Customer Retention Rate for a Small Business

The formula you will find everywhere, retention rate = ((customers at the end minus new customers) ÷ customers at the start) × 100, assumes your customers are on a list that they can leave. A landscaper, a dentist and a dog groomer have no such list. Their customers never cancel anything. They just stop coming, and there is no event to count. The small-business version replaces the missing cancellation with a deadline you set yourself. Pick a repeat window that matches how often a happy customer would normally come back, usually your typical service interval times about 1.5. Take every distinct customer you served in one fully aged month. Count how many of them came back inside the window. Divide, multiply by 100, and that is your retention rate. For a groomer on a six-week cycle the window is about four months. For a dentist on a six-month recall it is twelve to eighteen. The number takes five minutes once your records carry a customer name and a date, and the list of people who did not come back is worth more than the percentage.

Why doesn't the standard retention formula fit a local business?

The formula was built for a subscriber base. Software, gyms, phone contracts, insurance. In that world a customer is a row in a table with a status, and when they leave they do something: they cancel, the card fails, the renewal date passes. Both counts in the formula are unambiguous, because on any given day the business can say exactly how many live accounts it has.

Now try it on a mobile dog groomer. How many customers did she have on 1 March? The question has no answer. There is no status field. There is a woman with a labradoodle who came four times last year and has not been in since November, and nobody, including the customer, has decided whether that relationship is over. She has not churned. She has just not booked.

That is the whole problem. The metric depends on an event that never happens in appointment-based and job-based work. Plugging a guess into the formula gives you a number that looks like a retention rate and is not one, which is worse than having no number at all, because you will make decisions with it.

There is a second mismatch that matters just as much. In a subscription business the period is arbitrary. Monthly, quarterly, annual, it does not matter, because the cancel event lands wherever it lands. In a service business the period has to match the rhythm of the work. Measure a landscaper's retention over one month and almost nobody has come back, because almost nobody was due. Measure a barber's retention over a year and almost everybody has, because a year is eight visits. The same business scores 12% or 94% depending on a number somebody picked without thinking about it.

So we fix both things at once. Replace the cancellation event with a deadline. Set the deadline from your own service interval instead of the calendar.

What is a repeat window and how do you set yours?

A repeat window is the amount of time after a visit within which a customer who intends to stay with you would reasonably have come back. Past that line, you treat them as lapsed. It is a judgement call, and it is your judgement, not an industry constant.

Start with your normal service interval, then add roughly half again. The padding absorbs real life: a holiday, a busy quarter, a rescheduled appointment. Some rough shapes:

Business	Typical interval	Sensible repeat window
Barber, nail tech	3 to 5 weeks	2 to 3 months
Hair salon, dog groomer	6 to 8 weeks	3 to 4 months
House cleaner (fortnightly)	2 weeks	6 to 8 weeks
Mobile detailer	3 to 6 months	9 months
Dentist, hygienist	6 to 12 months	12 to 18 months
Landscaper (seasonal)	one season	the following season
HVAC service plan	annual	15 to 18 months

Do not take those as authority. Take them as a starting point you will correct after one look at your own book, which is the better source anyway. Sort your existing customers by the gap between their visits and read the middle of the distribution. If most of your returning clients come back between five and nine weeks, your window is three months and you knew it before you read this table.

It is worth saying plainly that even clinical service intervals are less fixed than they sound. The American Dental Association's review of the evidence on how often patients should be seen notes that systematic reviews have not reached consensus on an optimal recall interval, and that the sensible approach is to tailor the interval to the individual patient's assessed risk. If the profession with the most-studied service interval in the world sets it per patient, you are entitled to set your own window from your own data rather than looking for a rule.

One rule about the window that is not negotiable: once you pick it, leave it alone. A retention rate is only meaningful against its own history, and changing the window changes the number without anything changing in the business.

Which retention question can your records actually answer?

Before you calculate anything, find out which version of the question your records support. There are three, they are not interchangeable, and most owners try to answer the one their records cannot reach.

The good news is that you almost certainly have more raw material than you think. The IRS is explicit that, for federal tax purposes, the law does not require any specific kind of records and that you can use any system suited to your business that clearly shows income and expenses. Everyone in business is keeping something. The question is only what shape it is in.

Work down this tree and stop at your first yes.

1. Do you have agreements with a start and an end date? Contracts, retainers, memberships, service plans, recurring cleaning schedules.

→ Yes. You can answer the textbook question. Count the agreements live on the first day of the period and the ones still live on the last day, subtract agreements signed during the period, and the standard formula works exactly as published. Use it. You have a subscriber base even if you never called it that.

