Retention Benchmarks

What published retention research can and cannot tell you

Bain: the economics of keeping customers

Fred Reichheld’s 2001 Loyalty Rules! chapter reports that a 5% increase in retention can increase profits by 25% to 95%. It is a historical finding about customer economics, not a universal forecast for a salon, studio or software buyer. Costs, margins, customer tenure and the starting retention rate all matter. Read the original Bain-hosted chapter.

McKinsey: customers expect relevant interactions

The 2021 Next in Personalization report found that 71% of surveyed consumers expected personalized interactions and 76% felt frustrated without them. The September 2021 survey included 1,013 US adults. It measures expectations, not the causal impact of a retention product. Read the report.

Neither source evaluated RetentionTek. Neither establishes that one vendor is the only option or guarantees your return on investment. Use them to motivate a measured retention program, then validate your own results.

Build a retention report that supports a business decision

For each reporting period, show the cohort definition, service or industry context, eligible customer count, completed returns, observation window, missing data and formula. Separate new customers from established customers. Report percentage-point changes alongside the underlying counts.

• Appointment businesses: first-to-second visits and completed returns within an agreed service window.
• Retail: repeat purchase cohorts; exclude cancelled and refunded orders consistently.
• Subscriptions: customer renewals and revenue retention are separate measures; expansion revenue can mask customer losses.
• Outreach: eligible contacts, messages delivered, opt-outs and completed returns. Delivery is not a revenue outcome.

Illustrative, not observed: 42 returns out of 60 eligible customers is 70%. A later comparable cohort with 48 of 60 returns is 80%, an increase of 10 percentage points. It does not by itself prove that outreach caused the difference. Check seasonality, customer mix, capacity and a comparison group.

Where transaction data is unavailable, leave revenue impact unmeasured. For workflow examples, use the industry outreach playbook; for a product evaluation, review retention software versus a CRM.

Why this page publishes no industry averages

Any “average med spa retention rate is X%” figure you meet online deserves interrogation it rarely survives: whose clients, which definition, what period, what denominator, what sample size, self-reported by whom, selling what? Retention definitions vary so widely, flat-day rules vs cadence windows, blended vs cohort, visits vs revenue: that two businesses reporting “80%” may be measuring different phenomena. Publishing a number here would launder all of that variance into false authority. Instead, this page teaches benchmark method: the comparisons that survive scrutiny, with your own data, under definitions you control.

Internal benchmarks: your baseline is the best benchmark

Historical self-comparison. Your own past cohorts under fixed definitions are the only perfectly comparable population in existence, same services, same market, same rules. Trend your retention rate, rebooking rate, and first-to-second conversion (framework) quarterly with frozen definitions, and judge every initiative against that line.

Why it beats external averages. External numbers differ in definition, mix, and market, three uncontrolled variables against your one question (“are we improving?”). Your baseline answers it directly, immediately, and honestly. A business beating a dubious industry average while its own trend declines is losing; a business trailing one while its own trend climbs is winning. Trend first, peers a distant second.

Cohort benchmarks

Compare equal-age cohorts: January joiners at 6 months vs April joiners at 6 months. Cohort benchmarks isolate operational change from mix change: the March process fix shows up in March-and-later cohorts, not in a blended number that dilutes it across two years of history. Rules: same eligibility, same window, same maturity minimums (RetentionTek requires 30 mature customers per score component before reporting: the same discipline applies to any benchmark you take seriously). Small cohorts get wide mental error bars; never benchmark a cohort of nine.

Location and practitioner benchmarks

Multi-location or multi-practitioner businesses hold a natural experiment: same brand, different execution. Benchmark locations and practitioners against each other, fairly, with case-mix awareness (a location doing mostly consultations will convert differently than one doing maintenance series). Uses: spread best practices from the top, coach the middle, investigate the bottom. Misuses: ranking as punishment, which teaches locations to game denominators instead of serving clients. Benchmarks are diagnostic instruments; point them at processes, never as weapons at people, and say so out loud.

First-to-second and established-return benchmarks

Benchmark the funnel stages separately, because they answer different questions. First-to-second benchmarks judge acquisition quality and checkout operations; established-return benchmarks judge base rhythm and follow-through. A business can lead on one and lag on the other: the composite hides this, which is why RetentionTek's score keeps the 40/60 components visible rather than merging them into an uninterpretable whole. Set separate targets, review them in separate agenda items, fix them with separate playbooks.

Trend benchmarks and the lag rule

Trend benchmarks compare the business against its own trajectory: is the slope right? Respect the cadence-cycle lag, process changes need a full return window to register, so quarterly judgment on quarterly-cadence businesses is the fastest honest cadence. Faster judgments reward noise-chasing: the checkout script changed Tuesday cannot have moved quarterly retention by Friday, and anyone claiming otherwise is reading variance as signal. Smooth with rolling windows, annotate interventions on the timeline, and let the slope speak.

The comparability checklist

Before accepting any benchmark, yours, a peer's, or a vendor's, verify:

Same definition. Identical numerator, denominator, eligibility, and lapse rules. “Retention” is not one metric.

Same window. Equal periods and equal cohort ages. A 90-day rate and a 12-month rate are different instruments.

Adequate sample. Mature-customer minimums (tens, not single digits) per compared group. Small samples produce confident-looking noise.

Mature cohorts only. Unexpired windows excluded from both sides. Counting the unmatured punishes the recent.

Comparable mix. Similar services, client tenure, and acquisition channels, or segmented until they are.

Known provenance. Who measured, with what incentive, selling what alongside? Self-reported vendor benchmarks are marketing with error bars removed.

Any benchmark failing two or more of these is entertainment. Treat it accordingly.

Frequently asked questions