premiums written grew 5% in Q2 2026 against 15% in Q2 2025; policies in force grew 8% against 16%. The company is spending 16% more on advertising to produce a third of the premium growth. That is what a market normalising from a hard cycle looks like from the inside.

The arithmetic of the risk, stated plainly. At the current run rate of roughly $86B of annual net premiums earned, each point of combined ratio is approximately $860M of pre-tax underwriting profit, or about $1.16 of after-tax EPS on 581 million shares. Moving from 87.3 to the company's own 96 goal would remove roughly $10 of annual EPS — more than half of it. Nobody, including us, expects that to happen in one step. But it establishes the size of the lever, and it is the reason a 10.6x trailing multiple on this company is not the bargain it appears to be.

3. Growth, margins and the earnings trajectory

That table is the entire cycle in six rows. Progressive earned $9.71 in 2020, collapsed to $1.19 in 2022 when inflation ran through claim severity faster than rate could be filed, and recovered to roughly $19.24 by 2025. Revenue compounded 15.5% a year through the whole thing. The 2022 row is the honest answer to anyone who describes auto insurance as a stable business: net margin was 1.5%, and the combined ratio exceeded 100.

Quarterly, from the filings:

Note the Q3 2025 miss: $4.05 delivered against $4.99 expected — a 18.8% shortfall. That is the single largest deviation in the earnings file and it is a reminder that catastrophe-exposed quarters are not forecastable.

Six-month 2026 against 2025, from the income statement: total revenues $45,797M against $42,413M (+8.0%); net premiums earned $42,541M against $39,719M (+7.1%); losses and loss adjustment expenses $28,399M against $26,409M (+7.5%); other underwriting expenses $6,057M against $5,408M (+12.0%); net income $6,129M against $5,742M (+6.7%); diluted EPS $10.47 against $9.77 (+7.2%).

Two things in that paragraph matter. First, losses are growing slightly faster than premiums earned (7.5% against 7.1%) — that is the combined ratio deteriorating, expressed in dollars. Second, other underwriting expenses grew 12.0%, well ahead of premium, which is the advertising spend showing up.

4. Balance sheet — and why enterprise value is the wrong lens

Filing-verified, 30 June 2026 (10-Q filed 2026-08-03):

Book value per share is $59.09 ($34,333M ÷ 581.0M shares outstanding), against a price of $210.75 — a price-to-book of 3.567x. The vendor's bookValuePerShareTTM of $58.890 uses average basic shares and is within 0.3%.

The company's own capital summary, verbatim from the 10-Q: "At June 30, 2026, total capital (debt plus shareholders' equity) was $42.7 billion, an increase of $5.5 billion from year-end 2025. The increase was primarily driven by $5.2 billion of comprehensive income earned during the first six months of 2026 and the issuance of $1.5 billion of senior notes during the first quarter 2026. These increases were partially offset by the repurchase of 5.4 million of our common shares at a total cost of $1.1 billion."

Enterprise value is REJECTED as a lens for this company, and this is a considered judgement rather than a data correction. km_ttm.enterpriseValueTTM of $130,739,050,000 is market capitalisation plus net debt of $8,209M, which reconciles as debt of $8,387M less cash of $178M and is arithmetically correct. But it is economically meaningless. Progressive holds $97,221M of investments against $45,567M of loss reserves and $27,401M of unearned premiums — the investment portfolio exists to fund policyholder obligations, not to reduce the claim of equity holders. Treating $8.4B of holding-company debt as the only offset while ignoring $97B of investments and $73B of insurance liabilities produces a number with no interpretation. Consequently evToSalesTTM (1.44x), evToEBITDATTM (8.49x), netDebtToEBITDATTM (0.533x) and enterpriseValueMultipleTTM are all discarded, and the valuation in Section 6 runs on price-to-earnings and price-to-book, which are the standard and correct lenses for a property-casualty insurer.

Two further vendor artefacts, both rejected.

currentRatioTTM of 0.237 and workingCapitalTTM of MINUS $62.7 BILLION are meaningless. An insurer does not have a working-capital cycle; unearned premiums and loss reserves are classified current but are funded by a long-duration investment portfolio. These fields would flag Progressive as insolvent on any generic screen, and they are wrong in concept, not in arithmetic.

The vendor's shortTermInvestments classification is internally incoherent and is rejected. bal_a reports $62,168M (FY2023), $75,947M (FY2024) and $19,711M (FY2025) — a $56 billion swing in one year. The 10-Q's actual short-term investments line reads $10,005M at 31 December 2025 and $1,978M at 30 June 2026. The vendor is aggregating portions of the available-for-sale portfolio into a "short-term" bucket on a basis that changes year to year. We use the filing's investment classification throughout and use none of the vendor's current-asset fields.

And intangibleAssets of $2,044M for FY2025 is not intangibles — it is DEFERRED ACQUISITION COSTS, which the 10-Q reports at exactly $2,044M at 31 December 2025. km_ttm.intangiblesToTotalAssetsTTM of 1.77% is therefore measuring capitalised commissions, not intangible assets. Related: bal_a shows goodwill of $228M in FY2023 and FY2024 falling to $0 in FY2025, and Progressive's balance sheet carries no separate goodwill line at all. Neither field is used.

What the balance sheet actually says. Debt of $8,387M against equity of $34,333M is a 0.244 debt-to-equity ratio with 49.8x interest coverage — no financial risk. The genuine risk is on the liability side and no ratio measures it: loss and loss adjustment expense reserves of $45,567M are 1.33x shareholders' equity. A 10% adverse development on those reserves would consume 13% of book value. Progressive's reserving record is good, but that is the exposure, and it is the only one that has ever seriously damaged a well-capitalised auto insurer. Note also that accumulated other comprehensive income swung from +$103M at year-end to −$847M at 30 June on unrealised fixed-maturity losses — a $950M reduction in book value that runs through equity rather than earnings.

