Customer Retention Strategies: 10 Levers, the Formulas and a Worked Example
Customer retention strategies are the steps you take so a customer who bought once buys again. You measure them with the retention rate: (customers at the end of a period − new customers in that period) ÷ customers at the start × 100.1 Keep a tactic only if the margin from the customers it keeps is larger than what it costs.
- Get new customers to their first result fast
- Ask why customers leave, then close the loop
- Reach out before customers lapse
- Make support fast and human, with self-serve options
- Personalize on what customers already did
- Make the second purchase easy
- Build a loyalty program people use, and price it first
- Recognize your best customers
- Give customers a place to meet
- Win back customers who lapsed
The 10 levers, ordered by the stage where customers leave:
To see what keeping customers longer is worth in dollars, run your own numbers in the unit economics calculator.
What customer retention is and why it pays
Customer retention is a business’s ability to keep its customers over a period, measured by the retention rate. A retained customer was a customer at the start of the period and is still buying at its end.
In 2014, Harvard Business Review put the cost of acquiring a new customer at 5 to 25 times the cost of retaining an existing one.2 In a BIA/Kelsey survey, 61% of small and midsize businesses said more than half of their revenue comes from repeat customers.3 Salesforce found that 88% of customers rate a company’s experience as important as its products or services.4
SAP’s Customer Loyalty Index 2025 surveyed 10,006 consumers in five markets and found True Loyalty at 29%, down five points from the year before; 28% had switched brands because they lost interest, with nothing going wrong.5 Watch each customer’s reorder cadence, not only complaints.
Retention moves numbers you already track:
- Repeat revenue. Adobe found that returning and repeat purchasers, 8% of site visitors, generated 41% of online revenue.6
- CLV and LTV/CAC. Each extra period a customer stays adds a period of margin to customer lifetime value (CLV), while you paid the acquisition cost once.
Acquisition and retention compared:
- Relative cost. Acquisition costs 5 to 25 times as much, per a 2014 estimate.2
- Main channels. Acquisition: ads, search and referrals. Retention: email, SMS, support and the product itself.
Key takeaways
- Track the retention rate on one fixed period, with new customers taken out of the count.
- Find the stage with the biggest drop before you choose a tactic.
- Price every tactic against the gross margin of the customers it keeps.
- Prove it on a holdout group before you roll it out to everyone.
How to calculate customer retention rate
The standard formula:1
Customer retention rate = (E − N) ÷ S × 100, where × 100 turns the result into a percentage.
- S is the number of customers at the start of the period.
- N is the number of new customers acquired during the period.
- E is the number of customers at the end of the period.
Leave new customers in, and a quarter with heavy acquisition shows high retention even when most old customers left.
Example by our method (assumption figures): an online store starts the quarter with 1,200 customers, gains 300 new ones and ends with 1,140. Customers kept: 1,140 − 300 = 840, so retention is 840 ÷ 1,200 × 100 = 70%.
Churn comes from the same counts:
Churn rate = (S − (E − N)) ÷ S × 100.
In the same assumption example, 1,200 − 840 = 360 customers were lost, so churn is 360 ÷ 1,200 × 100 = 30%. Retention and churn must sum to 100%; if yours do not, new customers probably slipped into one count.
Pick the period by the purchase cycle
Use a month for frequent purchases such as coffee or meal kits, and for monthly software. Use a quarter or a year for infrequent ones such as furniture or annual contracts. A period shorter than the normal gap between orders counts people as lost who have not come back yet.
Without subscriptions, count a customer as active if they ordered within your usual purchase cycle, and keep that rule fixed.
Read it by cohort
Group customers by first-purchase month and track the share still buying in each later period. Compare newer cohorts with your own older ones, not with industry averages.
Example by our method (assumption figures), share of each cohort still buying:
| First-purchase month | Month 1 | Month 2 | Month 3 |
|---|---|---|---|
| January | 100% | 45% | 38% |
| February | 100% | 52% | 44% |
Read across a row to find the leak: both cohorts lose the most between month 1 and month 2, so fix the second order first. Read down a column to see progress: February holds better than January at every step.
Six metrics to track next to the retention rate
Use the same period for every metric. The example column uses the quarterly assumption figures from the worked example (an example by our method).
| Metric | Formula | What it tells you | Example (assumption) |
|---|---|---|---|
| Churn rate | (S − (E − N)) ÷ S × 100 | share of starting customers lost | 30% |
| Repeat customer rate | customers with 2+ orders ÷ customers who ordered × 100 | share who came back within the period | not modeled |
| Purchase frequency | orders ÷ unique customers | how often a customer buys | 1.5 a quarter |
| Average order value (AOV) | revenue ÷ orders | what one order brings in | $40 |
| CLV on gross margin | AOV × orders per period × gross margin × (1 ÷ churn per period) | margin one customer returns over their life | $90 |
| Net revenue retention | (starting recurring revenue + expansion − contraction − churned revenue) ÷ starting recurring revenue × 100 | whether existing accounts grow | subscriptions only |
Revenue includes the cost of goods, so compute CLV on gross margin. The 1 ÷ churn term assumes steady churn, while real cohorts lose more people early, so treat CLV as an estimate.
