Commercial auto's 104% combined ratio: the submission data problem underwriters miss

The commercial auto submission data problem nobody talks about

Commercial auto just recorded its 14th consecutive year of underwriting losses. The line posted a $4.9 billion loss in 2024, and S&P Global projects combined ratios climbing from 104.4 in 2026 to 106.3 by 2029. Every carrier running commercial auto knows the numbers are bad. What most underwriting teams haven't reckoned with is how much of the problem starts at submission intake.

A $4.9 billion problem with a data quality root cause

The insurance industry's standard explanation for commercial auto losses focuses on social inflation and nuclear verdicts. That explanation is correct but incomplete. Yes, 135 lawsuits against corporate defendants resulted in nuclear verdicts exceeding $10 million in 2024, a 52% increase over 2023, with total verdict value reaching $31.3 billion. Yes, average claim severity for commercial auto liability has more than doubled over nine years. These are structural forces that no single carrier controls.

But here is what carriers do control: the quality and speed of the data that feeds underwriting decisions. And in commercial auto, that data arrives in some of the most complex, fragmented formats in all of commercial P&C.

A typical commercial auto submission includes vehicle schedules (sometimes hundreds of units across multiple states), driver lists with MVR histories, loss runs from prior carriers in inconsistent formats, ACORD applications with fleet-specific endorsements, and supplemental schedules for hired and non-owned auto. Each document type arrives in a different format depending on the broker, the prior carrier, and the state. A fleet account with operations in Texas, Florida, and Georgia — three jurisdictions with among the highest nuclear verdict frequency, requires underwriters to process and synthesize data from multiple sources before they even begin risk evaluation.

When that data extraction happens manually, errors compound. A miskeyed vehicle class code changes the rate. A missed driver with a poor MVR history goes unpriced. A loss run transcription error understates prior severity. None of these mistakes are dramatic on their own. Across a book of 500 commercial auto accounts, they are the difference between a 104% combined ratio and a 98%.

Why casualty rate increases don't fix a data problem

US commercial insurance rates moderated to 2.9% overall in Q4 2025, according to WTW's Commercial Lines Insurance Pricing Survey. But casualty lines tell a different story: rates rose 7%, and 11% when excluding workers' compensation. Commercial auto pricing has been aggressive for years. The industry has been pushing rate, and the combined ratio still sits above 100.

Rate alone doesn't fix a structural problem. If your underwriting team is pricing a 50-vehicle fleet in a high-litigation state and the submission data contains a transcription error that misclassifies three heavy trucks as light vehicles, the rate increase you applied is mathematically irrelevant. You priced the wrong risk.

AM Best projects that commercial auto remains under-reserved by $4 billion to $5 billion industry-wide. That gap exists because historical loss data, the foundation of reserving models, is itself subject to the same data quality challenges as new business submissions. Stale loss runs, inconsistent formatting across carriers, and manual transcription from PDF to rating system introduce noise at every step.

The carriers managing commercial auto profitably aren't the ones with the highest rates. They're the ones with the cleanest data pipelines, where every vehicle schedule, loss run, and driver record flows into the underwriting workstation accurately, consistently, and fast enough that the underwriter spends time on risk analysis instead of data assembly.

The three submission data failures that cost commercial auto underwriters the most

1. Vehicle schedule extraction errors

Commercial auto vehicle schedules are uniquely difficult documents. A mid-size fleet account might submit a 200-line Excel spreadsheet with VINs, vehicle years, makes, models, class codes, garaging locations, and radius of operation. Formats vary wildly by broker. Some arrive as PDFs of scanned printouts. Others come as CSVs exported from fleet management systems with non-standard column headers.

Manual extraction of a 200-vehicle schedule takes an experienced underwriting assistant 45-60 minutes. Error rates on manual vehicle schedule entry run 2-5% per field, according to operational data from carriers Pibit.ai works with. On a 200-vehicle schedule with 8 fields per vehicle, that is 32 to 80 field-level errors per submission. Each error has pricing implications: wrong class codes, incorrect garaging states, missed radius-of-operation flags.