5. Cash flow, dividends and the yield that is not a yield

Operating cash flow of $17.5B on net income of $11.3B gives incomeQualityTTM of 1.397 — the gap is the growth in unearned premiums and loss reserves, which is cash received ahead of claims paid. For a growing insurer this is normal and healthy, and it is why free cash flow (13.0% yield) looks so much larger than earnings (9.5% yield). It is not distributable in full: reserves must eventually be paid.

The dividend, and the single most misleading number in this payload. The vendor reports lastDividend: 13.9, dividendPerShareTTM: 13.90 and dividendYieldTTM: 6.595%. Read naively, that is a 6.6% yielding stock. It is not.

The 10-Q gives the actual structure:

The recurring dividend is $0.10 per quarter — $0.40 a year, a 0.19% yield. On top of that Progressive declares an annual variable dividend each December whose size is a direct function of that year's underwriting profit: $13.50 for 2025, $4.50 for 2024 — a threefold swing in one year. The $13.90 the vendor reports is the sum of one $13.50 special and four $0.10 quarterlies, correctly computed and catastrophically mislabelled as a yield.

This matters to the thesis, not just to the arithmetic. If the combined ratio normalises toward 96 as Section 2 argues, the variable dividend shrinks with it, and any investor who bought this stock for a 6.6% yield will discover the yield was the cyclical peak in another costume. dividendPayoutRatioTTM of 69.7% and dividendYieldTTM of 6.595% are both rejected as forward-looking measures and are shown here only to be corrected.

The buyback is small and opportunistic: $166M in FY2025, then $1,100M for 5.4 million shares in the first half of 2026 — an average of roughly $204 per share, close to the current price. Shares outstanding fell from 586M to 581M.

6. Valuation — priced in or room?

At $210.75 (market cap $122.53B, 581.0M shares, book value $34,333M):

Look at the P/E row and note what it does: 11.47x trailing becomes 12.05x, then 13.01x, then 13.15x. The multiple rises going forward at a constant price, because consensus earnings shrink. That is the single most compact statement of the problem in this dive. It is also the opposite of the "forward-multiple rolldown" the format guardrails warn about — there is no rolldown here, there is a roll-UP, and it is not mechanical: it is the market's own analysts saying this company earns less in 2028 than it earned in 2025.

A necessary basis correction before any of these numbers can be used, and it took the filings to establish it. The earn_cal block reports epsActual of $4.85 for the quarter released 2026-07-15. The 10-Q reports GAAP diluted EPS of $5.67 for the same quarter. The $0.82 difference is exactly the after-tax effect of $604M of pre-tax net realised gains on securities (roughly $477M after tax, $0.82 per share on 584.2M diluted shares). Confirmed: earn_cal EPS — and therefore est.epsAvg — is an OPERATING measure excluding net realised gains; inc_q net income is GAAP. The two are not comparable and mixing them would corrupt every growth rate in this section.

A second, larger basis problem in earn_cal, also settled by the filings. earn_cal.revenueActual for 2026-07-15 is $21,077M. The 10-Q's total revenue for that quarter is $23,609M and net premiums earned are $21,573M. Neither matches. The 8-K of 2026-07-15 resolves it: net premiums WRITTEN for the quarter were exactly $21,077M. earn_cal revenue for Progressive is net premiums written, inc_q revenue is total revenues, and they differ by roughly 11%. Any comparison between the two blocks is invalid and none is made in this dive.

est.revenueAvg is on an undetermined basis and is NOT used. FY2025's revenueAvg of $83,927M matches neither the reported total revenue of $87,637M nor the net-premiums-earned base. Rather than guess which, we run all forward valuation on EPS only and say so.

est.ebitdaAvg, est.ebitAvg and est.netIncomeAvg are ALL rejected. ebitdaAvg is exactly 10.72% of revenueAvg in every year from FY2023 to FY2028; ebitAvg is exactly 10.22%. That is the fixed-ratio fabrication signature. Worse, netIncomeAvg EXCEEDS ebitAvg in three separate years — $10,440M against $9,009M for FY2026, $10,572M against $8,579M for FY2025, and $9,501M against $9,499M for FY2027. A taxpaying company cannot earn more after tax and interest than it earned before them. This is the "netIncome > EBIT" defect named in the data contract, present here three times. Only epsAvg survives, and even that carries the operating-basis caveat above.

Peer context — and the vendor peer set is unusable. It contains Bank of America, BBVA, Interactive Brokers, ICICI Bank, Toronto-Dominion and UBS — six banks and a broker. Only Chubb and Travelers are property-casualty insurers. The set has been matched on sector code ("Financial Services"), not on business, and six of eight names have no bearing on Progressive. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn.

