Policy vs premium retention — the number that lies.

Underneath every retention spreadsheet is one question: are clients leaving? New business is nice and revenue growth is nice, but a book bleeding clients faster than sales can replace them is a treadmill, and treadmills don't compound. Retention rate is supposed to answer the question — except the number on most reports doesn't measure clients. It measures dollars. This piece separates the two metrics, shows where the deception hides, and lays out what to request to get a defensible answer.

§ 01 · Two metrics, one nameDollars renewed vs clients renewed.

Premium retention is the share of premium dollars that renewed year over year — calculated on revenue. That feels right, because revenue is what you're buying. It isn't right, because premium per policy moves with carrier rates. Policy retention — also called unit or client retention — is the share of clients who renewed, unaffected by rate. The same book in the same year produces two numbers that can tell diametrically opposite stories, and only the policy number answers "are clients leaving?"

§ 02 · The hard-market maskHow rising premiums hide departing clients.

When carriers push rate to offset losses, premium per policy climbs — a $2,200 home policy renews at $2,640. If every client renews at the new rate, premium retention soars past 100%; in a hard market, agencies routinely report 105%, 108%, even 115%. Those numbers look fantastic and hide something that isn't. Take an agency that began the year with 1,000 clients at 10% rate inflation: 850 renew, 150 shopped and left. On policy count that's 85% — a leaky-bucket signal — but because the 850 survivors each pay 10% more, premium retention lands around 94%.

§ 03 · Calculating it rightPolicy count, by line, over 36 months.

The formula is plain: policy retention equals policies renewed divided by policies available to renew. Four specifics matter. Count policies, not premium. Don't put new business in the denominator — it isn't "available to renew" yet, and including it inflates the denominator and deflates the rate. Pick one definition of "renewed" — buyers want client retention (the client stayed, even if the carrier changed), not carrier retention. And run it by line of business, because personal and commercial retain differently and a blended number hides line-specific problems. Every modern management system (Applied Epic, AMS360, QQCatalyst, Hawksoft) produces a native policy-count report; if the agency isn't running one, that absence is itself a soft signal.

The worked composite makes the stakes concrete: start the year at 1,200 policies, 1,020 renew, average premium up 9%. Policy retention is 1,020 ÷ 1,200 = 85% — the leaky-bucket threshold. Premium retention is (1,020 × 1.09) ÷ 1,200 = 92.7% — comfortably "normal." Same book, same year. A buyer who asks only for the second number overpays; a seller who reports only the first undersells.

§ 04 · The benchmarksWhat to underwrite to, and what to request.

Once you're measuring on policy count, the benchmarks are well established.

A single year isn't enough — ask for a 36-month trendline, sliced by line of business, with year-over-year comparisons; a flat 88% across three years beats a one-year spike to 93% off an 83% prior year. Your diligence request should be explicit: the 36-month policy-count report, a companion premium-retention report (not because you'll trust it alone, but because the gap reveals the rate inflation), a written methodology for how carrier changes and rewrites are counted, and any footnotes for book acquisitions or carrier exits. Treat the gap between the two metrics as its own diagnostic — a 10-point spread tells you rate is doing the work, so underwrite to the policy number. And if the seller resists producing policy-count retention at all, that refusal is a data point: every system can produce it, so reluctance usually means the number isn't flattering.

◆