Our method: find the leak, then price the lever
- Measure. Compute retention and churn on the same counts for one period.
- Locate the leak. In the cohort table, find the stage with the biggest drop and work on it first.
- Pick the lever for that stage from the map below, one lever at a time.
- Price the lever. Extra customers kept = lift in points × S. Margin per customer per period = AOV × orders per period × gross margin. The lever pays in the period if its cost is less than extra customers kept × margin per customer per period.
- Value the lift over the customer’s life. CLV = margin per customer per period × (1 ÷ churn per period). LTV/CAC and payback go to the unit economics calculator.
- Test on a holdout. Give the lever to a random share of customers, hold out the rest, and keep it only if the retention gap covers its cost.
| Where the cohort curve drops | Levers to try first |
|---|---|
| First 30 days, first order to second | onboarding, exit reasons |
| Active life | proactive outreach, support, personalization, the second purchase, loyalty, top-customer perks, community |
| Lapse or renewal | proactive outreach, win-back |
10 customer retention strategies
Get new customers to their first result fast
Define the first result, the moment the customer gets what they paid for, such as the first good cup brewed. Tie it to the job customers hire you for and check that onboarding delivers your value proposition.
Count the days from purchase to that result, then remove one step: a setup field, a manual approval, a slow default shipping option.
The metric is the second-order rate within one purchase cycle.
Ask why customers leave, then close the loop
When a customer cancels or passes their usual reorder date, ask one free-text question: “What made you stop buying from us?” For a longer script, see our customer interview questions.
Code each answer into one reason (price, quality, delivery, found an alternative, no longer needed) and fix the top reason first; our guide to customer pain points covers ranking them. Then tell the customers who raised it what changed, by email or on a public product roadmap.
The metric is the share of exits citing the top reason, measured the period after the fix.
Reach out before customers lapse
Three signals show before most customers leave:
| Trigger | Action | Who sends it |
|---|---|---|
| Fewer orders or logins than the customer’s own usual cadence | a check-in or a reorder reminder | an automated email; the account owner for large accounts |
| A second support ticket on the same issue | a personal call or email that owns the fix | a senior agent or the support lead |
| A failed payment | a retry plus a notice with an update link | the billing system; a person after a second failure |
Send on the channel the customer chose. In SAP Emarsys research, 69% of consumers named email as their preferred way for brands to contact them, and SMS tied with direct mail at 43%.7
With enough history, a churn model can score the same signals earlier; see our guide to AI for business.
The metric is retention of flagged customers who got outreach against flagged customers held out.
Make support fast, human and self-serve
Support tests the experience that 88% of customers rate as important as a company’s products or services.4
Write help articles for your top five ticket reasons. Publish a response time you can meet, keep a human within two clicks, and let the customer’s history follow them across channels.
The metric is the repeat-contact rate on the same issue.
Personalize on what customers already did
Suggest the next item from the customer’s last one, time messages to their buying cadence and send them on their preferred channel.
In SAP Emarsys research, 28% of consumers switched brands because they lost interest,5 so offer a customer who has bought the same item five times a related one to try.
The metric is the repeat-purchase rate of the personalized group against a generic-message holdout.
Make the second purchase easy
Add one-click reorder to order history and confirmation emails, time reminders to each customer’s cadence, and offer a subscription customers can skip or pause.
The published evidence is old, so treat it as a direction, not a forecast. A Bain & Company/Mainspring survey of 2,116 online shoppers in December 1999 found that repeat apparel customers spent 67% more in months 31 to 36 of the relationship than in their first six months.8 In a 2012 Adobe study, returning and repeat purchasers, 8% of site visitors, generated 41% of online revenue.6
Check it on your data: compare the next-year spend of customers who reordered within one purchase cycle with those who did not.
The metric is the median days between the first and second order.
Build a loyalty program people use, and price it first
Choose the type by purchase frequency: points for frequent buyers, tiers or standing perks for infrequent ones.
In the SAP Emarsys index, the top loyalty expectations were better deals (55%), great service (53%) and points or cashback (50%).5 True Loyalty fell five points to 29% in 2025;5 the same source cites 32% from a 2026 survey without saying whether the method matched, so don’t read the gap as a trend.
Set the reward as a share of spend and check it against gross margin. In our example, a program pays 5% of spend at a 45% gross margin (both assumptions), so it gives away 5 ÷ 45, about 11% of the margin on every rewarded order. Test demand for a perk with a fake door test before you build it.
The metric is the purchase frequency of members against a holdout of eligible non-members.
Recognize your best customers
Select the segment by RFM: score each customer on recency, frequency and spend, and take the top group. Check it against your ideal customer profile, so you reward the customers you want more of.
Engagement pays back. Gallup reports that a fully engaged customer represents an average 23% premium in share of wallet, profitability, revenue and relationship growth over the average customer.9
Cap the perk budget at a set share of that segment’s margin. Prefer non-cash perks: early access, a direct line to a named person, a handwritten note.
The metric is retention of the top segment.