6a. What today's price assumes (the inversion)

At $210.75 — 13.0x the FY2027 consensus and 3.57x book — the price embeds:

• Operating EPS falls to $17.49 in 2026 and $16.20 in 2027, then roughly stabilises at $16.03 in 2028. (Consensus; 15, 15 and only 7 analysts.) The price already accepts the decline — this is not a case where the market is ignoring a deterioration. What it assumes is that the decline stops at around $16 and does not continue toward the company's own 96-combined-ratio goal.
• The combined ratio deteriorates from 87.3 toward roughly 90-91 and then holds. (Our number, derived from the consensus EPS path.) Each point is roughly $1.16 of EPS; getting from a trailing operating $18.38 to a 2027 consensus $16.20 is about 1.9 points of combined ratio, offset by continued premium growth. The price assumes normalisation is partial, not complete.
• Policies in force keep compounding at 6-8%. (Our number; the filing shows 7%.) This is the assumption doing the most work in the bull case and it is currently true — but Personal Lines net premiums written growth has already fallen from 15% to 5%, and the company is spending 16% more on advertising to hold it.
• Book value keeps compounding at roughly 15%, and the market keeps paying 3.5x for it. (Our number; equity went from $30,323M at year-end to $34,333M in six months, and from $20,277M at end-2023.) This is the fragile leg. At a normalised 20% return on equity rather than the current 35%, the justified price-to-book on standard assumptions falls to roughly 3.0x, which is $177 on today's book.
• Reserves are adequate. (Our assumption; nothing in the file contradicts it.) $45,567M of loss reserves against $34,333M of equity means a 10% deficiency is 13% of book. No ratio in the vendor payload measures this, and it is the risk that has broken every insurer that has ever been broken.

6b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: EPS growth (NEGATIVE, −7.4% from FY2026E $17.49 to FY2027E $16.20) + multiple drift (EXPANSION, from 12.05x to roughly 13.3x, +10%) + shareholder yield (roughly 1-4%, dominated by a variable dividend that shrinks with the earnings) ≈ +2% to +5%.

This is an unusual bridge and it should be read carefully. For most names the earnings grow and the argument is about the multiple. Here the earnings SHRINK and the base case requires the multiple to EXPAND — from 12.05x on the current year to roughly 13.3x on the next — simply for the price to hold. We assume that expansion because a business whose earnings have troughed and whose unit growth continues normally re-rates on the way through the trough, not on the way into it, and because 3.57x book for a company that will still earn a mid-twenties return on equity after normalisation is not an expensive starting point.

But it means almost none of the expected return is earnings. At a constant 12.05x multiple applied to FY2027E of $16.20, the price would be $195 (−7.4%). At 14x it would be $227 (+7.7%). The base case depends on a re-rating, and that is the fragile leg — stated explicitly, as the format requires.

6c. Variant perception (where we differ, what would surprise)

• We do not differ from consensus on direction, and that is itself the finding. Fifteen analysts model three consecutive years of declining EPS; we agree, and the company's own monthly filings support it. Where we differ is on how much of the decline the price has absorbed. At 13.0x FY2027E and 3.57x book, the price implies partial normalisation. The 870 basis points between the current 87.3 combined ratio and the company's own 96 goal is a larger number than the price is discounting, and while we do not expect it to close fully, we think the risk is skewed toward more deterioration than consensus rather than less.
• We differ sharply from every screen on the dividend. dividendYieldTTM of 6.595% describes a stock that does not exist. The recurring dividend is $0.40 a year — 0.19% — and the balance is an annual variable dividend that was $13.50 for 2025 and $4.50 for 2024, set by that year's underwriting profit. An income investor screening on yield would buy this stock for precisely the wrong reason, and the variable dividend falls with the same combined ratio that drives the earnings.
• We differ from generic screens on solvency and leverage in Progressive's favour. currentRatioTTM of 0.237 and workingCapitalTTM of −$62.7B would flag this company as distressed. It has $97.2B of investments, $34.3B of equity, 49.8x interest coverage and just issued $1.5B of senior notes into a $5.5B increase in total capital. The financial risk here is close to zero; the risk is entirely in the underwriting result.
• The one knowledge-base claim we have runs with us, and it was made three months early. A 2026-04-25 claim from we_study_billionaires (speaker "Chris", bearish, conviction 65) argues that auto insurance is over-earning at 85-87 combined ratios, that regulators will refuse further increases, and that competitors will cut price and erode underwriting profitability. Progressive's own 10-Q, filed 2026-08-03, documents increased competition, rate decreases in Commercial Lines, advertising up 16%, and a combined ratio deteriorating from 86.2 to 87.3. The claim names GEICO rather than Progressive, so it is sector context and not name conviction — but it is the rare case where a sector claim is directly testable against the subject's own disclosures, and the disclosures corroborate it.
• Positive surprise that would force a re-rate upward: July and August monthly combined ratios below 85; Personal Lines net premiums written growth re-accelerating above 8%; or policies in force adding more than 0.5 million in a quarter without a corresponding jump in advertising. Any of these would say the hard market has further to run than consensus allows.
• Negative surprise that would break the thesis: monthly combined ratios above 92 for two consecutive months; adverse prior-year reserve development disclosed in the Q3 filing; policies in force growth falling below 5%; or a materially reduced annual variable dividend declared in December 2026, which would be the company's own confirmation that the profit cycle has turned.

Synthos fair values

Progressive is a cyclical insurer at what appears to be a peak-margin point, so the anchors are built on price-to-book against normalised return on equity, cross-checked against consensus EPS multiples. A trailing P/E anchor would be actively misleading here and is deliberately not used as the primary method.

