Ten numbers that tell you whether an embedded insurance programme is a business — and five that look impressive in a board deck while telling you almost nothing.
Embedded insurance programmes generate a lot of numbers, and most dashboards we see report the wrong ones. The test for a real KPI is simple: if the number moves, does someone know what decision to change? Ten metrics pass that test for a typical embedded programme. Five commonly-reported ones do not, and they are worth calling out because they actively mislead.
This is a practitioner's framework, not a benchmark study — we are deliberately not quoting target values here, because sensible targets differ by product line, market and journey design. What follows is which numbers to compute, how to define them so they cannot flatter you, and what each one is for.
Funnel KPIs: is the offer working?
- 1. Eligible-transaction coverage. The share of your transactions where an offer could be shown — the customer and product were in scope, and the quote service returned a price. This is your denominator integrity metric: attach rate computed on a shrunken denominator is fiction. If coverage is low, the growth lever is eligibility and quote reliability, not conversion tweaks.
- 2. Offer view rate. Of eligible transactions, how many customers actually saw the offer — not how many sessions technically included it below the fold. Instrument it as a genuine viewport event. A beautiful offer nobody scrolls to is a placement problem that no copy change will fix.
- 3. Attach rate. Policies bound divided by eligible transactions. The headline metric of the model, and the one to defend most fiercely against definition drift: keep the denominator at eligible transactions, count only bound-and-paid policies, and report it per product line rather than blended.
- 4. Quote-to-bind completion. Of customers who started the insurance flow, how many finished. This isolates friction inside your journey from the appeal of the offer itself. A high view rate with weak completion points at form length, pre-fill gaps or payment failures — all fixable in product.
Economic KPIs: is it a business?
- 5. Platform revenue per eligible transaction. Your commission earned divided by eligible transactions. This is the metric that lets you compare insurance against every other monetisation idea competing for the same screen space, in the same units. It rolls attach rate, premium mix and commission rate into one honest number.
- 6. Net revenue after givebacks. Commission net of cancellations, cooling-off refunds and clawbacks. The gap between gross and net commission is a quality signal: a wide gap means you are converting customers who regret the purchase within days.
- 7. Renewal rate on the eligible base. For annual products, policies renewed divided by policies that reached expiry. Long-run programme value is mostly this number; a programme reporting only first-sale conversion is reporting the cheap half of the story.
Quality KPIs: will it still be a business in three years?
- 8. Cancellation and complaint rate. Early cancellations and formal complaints per thousand policies. Regulators read these numbers, and so should you — they are the earliest warning that the offer is being bought by mistake or mis-sold by design.
- 9. Claims experience signal. Whatever you can lawfully observe of claim frequency, approval rate and settlement time on your book. The platform rarely runs claims, but the customer holds the platform's brand responsible anyway. Agree data-sharing on claims outcomes with your insurer or infrastructure layer at contract time, or you will fly blind on the metric that decides renewal.
- 10. Offer-experience impact. The effect of the insurance step on the host checkout: added time, drop-off delta against a holdout. Insurance revenue that costs you core-transaction conversion can be a net loss — this metric is the guardrail that keeps the programme honest with the rest of the company.
Five vanity metrics
- Gross written premium. GWP is the insurer's scale metric, not yours — your revenue is the commission line. Platforms headline GWP because it is a bigger number; that is exactly the problem.
- Total quotes generated. Quotes are an API call, not an outcome. Quote volume grows whenever traffic grows and says nothing about whether anyone wanted the product.
- Cumulative policies since launch. A monotonically increasing number that cannot go down is a chart decoration. Report the active in-force book instead — it can shrink, which is why it is informative.
- Products on the shelf. Shipping a seventh product line is only news if the first six attach. Breadth is an input; report it as roadmap, not performance.
- Announced partnerships. Signed logos measure business development effort. Until volume flows, a partnership is a press release with a KPI ambition.
Making the ten real
Three closing disciplines. Write the definitions down — denominator, numerator, counting window — before the first dashboard ships, because every later argument about performance is secretly an argument about definitions. Segment by product line and journey from day one; blended numbers hide everything interesting. And review the vanity list annually: every organisation regrows them, usually in the run-up to a board meeting.