Give customers a place to meet
A community fits a product with a learning curve or a shared identity, such as software or specialty coffee. Members answer each other’s questions, a reason to stay that a discount does not give. Skip it for commodity purchases.
You need one place, one host who makes sure every post gets a reply, and a weekly ritual so members know when to come back.
Pick a place your customers already use, start with your top RFM segment, and open it to everyone once the host keeps up.
The metric is the retention gap between members and non-members of the same cohort.
Win back customers who lapsed
Define “lapsed” as a customer whose gap since the last order is twice their usual gap. A customer who reorders every month counts as lapsed after two months without an order.
Send one message with a reason to return, such as a new product or the fix for their exit reason. Offer a discount last, and only after pricing it with step 4 of the method.
The metric is the reactivation rate within one purchase cycle, with the margin on reactivated orders reported next to it.
Example: what a 5-point retention lift is worth
This is an example by our method; every figure is an assumption or arithmetic on assumptions. The business is a small online store selling specialty coffee beans, measured by quarter.
| Figure | Value |
|---|---|
| Customers at start · new · at end (assumption) | 1,200 · 300 · 1,140 |
| Retention · churn (from the assumption counts) | 70% · 30% |
| AOV · orders per quarter · gross margin (assumption) | $40 · 1.5 · 45% |
| Margin per customer per quarter (from assumptions) | $27 |
| CAC (assumption) | $60 |
In our example (assumption figures), margin per customer is $40 × 1.5 × 0.45 = $27 a quarter. Lifetime is 1 ÷ 0.30 = 3.33 quarters, or 10 months, and CLV is $27 × 3.33 = $90. With an assumed CAC of $60, LTV/CAC is 1.5 and payback is $60 ÷ $27 = 2.2 quarters.
Now assume, for our example, that a lever lifts retention to 75%. Churn falls to 25%, lifetime rises to 4 quarters (12 months), and CLV rises to $108, up $18 per customer. LTV/CAC rises to 1.8.
Price the lever per quarter: extra customers kept are 5% × 1,200 = 60, and extra margin is 60 × $27 = $1,620 a quarter (assumption figures). That is the ceiling on what the lever may cost.
- A reminder flow that costs $300 a quarter (assumption) passes, since $300 is below $1,620.
- A 10% discount on every repeat order (assumption) costs $4 per order × 900 kept customers × 1.5 orders = $5,400 a quarter. That exceeds the $1,620 the whole lift returns, so it fails even if the discount caused the entire lift.
In the calculator, enter the figures from our example (price $40, purchases a month 0.5, margin 45%, CAC $60 and customer life 10 months), and it returns LTV $90; at 12 months it returns $108. Run your own numbers with your counts and margin.
Five retention mistakes that cost margin
| Mistake | Fix |
|---|---|
| A blanket discount for every repeat order | price it against the margin it keeps; in the example above it costs $5,400 against a $1,620 ceiling (assumption figures) |
| Judging your rate by an industry benchmark | compare with your own past periods and cohorts |
| CLV computed on revenue | multiply by gross margin before you divide by churn |
| No holdout group | keep a random share of customers out of each lever and compare |
| New customers mixed into the retention count | subtract new customers (N) before dividing by the starting count |
FAQ
What is a good customer retention rate?
There is no single number: benchmarks differ by industry, period length and whether new customers were excluded.
Your rate should beat your own trailing average, and CLV on gross margin should cover CAC with room to spare. In our example (assumption figures, example by our method), 70% quarterly retention gives LTV/CAC of 1.5, and 75% gives 1.8.
How do you calculate customer retention rate?
Subtract the period’s new customers from the customers at its end, divide by the customers at its start and multiply by 100.1 With our assumption figures, 1,140 minus 300 new gives 840, and 840 ÷ 1,200 = 70%.
What is the difference between retention rate and churn rate?
Retention is the share of starting customers you kept; churn is the share you lost. On the same counts they sum to 100%, so 70% retention means 30% churn.
Is keeping a customer cheaper than finding a new one?
Harvard Business Review put acquiring a new customer at 5 to 25 times the cost of retaining an existing one.2 Your own ratio depends on your CAC and your retention costs.
Do loyalty programs improve retention?
They can, if customers value the reward and it fits your margin. In SAP Emarsys research, deals, service and points topped consumers’ loyalty expectations.5 Price the reward against gross margin and test it on a holdout.
How can AI help reduce churn?
A churn model scores signals you already collect, such as falling orders, repeat tickets and failed payments, and flags at-risk customers earlier. Test the outreach on a holdout.
Which retention strategy should a small business try first?
Build a cohort table and find the biggest drop. If customers leave after one order, fix onboarding and ask why they left; if they leave at renewal, start with proactive outreach and win-back.
Sources
- revechat.com, 10/22/2026
- stylograph.ai, 10/22/2026
- mediapost.com, 10/22/2026
- salesforce.com, 10/22/2026
- emarsys.com, 10/22/2026
- adobe-newsroom.de, 10/22/2026
- emarsys.com, 10/22/2026
- media.bain.com, 10/22/2026
- news.gallup.com, 10/22/2026
Check your idea for free
Describe your idea and the check collects demand, competitors, complaints and prices, with sources.