• Bear ~$150 — 2.2x forward book value of roughly $67.50 per share. Cross-check: 9.3x FY2027E EPS of $16.20; roughly 21% below the 52-week low of $190.40. The scenario: the combined ratio moves to 93-95, close to the company's goal; operating EPS falls toward $12; the variable dividend collapses toward the $4.50 of 2024; return on equity normalises to the high teens; and the market pays a normal-cycle multiple of book. −28.8%.
• Base ~$215 — 3.2x forward book value of roughly $67.50. Cross-check: 13.3x FY2027E EPS of $16.20; 3.64x current book of $59.09. The construction: at a normalised return on equity of roughly 22% (against 35.4% today), a 12% cost of equity and 8% long-run growth, the justified price-to-book is (22−8)/(12−8) ≈ 3.5x, and we apply a discount to 3.2x for the fact that normalisation has not yet happened and its endpoint is unknown. Sensitivity, stated openly: 3.0x gives $203, 3.5x gives $236. The whole answer lives in a 3.0-3.5x band. +2.0%, plus a variable dividend that is genuinely unforecastable.
• Bull ~$270 — 4.0x forward book of roughly $67.50. Cross-check: 16.7x FY2027E; 7% above the 52-week high of $252.68 and 4% above the street's high target of $259. The scenario: competition proves slower to arrive than the filings imply, the combined ratio holds in the high eighties for another two years, policies in force compound at 8%, book value compounds at 18-20%, and the market pays a growth-compounder multiple for what is genuinely the best operator in the industry. +28.1%.

Base is 2.0% above spot; asymmetry roughly 0.98:1 (28.8% down, 28.1% up). A symmetric distribution around a base 2% above the price, on a name with an empty conviction lane and a deteriorating operating metric, is a Watch by arithmetic.

7. Knowledge base — an empty lane, and one bearish sector claim that lands

Raw entity hits naming Progressive: 0. Case-sensitive text hits on the word "Progressive": 0. Name-level conviction: NONE. Sector claims used as context: 1 of 3. Broad sweep discarded in full: 262.

The primary sweep covered entity matches on PGR, Progressive Corp, Progressive Insurance, Progressive Casualty and standalone Progressive, plus a case-sensitive text match on the word "Progressive" across all 51,928 distilled claims. All returned zero. The word "Progressive" — which is also a common adjective in political and tax contexts and would ordinarily generate collision noise — does not appear anywhere in the knowledge base. That is a clean, unambiguous negative result and it is reported as one.

A sector sweep on auto insurance, car insurance, P&C insurance, property and casualty, casualty insurance, underwriting cycle, combined ratio, GEICO and Allstate returned 3 hits, all from we_study_billionaires.

Two are discarded as historical narrative with no forward market content:

> 2025-10-18 · neutral · conviction 68 · entities: GEICO · no named speaker

> "Graham's GEICO shows one shrewd concentrated decision can outweigh a lifetime of diversified journeyman efforts — stay flexible on core principles."

> 2022-10-17 · bullish · conviction 85 · no entities · no named speaker

> "Quality, honest management is decisive; the Geico turnaround under Jack Byrne showed the value of backing a highly capable manager in a distressed business."

Both are investment-philosophy anecdotes about events of 1948 and 1976 respectively. Neither is a claim about the auto-insurance market today. Discarded.

One is directly on point and is used as labelled sector context:

> 2026-04-25 · BEARISH · conviction 65 · speaker: Chris · channel: we_study_billionaires

> "Auto insurance is over-earning (Geico ~85-87% combined); regulators won't grant more price increases, so competitors cut prices and underwriting profitability erodes going forward."

Why this one earns its place. It is falsifiable, dated, mechanistic, and it names the exact combined-ratio band Progressive is operating in. It was made 100 days before this dive, and in the intervening period Progressive's own 10-Q has documented every step of the mechanism it describes: increased competition, rate decreases in Commercial Lines, advertising up 16%, growth decelerating from 15% to 5%, and the combined ratio moving from 86.2 to 87.3. It names GEICO, not Progressive, so under the attribution discipline it is sector context and contributes NO name-level conviction — but it is the only claim in this batch so far where a sector observation has proved directly testable against the subject company's subsequent filings, and it passed.

A broad sweep on "insurance" and "insurer" returned 262 hits and is DISCARDED IN FULL as a lane. The channel distribution (odd_lots 33, business_breakdowns 22, brent_johnson 13, eurodollar_university 13, andreas_steno 12) makes the composition obvious: health insurance, life insurance, deposit insurance, crypto-custody insurance, and macro claims about insurers as marginal buyers of long-duration bonds. Not one names Progressive, and using them would be padding.

Conclusion, stated as a finding rather than dressed up. The Synthos knowledge base has no view on Progressive, and no view on personal automobile insurance as an industry beyond a single bearish claim. That claim happens to be right so far, and it happens to point the same way as the filings and the consensus estimate stack — but one sector claim at conviction 65 from an unnamed-channel-attributed speaker is not a thesis, and we are not going to pretend otherwise. Conviction rating Low, breadth 1, net conviction negative-low, name-level claim count zero.

8. Data integrity — what we rejected and why

Eleven findings. Progressive's file is one of the more defective in this batch, and the reason is structural: the vendor's generic industrial/commercial schema does not fit an insurance balance sheet, and several fields are wrong in concept rather than in arithmetic.

1. earn_cal.revenueActual is NET PREMIUMS WRITTEN, not revenue — established from the filings. The 2026-07-15 row reports $21,077M. The 10-Q's total revenue for that quarter is $23,609M and net premiums earned are $21,573M. The 8-K of 2026-07-15 reports net premiums written of exactly $21,077M. earn_cal revenue and inc_q revenue are on different bases and differ by roughly 11%; no comparison between them is made in this dive.

2. earn_cal.epsActual and est.epsAvg are OPERATING EPS excluding realised gains; inc_q net income is GAAP — established from the filings. earn_cal reports $4.85 for the June quarter; the 10-Q reports GAAP diluted EPS of $5.67. The $0.82 gap is the after-tax effect of $604M of pre-tax net realised gains on securities, which reconciles to within a cent on 584.2M diluted shares. Trailing operating EPS is $18.38 and trailing GAAP diluted EPS is $19.94; both are shown, labelled, and never mixed.

