What It Costs to Replace a Customer Versus Keep One

Why this matters

Every shop has heard that keeping a customer is cheaper than finding a new one, and almost none can say by how much at their own shop. That gap matters, because the ratio is what decides where the next unassigned hour of office time goes. Worse, the familiar version of the claim compares two things that are not the same unit: acquiring a customer and re-booking a job are different objects on different clocks. This card shows how to compute both sides from your own service history without a CRM, and why the honest comparison changes what you do with the answer.

The two sides are not measured in the same thing

Acquisition produces a customer. The output is a relationship with an unknown number of future jobs in it.

Retention produces a job. The output is one booking, plus a reset clock, plus a marginal extension of the relationship.

You cannot divide one by the other and get a meaningful ratio until you convert both to the same unit. The unit that works for a service shop is jobs booked per non-billable hour spent, and the conversion factor between the two sides is the number of jobs a customer produces over their whole relationship with you. That factor is not a property of customers. It is a property of your shop, and it is the thing most owners assume and never measure.

Building the acquisition side from records you already have

You need two numbers for one full year. Use a year rather than a quarter, because acquisition effort and the bookings it produces are separated by weeks or months and a short window mismatches them.

Net-new customers. Count of customer records whose first job fell in the year. Not total customers served, not invoices. First jobs.

Non-billable acquisition hours. Add up every hour the shop spent creating those first jobs and would not have spent otherwise. In a small shop that is usually three lines: owner time on marketing, advertising, referral relationships and quoting for strangers; office time handling inbound calls from people who are not yet customers; and any crew time spent on unpaid site visits to prospects. Paid advertising spend is real but is deliberately excluded from this calculation, because the whole point is to price your own scarce hours against each other, and money and hours do not substitute freely in a shop where the owner is the constraint.

Divide hours by customers. That is your acquisition cost per customer, in hours.

Building the retention side

Retention hours. Time spent on the list rather than on strangers: list reviews, reminder calls to past customers, follow-ups on deferred work, the outreach on the quarterly call list.

Re-bookings produced. Jobs booked by an existing customer that are traceable to that outreach. Traceable is the hard word. If your office logs an outcome code on every touch, this is a count. If it does not, you cannot compute this side at all, and building the logging habit is the prerequisite, not an optional refinement.

Divide hours by re-bookings. That is your retention cost per booked job, in hours.

The worked example, carried through

A shop pulled one full year.

Acquisition: 96 net-new customers. Non-billable acquisition hours were 156 from the owner, which is 3 hours a week across the year, plus 58 hours of office time handling first-contact calls from non-customers. Total 214 hours. That is 214 divided by 96, or about 2.2 hours of shop time per new customer acquired.

Retention: 130 hours, which is about 2.5 hours a week of office and owner time on list work. Those hours produced 148 re-bookings with an outcome code tying them to a touch. That is 130 divided by 148, or about 0.88 hours per booked job.

The tempting sentence here is "retention is 2.5 times cheaper," and it is the sentence to refuse. Compare the two figures as rates instead, and watch the units: acquisition ran at 96 customers per 214 hours, which is 0.45 customers per hour. Retention ran at 148 jobs per 130 hours, which is 1.14 jobs per hour. Those are different nouns. One hour of acquisition work does not produce 0.45 jobs, it produces 0.45 relationships.

The conversion factor is the whole argument

To compare them, you need to know how many jobs a relationship contains. The shop pulled every customer whose first job was six or more years back and counted their total jobs. The typical relationship that survived ran about 1.4 jobs a year for about 6 years, which is 8.4 jobs.

Apply that: 0.45 customers per acquisition hour times 8.4 jobs per customer is about 3.8 jobs per acquisition hour, eventually. Against retention's 1.14 jobs per hour, acquisition looks better by a factor of about 3.3, not worse.

Except the 3.8 arrives spread over six years. Divided evenly, that is about 0.63 jobs per acquisition hour per year, against retention's 1.14 jobs per hour arriving inside twelve months. Inside a one-year window, retention wins by about 1.8 times. Over six years, acquisition wins by about 3.3 times. Both statements are true and they are answers to different questions, which is why shops argue about this endlessly with no one being wrong.

