One number tells you whether your insurance line is a product or a footnote: the share of eligible transactions that leave with cover. Defining it honestly is harder than it looks.
Attach rate is the fraction of eligible transactions that end with an insurance policy sold. That is the whole definition, and it is the single number that decides whether your embedded insurance programme is a business or a rounding error: your revenue is attach rate times eligible volume times revenue per policy, and of those three factors, attach rate is the only one your product decisions control directly. This piece is about measuring it honestly, decomposing it usefully, and knowing when a low number is actually fine.
We are deliberately not publishing benchmark figures here. Attach rates vary so much by product line, market, placement and price point that a context-free number is closer to misinformation than guidance; our separate benchmarks research carries the figures together with the methodology they need. What transfers across every programme is the logic below.
The denominator is where programmes lie to themselves
Everyone agrees on the numerator: policies sold. The denominator — eligible transactions — is where definitional choices quietly manufacture whatever number the slide deck needs.
Consider a car marketplace. Is the denominator all listings viewed? All purchases completed? Purchases by buyers who saw the insurance offer? Buyers who saw it, excluding those who already hold cover elsewhere? Each definition is defensible; each produces a different attach rate; and moving between them can triple the reported number without a single extra policy sold.
Our recommendation is to maintain two rates and never blend them. The commercial attach rate uses the widest honest denominator — every transaction where cover was relevant — and is the number for board decks and forecasts, because it reflects the revenue reality. The funnel attach rate uses offers actually seen as its denominator and is the number for product work, because it isolates what your UX controls from what your placement controls. When a programme reports one number and cannot tell you which of these it is, it usually is not either.
Why this metric, and not revenue
Revenue flatters. A programme can post growing insurance revenue purely because the platform's transaction volume grew, while the offer itself converts worse every quarter. Attach rate is volume-neutral: it measures whether the offer earns its place in the journey. A falling attach rate inside a growing platform is an early warning that revenue will follow — typically two or three quarters later, once volume growth stops covering for it.
Revenue tells you what the programme earned. Attach rate tells you whether it deserves to keep its place in the checkout.
The five levers
Every attach-rate movement we have seen decomposes into five levers. Diagnose before you optimise: each lever fails differently.
- Relevance: does the transaction genuinely imply the insurable need? Cover for a need the customer does not recognise at that moment converts near zero regardless of execution. This lever is set when you choose what to embed, and no downstream work compensates for choosing wrong.
- Placement: where in the journey the offer appears. Before payment intent forms, it is an interruption; after the transaction completes, it is an afterthought. The window is narrow and product-specific.
- Price salience: not the price itself, but how it reads against the anchor transaction. The same premium reads differently attached to a purchase a hundred times its size than to one three times its size.
- Friction: every additional field, screen and decision. Pre-filled data from the transaction is the structural advantage of embedded distribution; a flow that asks questions the platform already knows the answers to is spending its advantage.
- Trust: brand context, clarity of what is covered, and visible ease of claiming. Slowest lever to move, and the one that compounds.
When a low attach rate is fine
Attach rate is a means, not a scoreboard. Three situations where a modest rate is the correct outcome:
- High-premium, high-consideration products. A small share of transactions attaching a substantial annual policy can outearn a mass-attach microproduct many times over. Judge these programmes on revenue per eligible transaction instead.
- Deliberately quiet placement. Some platforms rationally choose a low-pressure offer to protect their core conversion. The attach rate is lower by design; the trade was made consciously and should be evaluated as a whole.
- Early honesty. A new programme measured against the wide commercial denominator starts low. That is the accurate baseline, not a failure — the failure is narrowing the denominator to feel better.
The inverse also holds: a very high attach rate is not automatically good news. If it is achieved through pre-selection, dark patterns or customers not understanding what they bought, it converts into cancellations, complaints and regulatory attention on a delay. Attach rate only means something alongside its quality shadow: early cancellation rate, claims acceptance experience, and complaint volume.
Making it operational
Three practices separate programmes that manage attach rate from programmes that merely report it. First, write the denominator definition down, version it, and flag every dashboard with which version it uses — future you will try to move the goalposts. Second, decompose weekly: offer visibility, offer engagement, flow completion. A single blended number hides which lever moved. Third, review attach rate next to its quality shadow in the same meeting, so the incentive to inflate one at the cost of the other never takes root.
Embedded insurance succeeds transaction by transaction, and attach rate is the count of those small verdicts. Measure it honestly and it will tell you, earlier than any other number, whether you have a business.