3. est.netIncomeAvg EXCEEDS est.ebitAvg in three separate years — REJECTED. FY2026: net income $10,440M against EBIT $9,009M. FY2025: $10,572M against $8,579M. FY2027: $9,501M against $9,499M. A taxpaying, interest-paying company cannot earn more after those charges than before them. This is the exact defect named in the data contract, present three times. netIncomeAvg is not used anywhere.

4. est.ebitdaAvg and est.ebitAvg are fixed-ratio fabrications — REJECTED in full. ebitdaAvg is exactly 10.72% of revenueAvg in every year FY2023 through FY2028; ebitAvg is exactly 10.22%. Combined with finding 3, the entire non-EPS portion of the estimate block is unusable.

5. est.revenueAvg is on an undetermined basis — NOT USED. FY2025's $83,927M matches neither the reported total revenue of $87,637M (4.2% below) nor the net-premiums-earned base. Rather than assume, we run all forward valuation on epsAvg alone and state the limitation.

6. All enterprise-value metrics are REJECTED as economically meaningless for an insurer. enterpriseValueTTM of $130.739B is arithmetically correct (market capitalisation plus $8,209M of net debt, being $8,387M of debt less $178M of cash) but treats a $97,221M investment portfolio held against $72,968M of insurance liabilities as if it were irrelevant to enterprise value. evToSalesTTM (1.44x), evToEBITDATTM (8.49x), evToOperatingCashFlowTTM (8.00x), evToFreeCashFlowTTM (8.20x), netDebtToEBITDATTM (0.533x) and enterpriseValueMultipleTTM are all discarded. Valuation runs on price-to-earnings and price-to-book.

7. currentRatioTTM (0.237), quickRatioTTM (0.237) and workingCapitalTTM (−$62,726M) are conceptually inapplicable — REJECTED. An insurer has no working-capital cycle; unearned premiums and loss reserves are classified as current liabilities but are funded by a long-duration investment portfolio. On a generic screen these fields would classify Progressive as insolvent.

8. bal_a.shortTermInvestments is internally incoherent — REJECTED, filing used. The vendor reports $62,168M (FY2023), $75,947M (FY2024) and $19,711M (FY2025) — a $56 billion year-on-year swing. The 10-Q's short-term investments line reads $10,005M at 31 December 2025 and $1,978M at 30 June 2026. The vendor is bucketing portions of the available-for-sale portfolio on a basis that changes year to year. None of the vendor's current-asset classification is used; the filing's investment table is used throughout.

9. bal_a.intangibleAssets is DEFERRED ACQUISITION COSTS, and goodwill is unreliable — both REJECTED. The vendor's intangibleAssets of $2,044M for FY2025 matches the 10-Q's deferred acquisition costs line at exactly $2,044M — capitalised commissions, not intangible assets. goodwill is reported as $228M for FY2023 and FY2024 and $0 for FY2025, while Progressive's balance sheet carries no separate goodwill line at all. km_ttm.intangiblesToTotalAssetsTTM of 1.77% is therefore measuring the wrong thing and is not used.

10. dividendYieldTTM of 6.595% presents a variable special dividend as a recurring yield — CORRECTED from the filing. The 10-Q shows the recurring dividend is $0.10 per quarter, $0.40 annually, a 0.19% yield, plus an annual variable dividend of $13.50 for 2025 and $4.50 for 2024 whose size is set by that year's underwriting profit. The vendor's $13.90 is correctly computed (one $13.50 special plus four $0.10 quarterlies) and badly mislabelled. dividendPayoutRatioTTM of 69.7% inherits the same problem. The corrected reading: 0.19% of certain yield, plus a variable payment that falls with the combined ratio.

11. seg_prod mixes segment structures across years and contains one channel-split row — partially rejected. FY2022 and FY2023 show a three-segment structure including a Property Segment ($2,551.4M in FY2023) which no longer exists; the 10-Q confirms the current two-segment structure with property folded into Personal Lines. That is a real reorganisation, not a data error, but the vendor supplies no note and a naive reader would see a vanished segment. The FY2021 row is a genuine defect: it reports "Agency Channel" $16,881.0M and "Direct Channel" $18,492.3M — a distribution-channel split summing to $35,373M against reported FY2021 revenue of $47,676M. That row is discarded. Note also that FY2025's two segments sum to $82,855M against $87,637M of total revenue; the $4,782M difference is investment income, realised gains, fees and service revenue outside the underwriting segments, which is correct disclosure structure and not an omission.

seg_geo is EMPTY — and this is legitimate. Progressive writes only in the United States. The CVS/PDD empty-geography pattern occurs here for a real reason and is recorded as such rather than flagged as a defect.

Not defects, correctly reported and independently confirmed: inc_q net income for the June 2026 quarter ($3,311M) matches the 10-Q exactly; Q2 2025 net income ($3,175M) matches; bal_a FY2025 total assets ($123,039M), total liabilities ($92,716M), equity ($30,323M) and debt ($6,897M) all match the 10-Q's December 2025 comparative column to the dollar; interestCoverageRatioTTM of 49.8x is consistent with $8,387M of debt and $158M of half-year interest expense; incomeQualityTTM of 1.397 correctly reflects operating cash flow of $17,548M against net income of $11,308M; profile.currency is USD and inc_a[0].reportedCurrency is USD — no currency mixing; and tech.hi52/lo52 ($252.68/$190.40) are internally consistent with the moving averages.