Template-agnostic extraction handles the format variation that breaks template-dependent OCR. The vehicle schedule that arrives as a scanned PDF from one broker and a Google Sheets export from another both produce the same structured output, every VIN, class code, and garaging location captured accurately regardless of source format.

2. Loss run reconciliation gaps

Commercial auto loss runs arrive in carrier-specific formats. A fleet account with five years of history across two prior carriers produces loss runs with different column structures, different claim status terminology, and different methods of reporting incurred versus paid losses. The underwriter needs to reconcile these into a single loss history to evaluate trend, frequency, and severity.

When done manually, underwriters often default to the broker's summary rather than reconciling the raw loss runs. The broker's summary is an interpretation of the data, not the data itself. It may omit open claims, round severity figures, or exclude ALAE. In a line where social inflation is driving 8% annual severity increases, the difference between the broker's summary and the actual loss run data is the difference between pricing for last year's risk and pricing for next year's.

Automated loss run processing normalizes disparate carrier formats into a consistent structure, surfaces open claims the broker summary may have omitted, and calculates trend lines that reveal whether a fleet's loss experience is improving or deteriorating, before the underwriter even opens the file.

3. Incomplete driver and fleet data

Driver records, MVR histories, and DOT compliance data are critical to commercial auto risk assessment. A fleet with three drivers holding recent DUI convictions presents a fundamentally different risk than one with clean records across the board. But this data often arrives in supplemental attachments, separate from the primary application, in formats ranging from state DMV printouts to carrier-generated driver experience reports.

When submission volume exceeds underwriter capacity, supplemental documents are the first casualty. They get deprioritized, skimmed, or missed entirely. In a market where a single nuclear verdict from one driver's accident produces a $10 million+ loss, incomplete driver data is not an administrative oversight. It is an underwriting failure with direct loss ratio consequences.

What this means for underwriting operations in 2026

The commercial auto market is not going to fix itself through rate increases alone. S&P Global's forecast of combined ratios rising through 2029 reflects a structural challenge, not a cyclical pricing correction. Social inflation is behavioral and legal, it responds to jury attitudes, litigation funding economics, and settlement anchor effects. These forces operate independently of carrier pricing actions.

What carriers control is the quality of the risk data that flows into their pricing models. In a line where every percentage point of combined ratio represents millions in underwriting margin, the operational question is practical: how accurately and quickly does submission data convert into decision-ready information?

Carriers still processing commercial auto submissions manually face a compounding problem. As rate adequacy erodes and claim severity accelerates, the margin for data error shrinks. A 3% field-level error rate on vehicle schedules was tolerable when combined ratios sat at 99%. At 104%, that same error rate is the difference between writing profitable business and subsidizing losses.

The underwriting teams managing commercial auto most effectively in 2026 share a common operational characteristic: they've removed manual data extraction from the submission workflow. Their underwriters receive structured, verified data, vehicle schedules, loss run summaries, driver records, fleet compliance data, and spend their time on risk selection, not data assembly. They respond to brokers faster because quote turnaround isn't bottlenecked by transcription. And they price more accurately because the data feeding their models reflects what the submission actually contains, not what a hurried assistant transcribed under time pressure.

Streamlining submission processing is not an efficiency initiative in commercial auto. It is a loss ratio initiative. When your combined ratio is above 100%, every improvement in data accuracy translates directly into better risk selection and tighter pricing. The math is straightforward: 85% faster submission processing means underwriters evaluate more risks per day. 100% data accuracy means the risks they select are priced correctly. Together, these produce the 700 basis points of loss ratio improvement that separates carriers growing profitably from carriers growing into bigger losses.

The commercial auto crisis will not be solved by any single technology, pricing strategy, or regulatory change. But the carriers that emerge from this cycle in the strongest position will be the ones who treated submission data quality as an underwriting discipline, not an administrative afterthought.