The number most service businesses guess at
Ask an owner what a customer is worth and you will usually get the average ticket. That is the value of a job, not a customer, and the gap between the two is where most marketing decisions go wrong. If your average job is $340 and you refuse to spend more than that acquiring a customer, you are pricing as though every customer buys once and vanishes. Some do. The ones who do not are carrying the business, and you cannot see them in a per-job number.
As of September 2026 the practical problem is not that owners disbelieve in lifetime value. It is that the formula they find when they go looking was written for software companies, it does not fit a trade at all, and the mismatch is obvious enough that most people try it once and abandon it. This guide builds the version that works when jobs are irregular and years apart, shows the data you have to be capturing for the number to mean anything, and covers what actually changes the day you know it.
Why the standard formula does not transfer
The lifetime value formula you will find everywhere is some variant of average revenue per user divided by churn rate. It is elegant and it depends on two things a service business does not have.
The first is a fixed recurring amount per period. A software customer pays the same amount every month, so revenue per period is a real number rather than an average over a wildly uneven distribution. Your customers do not do this. One calls twice in a fortnight during a bad week, then nothing for three years. Another has a single large job and never returns. A third calls every autumn like clockwork. Averaging those into a monthly figure produces a number that describes none of them.
The second is an observable cancellation. A subscription business knows the exact day a customer left, so churn rate is a measurement. You have no cancellation event at all — the customer simply stops calling, and you find out never. Feeding an estimated churn rate into a denominator gives the arithmetic a false precision, because the whole result now hinges on a guess and small changes to that guess swing the answer enormously.
There is a third mismatch nobody mentions. In software, usage is continuous and mostly free to serve; the marginal cost of one more month is near zero. In a trade, every single purchase occasion costs you a technician, a truck, drive time, and parts. Cost to serve is not a rounding error in your lifetime value — for many businesses it is roughly half the revenue, and any formula that ignores it is measuring the wrong quantity entirely.
So put the subscription formula down. Build from purchase occasions instead, because purchase occasions are what you actually record.
The practical trade version
Here is the version that works, in one line:
Lifetime value = average job value × jobs per year × expected years of relationship × gross margin − cost to serve the relationship
Four inputs and a subtraction. Take them one at a time, because each has a trap.
Average job value. Use the median rather than the mean if you have a small number of very large jobs, otherwise one commercial roof replacement drags every residential customer's apparent value upward. And segment before you average. A single blended figure across residential and commercial customers describes a customer who does not exist.
Jobs per year. This is where the irregularity bites, and the fix is to stop thinking annually. Count total jobs a repeat customer has bought and divide by the number of years since their first job. A customer with four jobs over six years is 0.67 jobs a year, which looks unimpressive until you multiply it by six years and a healthy margin. Fractional job frequency is normal in the trades and it is not a sign you have measured it wrong.
Expected years of relationship. The hardest input and the one owners inflate most. You cannot observe the future, so anchor on what your data already shows: how long have your longest-running customers actually stayed, and what share of first-time customers ever buy a second time? If your oldest customer relationship is five years because that is how old the business is, you do not get to assume eight. Use a term you have evidence for and note the assumption explicitly, because everything downstream multiplies by it.
Gross margin. Lifetime value is profit, not revenue. If you have not built job-level costing, this input is a guess and the whole calculation inherits the error — the job costing guide is the genuine prerequisite here, not optional background reading.
Cost to serve the relationship. The part almost everybody drops. Every repeat job costs something to book, dispatch, invoice and chase. If a customer generates six jobs over their lifetime and each carries a modest administrative and scheduling cost beyond the direct job cost already in your margin, that adds up to real money over a long relationship. Subtract it, even roughly.
Worked through
Take a residential customer, and treat every figure as an illustration to replace with your own.
Average job value $340. Jobs per year 0.7. Expected relationship five years. That is 3.5 jobs over the lifetime, or $1,190 of revenue. At 55 percent gross margin that is $654 of gross profit. Subtract an estimated $15 per job of relationship overhead — scheduling, reminders, invoicing, collections — which is about $53 across 3.5 jobs, and the customer is worth roughly $601 in gross profit over five years.
Compare that to the average ticket of $340, and against a gross profit per job of $187. The customer is worth more than three times a single job. That multiple is the entire reason this calculation matters, and it is the number that should be sitting behind every decision about what you are willing to pay for a lead.
What the number changes: the maximum you can pay per lead
The most immediate consequence is your advertising ceiling, and it is usually the moment the arithmetic stops being academic.