→ No. Go to 2.

2. Do you keep dated visits or jobs attached to a named customer? An appointment book, a job list, an invoice history, a calendar with names on it.

→ Yes. You can answer the repeat-window question, which is the real retention rate for a service business and the one this article is about. Method in the next section.

→ No. Go to 3.

3. Do your receipts identify the customer at all? A name on an invoice, an email on a digital receipt, a loyalty tag, a phone number in the booking note. Even partially.

→ Yes, for some. You can answer the repeat purchase rate question: of the transactions in this period, what share came from someone you had served before. Note carefully what this is. It is a rate over transactions, not over people, and it is biased toward your most frequent customers, because they show up in it repeatedly. It will not tell you how many individuals you kept. It will tell you reliably whether the trend is up or down, which is enough to act on.

→ No. You can answer nothing yet, and no formula will rescue you. What you can do is start today. Attach a name to every job from this week forward and you will have a first honest number one repeat window from now. Until then the usable proxy is the share of this month's revenue you can attribute to customers you recognise by name, counted by hand. It is crude, it is real, and it improves every week you keep it up.

Most one-person and two-person businesses land on branch 2 and are surprised, because they assumed retention needed software. It needs a list of dates with names on them.

How do you count returning customers from an appointment list?

Five steps, once your window is set. The example uses a four-month window and a dog groomer, but the method is the same for a physio, a handyman or a piano teacher.

Step one: pick a base month that has fully aged. With a four-month window, the most recent month you can measure is the one that ended at least four months ago. Today is September, so the latest usable base month is April. Measuring a month that has not finished aging is the single most common way owners frighten themselves with a bad number, because half the returns have not had time to happen yet.

Step two: list the distinct customers you served in the base month. Distinct is the operative word. If Mrs Halloran brought both dogs on separate days, she is one customer. Suppose April gives you 34 distinct customers.

Step three: for each, look forward exactly one window. May, June, July and August. Did that customer appear again? A yes is a yes regardless of what they bought or what they spent.

Step four: divide. Say 21 of the 34 came back. 21 ÷ 34 = 0.617, so a 62% four-month repeat rate for the April cohort.

Step five: repeat it for the three or four base months before April, so you have a short series rather than a single reading. One month is weather. Four months is a trend.

Two refinements once the basic count is running. First, measure first-time customers separately from established ones. A customer on their eighth visit returning tells you almost nothing, while a first-timer returning tells you whether your work and your booking process survive contact with a stranger. New-customer retention is usually much lower, and it is the number that actually moves when you change something. Second, if a customer left with their next appointment already booked, mark it. A booked-before-leaving rate is a leading indicator that tells you in minutes what the retention rate takes a full window to tell you.

What does a lapsed customer look like in your records?

A lapsed customer is defined by absence, which is why software so rarely surfaces one. Concretely, a customer is lapsed when the date of their last visit is older than your repeat window and they have no future booking on the calendar. That is the whole test, and it is a filter you can run on any list with a last-visit date in it.

The output is not a percentage. It is a list of names, and the list is worth considerably more than the rate. A retention rate of 62% tells you thirteen people did not come back. The lapsed list tells you which thirteen, which ones spent the most, and which ones you would actually like to have back. One of those is a report and the other is Tuesday morning's work.

Sort the list by last-visit date, oldest first, and read the top of it honestly. Some will have moved, some will have gone elsewhere, and some, more than you expect, simply forgot and will be glad to hear from you. Nothing about that call needs to be a campaign. It needs to be short, specific to them, and free of anything that sounds like a promotion.

If you want the mechanics of never letting the list get long in the first place, the system for that is a separate discipline: stop losing track of customer follow-ups covers capturing the next action at the moment it occurs to you, rather than rediscovering it four months later in a retention count.

When is a low retention rate fine?

Often enough that you should check before you treat a low number as a problem.

The work is one-time by nature. A wedding photographer, a conveyancer, a roofer, a moving company. Nobody is coming back next quarter, and a retention rate for that business measures nothing. The metric that carries the same information there is referral share: what proportion of new customers arrived because a past customer sent them. If you are in this category, measure that instead and stop feeling bad about the other one.

The business is deliberately seasonal. A landscaper's winter retention is a measure of winter, not of the landscaper. Compare season to matching season, never quarter to adjacent quarter.

You are growing fast. A month in which you served forty first-time customers and twelve regulars will produce a poor repeat rate through the base period, because first-timers return at a lower rate than established clients everywhere. That is arithmetic about your mix, not a verdict on your service. This is exactly the case that splitting new from established customers resolves.