8-K scan for corporate actions postdating the statements — completed. The three retained 8-Ks (2026-05-20, 2026-06-17, 2026-07-15) are monthly results releases, all three carrying the full Exhibit 99 press release with income statement and policies-in-force tables. No merger, acquisition, divestiture or capital-structure event appears. The one financing event of the period — the $1.5B senior-notes issue in Q1 2026 — is already reflected in the 30 June 2026 balance sheet's $8,387M debt figure and is disclosed in the 10-Q. Verified clean, and stated as a finding.

A note on the quote block that is NOT a defect. quote.yearHigh is $254.93 and tech.hi52 is $252.68 (intraday versus close); more usefully, tech.pct_from_hi is −16.6% while tech.max_dd_from_peak is −27.6%. These differ because the six-year peak is higher than the 252-day high — Progressive's all-time peak predates the last twelve months. This is the legitimate case and is the reverse of the KLAC situation where the two coincided. Both are correct.

Non-equity tripwire — checked and passed. PGR is common shares, $1.00 par value, NYSE-listed (10-Q balance sheet). The $210.75 price is not par-like; beta is 0.246, which is genuinely low and warranted a second look — but the 52-week range of $190.40 to $252.68 is a 33% band, far wider than a fixed-income instrument, the dividend is variable and swung threefold in one year, volume was 2.04M shares (~$430M of turnover), and the shares carry ordinary voting common-stock characteristics. This is common equity, and the low beta is a genuine property of personal-lines insurance rather than an instrument artefact.

9. Technicals

• Price $210.75. −16.6% from the 52-week high of $252.68; +10.7% above the 52-week low of $190.40; −27.6% from the six-year peak. Position within the annual range: 32.7th percentile.
• The moving averages are the striking feature: 50-day $211.06 and 200-day $210.66, against a close of $210.75. All three within 0.2% of each other. This is a fully flat structure — the stock has gone precisely nowhere for two hundred trading days and is sitting exactly on both averages. It carries no directional information at all, and any technical read of this chart is a fabrication.
• RSI 56.9 — neutral. MACD −0.70 — marginally negative and effectively flat.
• Relative performance, and this is the real technical story: 3-month +5.8% vs SPY +7.6%; 6-month +3.8% vs SPY +11.1%; 12-month −12.6% vs SPY +24.3% and QQQ +30.8%. Progressive has underperformed the S&P 500 by 36.8 points over twelve months — the worst relative performance of any name in this batch so far.
• Sentiment: 15 buy, 22 hold, 4 sell across 41 analysts, vendor label "Hold". Consensus target $225.92 (+7.2%), median $225, high $259 (+22.9%), low $198 (−6.0%).

What the technical picture does to the entry. This is neither a broken chart nor an attractive one; it is a stock that has spent a year converting a valuation excess into a valuation neutrality while its operating metric slowly deteriorated. The de-rating is genuine — 12.6% down against a market up 24.3% is a 37-point swing — and it is why the multiple is now 10.6x trailing rather than the 15x-plus it commanded at the peak. But the flat 50-day and 200-day mean the market has reached no conclusion, and our own base of $215 sits 2.0% above the price. There is no urgency in either direction, which is exactly what a Watch looks like.

10. Insiders

The insider block for Progressive contains no transactions usable for inference. No open-market purchase or sale by any officer or director appears in the last eight records available in the file. This is reported as an absence of signal rather than as a neutral signal, because the underlying data does not support even that much.

What can be said about ownership behaviour comes from the company rather than from individuals: Progressive repurchased 5.4 million shares for $1.1 billion in the first half of 2026, an average of roughly $204 per share — 3.2% below the current price. That is a modest programme (0.9% of shares outstanding) and is described by the company as offsetting equity-compensation dilution rather than as a valuation statement. Shares outstanding fell from 586 million to 581 million.

11. Verdict, kill-criteria and flip conditions

Watch.

What is genuinely excellent: the lowest-cost operator in US personal auto, with 40,086 thousand policies in force growing 7% and 2.8 million added in twelve months; direct auto growing 10% against agency's 8%, which is the expense advantage compounding; $97.2 billion of investments and $34.3 billion of equity against $8.4 billion of debt, with 49.8x interest coverage; total capital up $5.5 billion in six months; book value compounded from $20.3 billion to $34.3 billion in thirty months; a 35.4% return on equity; revenue compounded 15.5% a year from FY2020 to FY2025; and a management team that publishes its combined ratio every single month, which is the most transparent disclosure regime of any company in this batch.

What we are declining to pay for: a 87.3 combined ratio against the company's own 96 goal, with the gap already closing — 86.2 to 87.3 year on year, Personal Lines 86.0 to 87.6, June alone at 90.0 against 86.6 with net income down 31%; net premiums written growth falling from 15% to 5% while advertising rises 16%; a consensus estimate stack modelling EPS declines in 2026, 2027 and 2028 such that the forward P/E rises from 12.1x to 13.2x at a constant price; a 6.595% headline dividend yield that is really 0.19% plus a variable payment tied to the same underwriting profit that is normalising; a base fair value of $215, just 2.0% above spot; and a knowledge-base lane containing zero claims naming this company and one bearish sector claim that the filings corroborate.

The distinction that matters. Progressive is not a bad business and this is not a bearish note about its quality — it is the best-run company in its industry and the unit growth proves it. The issue is that the earnings being capitalised at 10.6x are the top of an underwriting cycle, the company's own monthly disclosures show that cycle turning, and its own analysts have already written the decline into their models. A cheap-looking multiple on peak cyclical earnings is the single most reliable value trap in property-casualty insurance, and the arithmetic here — 870 basis points of surplus margin, each point worth $1.16 of EPS — quantifies exactly how much room there is to fall. With our base 2.0% above the price, symmetric risk, and no name-level conviction, Watch is the arithmetic answer.