Work backwards. If a customer is worth $601 of gross profit, and you want a three-to-one return on acquisition spend, you can rationally pay up to about $200 to acquire one. That is the acquisition cost, not the lead cost, and confusing the two is the most expensive mistake in this whole area. If one in four leads becomes a customer, then $200 per acquired customer means you can pay up to $50 per lead. If your close rate is one in three, you can pay about $67. The close rate is doing as much work in that sentence as the lifetime value is, which is why the cost per lead guide belongs directly alongside this one — the two numbers are useless apart.
Two adjustments keep this honest.
Payback period matters as much as the ratio. A customer worth $601 over five years does not pay you $601 today. If most of that value arrives in years three through five, you are funding the acquisition cost out of working capital for a long time. Two businesses with identical lifetime arithmetic can have completely different survival odds depending on how fast the first job repays the spend. Track how much of the value lands in the first ninety days as a separate figure, and let it set how aggressively you are willing to scale spend.
Segment ceilings, do not blend them. If commercial accounts are worth several times residential, one blended maximum bid means you systematically underbid on the leads worth having and overbid on the ones that are not. Where your channels and segments produce different values, they deserve different ceilings.
Membership customers, transactional customers, and commercial accounts
The single biggest lever in the formula is not price. It is the years term, because it multiplies everything else — and the reliable way to extend it is to give the customer a reason to stay attached between emergencies.
| Over five years, illustrative | Transactional residential | Membership residential | Small commercial account |
|---|---|---|---|
| How they arrive | Paid search or a directory, one urgent job | Same, then converted on site at the end of a job | Referral or direct outreach |
| Recurring fee | None | $19 a month, $228 a year | None, but contracted scheduled work |
| Scheduled visits per year | 0 | 2 | 4 |
| Additional repair jobs per year | 0.7 | 1.0 | 1.5 |
| Average job value | $340 | $340 | $520 |
| Five-year revenue | $1,190 | $2,840 | $3,900 |
| Gross margin assumption | 55 percent | 52 percent blended with the plan fee | 50 percent |
| Five-year gross profit | $654 | $1,476 | $1,950 |
| Less relationship overhead | $53 | $110 | $150 |
| Five-year lifetime value | $601 | $1,366 | $1,800 |
| Realistic retention risk | Highest - no attachment between jobs | Lower - lapses at renewal if nobody calls | Lower, but concentrated - one contact leaving can end it |
Every figure in that table is illustrative and built to be replaced with your own. What it is meant to show is the shape, not the amounts.
The membership customer is worth more than double the transactional one, and almost none of that comes from the plan fee. The fee brings in $228 a year against costs that consume most of it. The value comes from the scheduled visits producing more repair work and from the relationship lasting longer, which is exactly the argument made from the other direction in the service agreement and membership plans playbook. If you take one operational decision away from this guide, it is that converting transactional customers to members is the highest-leverage change available, because it moves two multipliers at once.
The commercial account is worth the most per relationship and carries a different risk profile. Its value is concentrated in fewer relationships, and it can end in a single conversation when a facilities manager changes jobs. Higher value, lumpier loss.
The referral tail
There is a component of lifetime value that never appears in the formula and is real anyway: the customers a good customer brings you.
Be careful here, because this is also where the arithmetic most often turns into fiction. The disciplined way to include it is to measure, not to assume. If your intake reliably captures how each new customer heard about you, you can calculate what share of new customers arrive as referrals, and attribute those back. If a hundred customers produce twelve referred customers over their lifetimes, then each customer carries roughly 0.12 of another customer's value — about $72 on a $601 base — and you can add that to the figure with a clear conscience.
If you are not capturing source properly, leave the referral tail out of the number entirely and treat it as upside. An unmeasured referral multiplier is how a $600 customer becomes a $2,000 customer on a whiteboard and a business overspends on advertising for a year.
Reviews behave like a public, compounding version of the same effect. A customer who leaves a good review is doing referral work continuously for people you will never trace, which is genuinely valuable and genuinely unmeasurable at the individual level. The right response is to systematise asking rather than to invent a number for it — the review generation guide covers the mechanics.
The data you must be capturing
Everything above collapses without a few unglamorous data disciplines. This section is the one worth acting on first, because no formula rescues bad inputs.
A persistent customer record. This is the foundation and the most common failure. If the same household appears as three separate customers because they called from a mobile once, a landline once, and a spouse's phone once, your repeat rate is understated and your lifetime value is a third of the truth. Deduplication by phone, address and name is not housekeeping — it is the difference between measuring customers and measuring phone calls.
Every job tied to that record, with a date. You need the sequence, because jobs per year and years of relationship both come from the gaps between dates. A job recorded without a customer link is revenue you can count and a relationship you cannot.