Then the shop checked the assumption underneath the whole thing. The 8.4 jobs came from customers who survived six years, which is a survivor-biased sample by construction. When it instead counted every customer acquired six years ago including the ones who never came back, the median relationship contained 2.1 jobs, not 8.4.

Redo it with 2.1: 0.45 customers per acquisition hour times 2.1 jobs is about 0.95 jobs per acquisition hour, over the whole relationship. Retention delivered 1.14 jobs per hour inside a single year. Now retention is ahead on both clocks.

The conclusion the shop drew is the one worth taking: the case for retention spending is not that it is cheaper per unit. It is that the acquisition side's return is almost entirely determined by a number, jobs per relationship, that only the retention side moves. Acquisition builds the stock. Retention sets its half-life. Cutting retention work does not shift spending from one column to the other, it lowers the yield of the acquisition column.

What a lost customer actually costs

Losing a customer mid-relationship costs three separate things, and shops usually count only the first.

The remaining jobs. A customer lost at the two-year mark, having produced about 2.8 jobs against the 8.4 a surviving relationship carries, forfeits about 5.6 jobs. That is the number people mean by lost revenue.

The replacement hours. About 2.2 hours to acquire a replacement, which does not recover the 5.6, it starts a fresh stream from zero.

The efficiency gap on the replacement's first job. At this shop, first jobs at an unknown property ran about 0.6 hours longer than an equivalent visit to a property already in the records: no access notes, no equipment history, no known layout, more diagnosis from scratch. Every replacement pays that once.

The third item is small per job and is the one that compounds invisibly in a shop with high churn. A shop replacing a third of its customers every year is running a meaningful share of its total job hours at the higher first-visit cost permanently.

What changes the answer

A trade with genuinely low repeat frequency. If the honest jobs-per-relationship figure at your shop is close to 1, retention work has almost nothing to act on and acquisition is correctly where the hours go. The mistake is assuming this about your trade rather than counting it, because most shops that assume it have a repeat rate they have never measured.

A shop at capacity. When the crew is fully booked, neither number matters as much as job mix, and hours spent on either column should go to raising the quality of the work you accept rather than the quantity. Retention still wins on one narrow ground: a known property is faster to serve, so the same crew hours produce more completed jobs.

Heavy commercial or property-management concentration. Relationship length is governed by contract cycles and by individual turnover at the client, not by service intervals. Compute the same numbers, but per account rather than per contact, and expect the retention hours to concentrate on very few names.

A shop whose churn is a quality problem. If customers leave because of workmanship or communication rather than drift, retention outreach is spending hours to re-contact people who made a decision. The numbers will look terrible and the correct read is that the outreach is not what is broken.

How to verify you got this right

Check that your re-booking count is traceable, not inferred. If retention re-bookings were counted as "any job by an existing customer," you have credited the retention program with every customer who would have called anyway, and the hours-per-booking figure is meaningless. The count must come from outcome codes on actual touches.

Check whether your jobs-per-relationship figure is survivor-biased. Ask explicitly: did the sample include customers who never came back? If the cohort was defined as "customers with more than one job," the answer is no, and the resulting number will be roughly the multiple this example saw, four times too high.

Check that both sides used the same year. Acquisition hours from a heavy marketing year compared against retention bookings from a quiet one produces a ratio that describes nothing.

Recompute annually, not once. These numbers move when the shop changes, and the most common way this analysis fails is being run once, producing a decision, and then being cited for years after the shop that produced it stopped existing.

References

• U.S. Small Business Administration, small-business marketing and customer-retention guidance
• Trade-standard practice for service-history recording and job costing
• See related: Customer Lifetime Value
• See related: Customer Acquisition Cost (CAC) Framework for Trade Businesses
• See related: Customer Retention Economics + Strategy
• See related: The Quarterly Customer List Review SOP