Pre-registered KILL criteria — what would take this to Avoid:

• Monthly combined ratios above 92 for two consecutive months, which would put the full-year figure on a path toward the low nineties and take roughly $5 off annual EPS.
• Adverse prior-year reserve development disclosed in any quarterly filing, against loss reserves of $45,567M that are 1.33x shareholders' equity.
• Policies in force growth falling below 5%, which would remove the one metric currently offsetting the margin decline.
• Personal Lines net premiums written growth turning negative, after falling from 15% to 5% in a single year.
• An annual variable dividend declared in December 2026 below roughly $6 per share — the company's own confirmation, in cash, that the profit cycle has turned.
• Advertising expense growing faster than 20% year on year while premium growth decelerates further, which is the signature of buying share at a losing price.

Pre-registered FLIP conditions — what would take this to Buy — Tactical:

• Price below approximately $180 (roughly 2.7x forward book, 11.1x FY2027E) with policies in force still growing above 6%. At $180 the base case is +19% and the downside case is only 17% further. That is 14.6% below today's close and 5.5% below the 52-week low.
• Two consecutive monthly combined ratios below 85 with net premiums written growth re-accelerating above 8%, which would argue the hard market has further to run than either consensus or the knowledge base allows.
• Consensus EPS estimates for 2027 revising UPWARD, reversing the three-year decline the current price is built on. This is the cleanest single confirmation available.
• Evidence that the property book is scaling — policies in force growth above 5% against the current 1% — which would add a genuinely new market to a company that currently has one.
• Genuine name-level knowledge-base conviction appearing. At present the lane is completely empty; a single credible claim about Progressive specifically would be worth revisiting the weighting.

Where PGR fits in the Synthos Framework Portfolio. The insurance sleeve, at 0% today with a 1.5% target on a fill near $180 or on two consecutive sub-85 monthly combined ratios. Batch overlap note: there is no other insurer in this batch, so the position competes for the low-beta defensive allocation with MO rather than with a direct peer — and the two are very different animals, one a declining-volume cash machine and the other a growing-unit cyclical. On the specific question a reader is most likely to ask — is Progressive cheap? — the answer this dive gives is: it is cheap against the earnings it just made and roughly fair against the earnings it is about to make. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $210.75.

Single biggest risk: the 870 basis points of excess underwriting margin close faster than the price allows. Progressive is running a 87.3 combined ratio against a self-declared 96 goal. Every point of that gap is roughly $860 million of pre-tax underwriting profit and $1.16 of after-tax EPS. The company's own filings document competitors cutting price, its own Commercial Lines cutting rate, its own advertising up 16%, its own premium growth cut by two-thirds, and its own combined ratio moving 1.1 points the wrong way in twelve months — with the most recent single month, June, printing 90.0 and net income down 31%. Consensus models a partial normalisation to roughly $16 of EPS. A full normalisation to the company's own target would take about $10 off annual earnings. At 3.57x book and 13.0x a forward estimate that is already falling, there is very little in the valuation to absorb the difference between partial and full.