True cost per job. Loaded labour including drive time, parts at real cost, and the vehicle. Without it you have lifetime revenue, which can point you toward exactly the wrong customers — the high-revenue, low-margin ones you are busiest serving and least profitable keeping.
Acquisition source on the first job. Not the most recent job, the first. Attribution belongs to whatever brought them in originally, and if you overwrite the source field on each new job you destroy the ability to compare channels by the quality of the customers they produce rather than the volume of leads.
The lost customers, still in the system. Lifetime value calculated only on customers you still serve is biased upward by definition, because it excludes everyone who left. Keep the dormant ones in the denominator. Identifying them is its own discipline, covered in the at-risk customer detection guide, and the two calculations feed each other: knowing what a customer is worth is what tells you how hard to work to keep one.
This is the practical argument for the customer history, job records, quotes and call log living in one operations layer rather than three tools that each hold a piece. On Run with Jarvis the CRM, the booking calendar, the invoicing and the call records write to the same customer record, which is what makes a five-year job sequence readable at all. Where it gets useful day to day is the querying: you can ask the Jarvis brain how many jobs your repeat customers average, or which acquisition sources produced the customers who came back, and get an answer in a sentence rather than commissioning a report. Plan details are on the pricing page.
Six ways owners overstate the number
Each of these alone can double the figure. Two together and you are budgeting off fiction.
Using revenue instead of gross profit. The most common by a distance. A customer generating $1,190 of revenue at 55 percent margin is worth $654 in profit, and only the second number can be compared against acquisition spend. Any lifetime value figure quoted without a margin attached should be assumed to be revenue.
Assuming a relationship length nobody has demonstrated. If your business is four years old, you have no five-year customers, and a ten-year assumption is a wish. Use the longest term your records support, and if you want a longer projection, state it separately as a scenario rather than folding it into the headline number.
Averaging across customer types. Blending commercial and residential produces a figure that overstates residential value and understates commercial. Since you will use these numbers to set bids on channels that reach the two groups differently, the blended number is worse than useless — it is actively misleading in both directions.
Ignoring cost to serve on repeat jobs. Every subsequent job costs money to book, dispatch and collect. Over six jobs that is a real subtraction, and leaving it out is how a formula that already flatters you flatters you further.
Survivor bias. Calculating from your loyal customers and calling the result the average. If a large share of first-time customers never return, they belong in the denominator. Their zero is part of the truth.
Counting the referral tail you never measured. Covered above, and worth repeating because it is the one that produces the most spectacular overestimates. If you cannot show the referral rate from your own intake data, it is not in the number.
The correcting habit for all six is to state your assumptions next to the result. Write the number as "roughly $600 in gross profit over five years, assuming 0.7 jobs a year, 55 percent margin and a five-year term" rather than as "our lifetime value is $600". The first can be argued with and improved. The second becomes a fact nobody remembers the basis for.
What to do differently once you know it
Re-set the acquisition ceiling and act on it. Most owners are underspending on channels that work because they benchmark against average ticket. If the arithmetic supports a higher cost per acquired customer, the correct response is to raise the bid where the leads convert, not to keep the ceiling and complain about volume.
Rank channels by the customers they produce, not the leads they deliver. A source that delivers cheap leads who buy once and leave is worse than an expensive source producing customers who stay five years, and only lifetime value can see the difference. This requires the acquisition source discipline above, which is why that field matters more than it looks.
Fund retention properly. If a customer is worth $601 and reactivating a dormant one costs a fraction of acquiring a new one, retention deserves budget and calendar time rather than whatever is left after the marketing spend. Most service businesses have this backwards, and the arithmetic is not close.
Push the membership conversion. Two multipliers move at once — job frequency and relationship length — which is why membership programmes outperform any pricing tweak you could make instead.
Put it on the dashboard and watch it move. Lifetime value is not a one-off calculation, it is a metric that drifts as your job mix, margins and retention change. Recalculate quarterly, segment it, and keep it next to acquisition cost so the ratio is visible rather than reconstructed. Where it belongs alongside your other operating numbers is covered in the KPI dashboard guide. For broader context on small business financial planning, the guidance published at sba.gov is a reasonable starting point.
Put it together
Do not use the subscription formula. It assumes a fixed monthly payment and an observable cancellation, and you have neither. Build the number from what you actually record: average job value, jobs per year, a relationship length your own data supports, gross margin, minus the cost of serving the relationship. Segment it, because a blended figure describes nobody. State your assumptions alongside the result so it stays arguable.
Then use it. The number is only worth calculating because it changes three decisions — what you will pay for a customer, which channels you keep funding, and how much retention deserves. Get it roughly right with honest inputs and it will serve you better than a precise number built on a five-year term you invented.
