Insurtech has spent the last decade proving that the insurance industry could be rebuilt with better technology. What the sector is working through now is a different problem: proving that it can be run. Loss ratios that looked acceptable during hypergrowth have become liabilities as reinsurance costs rise and capital markets demand a clearer path to underwriting profitability. CAC that was forgiven when premium volumes were doubling now needs to be justified against cohort LTV. Claims operations that were scaled on headcount need to convert into data systems.
The companies navigating this transition well are not necessarily the ones with the most sophisticated models. They are the ones that have built operating intelligence — a structured, real-time view of what is making money, what is leaking margin, and what to do about it. This post lays out the metrics framework required to operate that way.
The Core Problem: Data Lives in Four Different Systems
Most insurtechs have at least four distinct data environments: a policy administration system (PAS), a claims management system, a distribution or producer management platform, and a financial or accounting system. These were built for different functions. They rarely share a common identifier. And they are almost never read together in real time.
The practical consequence is that a COO trying to understand whether loss ratio deterioration in a particular product line is driven by adverse selection, a specific distribution channel, or a claims handling issue has to manually cross-reference three exports and wait for a finance report. By the time that answer exists, the quarter is over and the damage is done.
Operating intelligence for insurtech means solving this integration problem first. The metrics framework below is only useful if the underlying data is connected. Defining what to measure is step one. Building the data layer that makes measurement continuous is step two — and for most insurtechs, it is the harder step.
The Underwriting Metrics Layer
Underwriting performance is the foundation of the insurtech operating model. Everything else — expense management, distribution strategy, reinsurance structure — is downstream of whether the core underwriting economics work.
Loss Ratio
The loss ratio is incurred claims divided by earned premium, expressed as a percentage. It is the single most important indicator of underwriting health.
Benchmarks: For P&C carriers and MGAs, a well-performing gross loss ratio sits in the 55–65% range. Early-stage books commonly run at 70–80% as pricing models are calibrated against real claims data. A gross loss ratio sustained above 85% over multiple quarters is a structural warning sign — it typically indicates that either pricing is inadequate, risk selection criteria are too loose, or a specific distribution channel is bringing adverse risks that aggregate data is masking.
Track loss ratio at multiple levels of granularity simultaneously: by product line, by cohort (policy inception quarter), by distribution channel, and by geography. A blended loss ratio that looks acceptable at the portfolio level can hide a severely deteriorating cohort or channel beneath it.
Combined Ratio
The combined ratio adds the expense ratio to the loss ratio. A combined ratio below 100% means underwriting is profitable before investment income. Above 100%, the company is paying out more in claims and expenses than it collects in premium.
Benchmarks: The US P&C industry closed 2025 with a combined ratio of 92.9%, its strongest underwriting performance in over a decade. For insurtechs specifically, the 95% combined ratio is the threshold investors use to assess whether underwriting discipline is established. Series B and later investors increasingly want to see a credible roadmap to sub-95% combined ratios within 24 months, even if the current number is above that level.
A combined ratio consistently above 110% for a carrier or MGA — without a specific and documented path to improvement — is difficult to finance and harder to reinsure.
Expense Ratio
The expense ratio captures operating expenses (including acquisition costs, technology, and G&A) as a percentage of earned premium. Traditional P&C carriers run expense ratios of 20–40%. Insurtechs frequently target sub-15% expense ratios by eliminating legacy infrastructure costs and automating underwriting workflows — but this advantage only materializes at scale.
Early-stage insurtechs often have high expense ratios due to fixed technology and compliance costs spread across a small premium base. The key metric to watch is the expense ratio trend as premium volume grows. If the expense ratio is not declining materially as premium doubles, the operating model has a structural efficiency problem.
Premium Per Policy
Average premium per policy tells you the average revenue contribution of each policy in force. It is most useful as a mix indicator — significant changes in premium per policy often signal shifts in the underlying risk profile or product mix before the loss ratio reflects them. Declining premium per policy alongside a stable loss ratio may indicate that the book is shifting toward lower-risk, lower-premium segments. Rising premium per policy with a stable loss ratio may indicate pricing power or an improving risk selection model.
The Claims Metrics Layer
Claims operations are where the loss ratio is ultimately determined. Technology investments in claims handling directly affect not just cost per claim but speed of settlement, fraud detection, and customer NPS.
Claims Frequency
Claims frequency is the number of claims filed as a percentage of policies in force during a given period. It is the primary volume driver of loss ratio and the first metric to watch for early deterioration signals.
Frequency benchmarks vary sharply by product line — auto claims frequency is measured in a completely different range than homeowners or renters — so the most useful comparison is frequency trend against prior periods for the same product and cohort, not against industry averages. A sudden increase in claims frequency on a specific product or geography cohort is the earliest quantitative signal that adverse selection has occurred or that an external loss event is developing.