Provenance & disclosures

• Traceability: ZERO tagged knowledge-base claims name Progressive (raw entity hits 0, case-sensitive text hits on the word "Progressive" across all 51,928 claims: 0; name-level conviction: none). This is an empty lane and it is reported as empty. A sector sweep on auto insurance, car insurance, P&C insurance, property and casualty, casualty insurance, underwriting cycle, combined ratio, GEICO and Allstate returned 3 hits, all from we_study_billionaires: two are historical investment-philosophy anecdotes about GEICO (Graham's 1948 position and the Jack Byrne turnaround) with no forward market content and are discarded; one — 2026-04-25, speaker "Chris", BEARISH, conviction 65 — argues that auto insurance is over-earning at 85-87 combined ratios, that regulators will not grant further increases, and that competitors will cut price and erode underwriting profitability. That claim names GEICO rather than Progressive, so it contributes zero name-level conviction and is used only as labelled sector context — but it is directly testable against Progressive's subsequent filings and the 10-Q filed 2026-08-03 corroborates every step of the mechanism it describes. A broad sweep on insurance / insurer returned 262 hits and is discarded in full as a lane — health, life, deposit and crypto-custody insurance plus macro claims about insurers as bond buyers, dominated by odd_lots (33), business_breakdowns (22) and brent_johnson (13), with not one naming Progressive. No claim in any surviving set carries a management speaker_role, so the half-weight management discount does not apply; no channel-concentration test was required because the usable set is a single claim from a single channel, which is itself disclosed as the weakest possible breadth. All quotes are verbatim from the stored claim text.
• Data as-of: most recent filing-verified financials 2026-06-30, from the 10-Q filed 2026-08-03 (income statement, balance sheet, segment combined ratios, dividend schedule, capital summary), supplemented by three 8-K monthly results releases whose Exhibit 99 press releases were fully extracted and supply the monthly combined ratios and the policies-in-force table · annual statements FY2025 ended 2025-12-31 from inc_a/bal_a/cf_a, corroborated by the 10-K filed 2026-03-02 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873602 = 2026-08-04T20:00:02Z ($210.75, +0.14%; 50-DMA $211.06; 200-DMA $210.66; RSI 56.9; MACD −0.70) · knowledge-base claims 2026-08-04. Fiscal note: Progressive's fiscal year is the calendar year, and the company reports results MONTHLY as well as quarterly. All figures come from the Synthos vendor data file for PGR or from the SEC filings in the PGR archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-03-02 (fiscal 2025); 10-Q filed 2026-05-04 (March 2026 quarter); 10-Q filed 2026-08-03 (June 2026 quarter); 8-K filed 2026-05-20 (April monthly results, Exhibit 99 fully extracted); 8-K filed 2026-06-17 (May monthly results, fully extracted); 8-K filed 2026-07-15 (June monthly and second-quarter results, fully extracted, including the policies-in-force table). All three 8-Ks carry the complete press release rather than a cover page, which for a monthly reporter means this archive contains three independent monthly datapoints that no other name in this batch offers. The July monthly release, due around 2026-08-19, is not yet in the archive.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): earn_cal.revenueActual of $21,077M identified as NET PREMIUMS WRITTEN — the 8-K of 2026-07-15 states that figure exactly — against inc_q total revenue of $23,609M, an 11% basis difference, with no comparison drawn between the blocks; earn_cal.epsActual and est.epsAvg identified as OPERATING EPS excluding realised gains, reconciled to GAAP diluted EPS of $5.67 through $604M of pre-tax net realised gains ($0.82 per share after tax); est.netIncomeAvg rejected for exceeding est.ebitAvg in FY2025, FY2026 and FY2027; est.ebitdaAvg and est.ebitAvg rejected as fixed-ratio fabrications at exactly 10.72% and 10.22% of revenue in every year; est.revenueAvg not used because its basis matches neither total revenue nor net premiums earned; all enterprise-value metrics rejected as conceptually inapplicable to an insurer holding $97,221M of investments against $72,968M of insurance liabilities; currentRatioTTM of 0.237 and workingCapitalTTM of −$62,726M rejected as conceptually inapplicable; bal_a.shortTermInvestments rejected for swinging $75,947M → $19,711M year on year against a filed short-term investments line of $10,005M; bal_a.intangibleAssets of $2,044M identified as DEFERRED ACQUISITION COSTS, matching the 10-Q line exactly, and goodwill rejected as unreliable since Progressive carries no separate goodwill line; dividendYieldTTM of 6.595% corrected — the recurring dividend is $0.10 per quarter (0.19% yield) plus an annual variable dividend of $13.50 for 2025 against $4.50 for 2024; seg_prod FY2021 row discarded as a distribution-channel split ($35,373M against $47,676M of revenue) rather than a segment split, and the FY2022-FY2023 three-segment structure identified as a genuine reorganisation with the Property segment folded into Personal Lines. Where vendor and filing AGREED — worth recording: bal_a FY2025 total assets, total liabilities, shareholders' equity and debt all match the 10-Q's December 2025 comparative column to the dollar; inc_q net income for Q2 2026 ($3,311M) and Q2 2025 ($3,175M) match the income statement exactly; enterpriseValueTTM reconciles precisely to market capitalisation plus $8,387M of debt less $178M of cash (the objection is conceptual, not arithmetic); seg_geo is correctly empty because Progressive writes only in the United States. The 8-K scan for corporate actions postdating the statements returned NOTHING beyond the $1.5B senior-notes issue already reflected in the June balance sheet — verified clean.
• Basis note: consensus estimates and earn_cal actuals are operating EPS excluding net realised gains on securities; reported diluted EPS is GAAP. Trailing operating EPS is $18.38 against trailing GAAP diluted of $19.94, an 8.5% gap that is entirely investment marks. The earn_cal basis is consistent across all reported rows, so the beat/miss series is like-for-like — one miss (Q3 2025, $4.05 against $4.99, an 18.8% shortfall), one narrow miss (Q1 2026, $4.81 against $4.85) and two beats (Q4 2025 and Q2 2026). Forward multiples in this dive use consensus operating EPS; the trailing multiple is shown on both bases (11.47x operating, 10.57x GAAP diluted).
• Estimate coverage: 15 analysts on FY2026 EPS and 15 on FY2027, but only 7 on FY2028 EPS and 7 or fewer on every revenue line (7 for FY2026, 7 for FY2027, 5 for FY2028). Revenue coverage is thin throughout and the revenue basis is undetermined, which is why no revenue estimate is used for any conclusion. The FY2028 EPS row, which contributes to the fair-value cross-check, carries seven analysts and that thinness is disclosed.
• Peer note: the vendor peer set is unusable — Bank of America, BBVA, Interactive Brokers, ICICI Bank, Toronto-Dominion and UBS are banks and brokers, matched on the "Financial Services" sector code rather than on business. Only Chubb and Travelers are property-casualty insurers, and even those are commercial-lines-weighted rather than personal auto. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn.
• Fair-value caveat: the $150 / $215 / $270 anchors are price-to-book multiples applied to an estimated forward book value of roughly $67.50 per share — 2.2x, 3.2x and 4.0x respectively — cross-checked against the FY2027 consensus operating EPS of $16.204 (9.3x, 13.3x and 16.7x). Price-to-book is used as the primary method deliberately, because a trailing price-to-earnings anchor on an insurer at a cyclical margin peak is actively misleading. The base's 3.2x is derived from a normalised return on equity of roughly 22% against the current 35.4%, a 12% cost of equity and 8% long-run growth, discounted for the fact that normalisation has not yet occurred and its endpoint is unknown. Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 3.0x gives $203, 3.5x gives $236. The base case requires the forward multiple to EXPAND while earnings fall, and that is flagged in Section 6b as the fragile leg.
• Timing: second-quarter and June monthly results were released 2026-07-15, twenty days before this dive; the 10-Q was filed 2026-08-03, one day before. The next catalyst is the JULY monthly release due around 2026-08-19, roughly fifteen days out — the vendor's earn_cal next date of 2026-10-14 is the quarterly print and understates catalyst frequency by a factor of three, which is corrected in the front matter. The 2026-10-14 consensus of $3.62 operating EPS implies a 25% sequential decline from the June quarter's $4.85.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.