You deliberately fired the bottom of your book. If you raised prices or dropped a service, some customers left on purpose. Retention should fall. The thing to check is margin, not headcount.

You changed the window. Already covered, still the most common cause of a retention rate that "collapsed" without anything happening.

What do you do with the number once you have it?

Three things, in order of how much they are worth.

Work the lapsed list. It is the only one of the three that generates revenue this week. Everything else is measurement.

Watch the direction, not the level. Your rate compared with last quarter's rate is information. Your rate compared with a published industry average is almost always noise, because the benchmark's period, window and definition of a customer are different from yours and usually undisclosed.

Split the rate by something you control. By service, by first-visit source, by which day of the week they came, by who did the work. The overall rate says you have a problem. The split says where it is. A groomer whose full-groom clients return at 70% and whose nail-trim clients return at 25% has just learned what to promote.

It is worth being careful with the argument that usually accompanies this topic. You will see it asserted that a 5% increase in retention raises profits by 25 to 95 percent, quoted so often that it reads as settled fact. It traces back to one source: Frederick Reichheld and W. Earl Sasser's 1990 Harvard Business Review article "Zero Defections: Quality Comes to Services", whose own claim is that companies can boost profits by almost 100% by retaining just 5% more of their customers. That was a real piece of analysis of specific service companies in 1990, and the direction it points is sound. It was not a measurement of your dog grooming business, and the wide percentage range that circulates today is a restatement of a restatement. Use the idea. Do not quote the figure at yourself as though it were a forecast.

Where this lives in SMBDashboard

The three ingredients are a customer record, a date, and a last-touched field you can sort on. In the customers module each customer carries their history and the date they were last dealt with, so the lapsed list is a sort rather than a project: order by last touched, draw the line at your repeat window, and everything below it is your call list. Appointments and money entries attach to the same customer record, which is what makes the base-month count a filter instead of a reconstruction.

It is free and there is no account. Your data stays in your browser unless you turn on Pro sync, which is off by default. The free tier holds 25 customers and 200 money entries; Pro removes both caps and adds CSV export, recurring entries and your own branding on the printed weekly report. If you are still deciding whether a customer list of any kind is worth keeping, a simple CRM is enough for most small businesses makes that case with the same records in mind, and for the appointment-heavy version of this arithmetic, salon KPIs you can work out without salon software runs rebooking rate and visit frequency off an appointment book.

Frequently asked questions

How do I calculate retention without subscriptions?

Set a repeat window, then count. Take your typical service interval and add about half again to get the window. Choose a base month that ended at least one full window ago. List the distinct customers you served in that month, then check each one for a return visit inside the window. Returned ÷ base, times 100. The window replaces the cancellation event that a subscription business gets for free, and it is the only adjustment the standard formula needs to become usable. Run it for three or four consecutive base months so you are reading a trend rather than a single month.

What is a good repeat customer rate?

There is no benchmark worth comparing yourself to, and the ones published are not measuring what you would be measuring. They rarely disclose the period, almost never disclose the window, and frequently define a customer differently from the way you do. Figures circulating for "average retention across all industries" sit anywhere from the high 30s to the high 80s depending on which page you land on, and the pages citing them mostly cite each other.

Your own history is the benchmark that carries information. Take three or four base months, get a baseline, then watch the direction. A move of several points between consecutive cohorts is worth investigating. The absolute level mostly tells you what kind of business you are in. A barber and a dentist will never have comparable rates, and neither should try.

How long before a customer counts as lost?

When their last visit is older than your repeat window and there is nothing on the calendar for them. That is a working definition for making a call list, not a judgement about the relationship. Plenty of people on a lapsed list come back. The point of the line is to force the question while it is still easy to answer, instead of noticing eighteen months later.

Do resist the urge to move the line when the number comes out badly. A window chosen after seeing the result is not a measurement, it is a mood.

Is it cheaper to keep a customer than find one?

Almost certainly in your business, and the widely quoted version of that claim is weaker than it sounds. The figure you will see most often is that acquiring a customer costs five times more than retaining one. It circulates without a traceable primary study behind it, it is stated as a universal constant when acquisition costs vary enormously by trade and by channel, and it is not a number you should plug into any decision.

What you can do is measure your own version, which takes an afternoon. Add up what you spent on getting new customers over the last quarter: advertising, directory listings, the discount on first visits, the hours you spent quoting work you did not win. Divide by the number of new customers you actually gained. Now compare that with what a reminder call costs you, which is a few minutes. The comparison is usually decisive without needing anyone else's multiplier, and it is yours, which means it survives an argument.