Severity Per Claim
Average claim severity (total paid losses divided by number of claims) interacts with frequency to produce the loss ratio. Insurtechs with strong data on both can decompose loss ratio changes into frequency-driven and severity-driven components — a distinction that points to different operational responses. Rising frequency with stable severity suggests a distribution or underwriting issue. Stable frequency with rising severity suggests inflation exposure in the underlying coverage or a claims handling problem.
Cycle Time and Reopened Claims Rate
Claims cycle time — the average number of days from first notice of loss to claim closure — is both a cost efficiency metric and a customer experience metric. Faster cycle times reduce loss adjustment expenses and correlate with higher NPS scores. The reopened claims rate measures how often closed claims require additional payment, which indicates whether the initial settlement process is thorough.
The Distribution and Customer Metrics Layer
Customer Acquisition Cost by Channel
Insurance companies typically spend $487 to $900 per new personal lines customer acquired through standard digital channels, with more complex products reaching $1,200 or above due to underwriting requirements. These are blended averages. The more actionable number is CAC broken down by channel — paid search, direct, broker or agent, embedded or API distribution, and referral — because these cost structures differ by an order of magnitude.
An insurtech running blended CAC without channel visibility is almost certainly cross-subsidizing expensive paid acquisition channels with cheap direct or embedded channels. That cross-subsidy disappears the moment paid acquisition scales and the mix shifts.
LTV:CAC Ratio
A 3:1 LTV:CAC ratio is the minimum threshold for a sustainable insurance customer acquisition model. A 4:1 ratio represents genuine acquisition efficiency. Below 3:1, the company is consuming capital to acquire customers who do not generate enough lifetime premium to justify the cost.
LTV in insurance is calculated from net premium retained (after reinsurance cession) multiplied by expected policy duration, minus expected claims and expenses over that period. For most personal lines products, LTV is heavily front-loaded in the first renewal year — customers who renew once are significantly more profitable than first-year policies due to declining acquisition cost amortization.
Net Promoter Score
The insurance industry averages an NPS of 23–35. P&C carriers tend to score in the 35–73 range; life and health carriers typically score 14–40. For insurtechs, NPS matters most as a retention predictor rather than a brand metric.
Claims interactions are the primary driver of NPS movement in insurance. Research shows that customers who begin a claim as detractors can convert to promoters at a 30% rate when the claim is resolved with transparency and speed. This means claims handling quality is not just a loss ratio input — it is a renewal premium driver. Operators who treat NPS as a marketing metric rather than an operational one miss this connection.
The Reinsurance Economics Layer
Reinsurance decisions are capital structure decisions. They should be modeled with the same rigor applied to equity financing rounds.
Cession Rate and Net Retained Premium
Quota share reinsurance allows early-stage insurtechs to limit net loss exposure while scaling premium volume. The trade-off is that ceding a large share of premium compresses net margins and reduces the premium available to cover operating expenses. A carrier ceding 55% of gross premium retains only 45 cents on every dollar earned to cover claims, expenses, and margin.
As loss ratio credibility builds — typically after 12–24 months of consistent underwriting data — the economics of high cession rates become increasingly difficult to justify. Lemonade's reduction of its cession rate from approximately 55% to around 20% for new business is a public example of using demonstrated loss ratio improvement as the trigger to retain more premium. Operators should model the cession rate step-down explicitly, with loss ratio thresholds defined in advance as the criteria for reducing reinsurance dependency.
Gross vs. Net Loss Ratio
The difference between gross loss ratio (before reinsurance) and net loss ratio (after reinsurance) tells you how much risk is actually being transferred and at what cost. Lemonade's gross loss ratio fell to 67% in 2025, while the net loss ratio was 69% — a narrow gap indicating limited loss transfer benefit from reinsurance at that cession level. Early-stage operators with gross loss ratios above 80% and wide gross-to-net gaps are effectively paying reinsurers to manage underwriting problems that should be fixed at the pricing or selection level.
Building the Operating Intelligence Layer
The metrics described above do not exist in isolation. Loss ratio by cohort is only meaningful if it is connected to the distribution channel that sourced that cohort. CAC by channel is only actionable if it is connected to the LTV model that prices retention assumptions. Reinsurance cession decisions are only well-calibrated if the gross loss ratio trend is visible in near-real time.
Building operating intelligence for insurtech means creating a data layer that connects the PAS, claims system, distribution platform, and financial system into a single operating view. The architecture does not need to be complex. It needs to be fast enough that a COO can see a loss ratio moving before the quarter closes — not after.
The operational questions that matter in insurtech are not complex in concept. Which product lines are above the combined ratio threshold? Which distribution channels are producing adverse cohorts? Which claims handling workflows are driving reopened claims? The difficulty is not the analysis. The difficulty is having the data connected, current, and organized in a way that makes the answers immediate rather than historical.
Operating intelligence does not replace underwriting judgment. It eliminates the lag between what is happening in the business and what operators actually know.