Customer acquisition cost: the real math and how to lower it

Thursday, 10:10 p.m.: your customer acquisition cost isn't on the dashboard

Thursday, 10:10 p.m. The day is finally over and you open Google Ads on the tablet. Under "Cost / conv.": $60. So a customer costs you $60. What a deal.

Except that $60 is the price of an enquiry, not a customer. Google gets it by dividing your total cost by the number of conversions, and for you, a conversion is a form or a phone call. Between that enquiry and a signed contract there's a callback, a site visit, a quote and, often, a no. And your evenings.

Our position: a small business's real customer acquisition cost includes the owner's time, is worked out channel by channel and is judged against the margin on the first job. Done that way, the math often crowns a different winner. The "free" channel turns out to be the most expensive, and the fastest way to bring the cost down isn't buying cheaper clicks. It's signing more of the enquiries you already get.

The four-number sheet

BDC gives the basic formula: (sales costs + marketing costs) ÷ number of new customers acquired through those activities. Everything depends on what goes on top of that fraction. For each channel (Google Ads, social media, referrals, a trade show), fill in a sheet with four numbers covering at least three months:

1. Money spent: ads, but also the designer, printing, thank-you gift cards, platform fees.
2. Hours, multiplied by a rate: posting, managing ads, calling back, site visits, writing quotes. Including yours.
3. New customers: people buying from you for the first time, who came through this channel.
4. Gross margin on the first sale: the first job minus its direct costs (materials, crew wages on site, subcontractors).

Channel acquisition cost = (number 1 + number 2) ÷ number 3. Number 4 is for the test further down.

The number almost everyone leaves out is the second one. BDC's example counts a salesperson's salary; in an eight-person business, the salesperson is you, at night. What rate should you use? At the very least, what it would cost to pay someone to do the work: according to Quebec's statistics institute, employees in Quebec earned an average of $35.05 an hour in 2025, before payroll costs. If an hour of your time billed to a client is worth more, use that rate.

Two more traps. A returning customer isn't an acquired customer: counting them makes everything look better. And the average across all channels, the "blended" cost, almost always hides one channel paying for another.

If you report in French, a note on wording: Quebec's language office, the OQLF, recommends "coût de conversion du client" and discourages "coût d'acquisition du client," an anglicism (in French, you can't acquire a person). The math is the same.

A worked example: a Trois-Rivières painter, three channels

Take a four-person residential painting company in Trois-Rivières, from April to June. Average job: $3,500, with a 35% gross margin, so $1,225 per first job. The owner does every site visit: an hour and a half including the drive and the quote, at $50 an hour, so $75 per visit. All of these numbers are assumptions for the example.

• Google Ads: $2,400 in ads, 10 hours of management ($500), 40 enquiries, 24 visits ($1,800), 6 new customers. Total: $2,400 + $500 + $1,800 = $4,700, so $4,700 ÷ 6 ≈ $783 per customer.
• Facebook and Instagram, no ads: $0 spent, but 40 evening hours of posting ($2,000), 15 enquiries, 8 visits ($600), 2 new customers. Total: $2,600, so $1,300 per customer.
• Referrals: 5 thank-you gift cards at $50 ($250), 10 enquiries, 8 visits ($600), 5 new customers. Total: $850, so $170 per customer.

Altogether: $4,700 + $2,600 + $850 = $8,150 for 13 new customers, about $627 each.

The usual calculation divides the $2,400 in ads by the quarter's 18 jobs, including 5 past customers who came back on their own: $133. Same quarter, two numbers: $133 on the dashboard, $627 in real life.

And that blended $627 hides what matters most: a referred customer costs $170, a customer from "free" social media costs $1,300. The channel that never shows up on the credit card statement is the most expensive one in the business.

The first-job test

An acquisition cost isn't good or bad on its own: $783 for a painting contract is reasonable; for a haircut, it would sink the business. Hence a simple rule, the first-job test: a channel passes if its acquisition cost stays below the gross margin on the first sale.

In the example, Google Ads passes: it eats up 64% of the margin on the first job ($783 ÷ $1,225). Referrals pass easily, at 14% ($170 ÷ $1,225). Social media fails: at $1,300 against a $1,225 margin, each new customer costs $75 more than it brings in, before the truck and the insurance.

Why the first job rather than "lifetime value"? Because lifetime value is a forecast, and an optimistic forecast can justify any expense. BDC suggests, as a very general guideline, a lifetime value 2.5 to 3 times your acquisition cost, and notes that benchmarks vary. The 3-to-1 ratio you see everywhere comes mostly from subscription software: David Skok sets it out for SaaS businesses with gross margins of 80% or more, recovering the cost in under 12 months. A painter doesn't live in that world.

Instead, fit the test to how often your customers come back. If a customer repaints every five or six years, the first job has to pay. If they come back several times a year (a massage therapy clinic, a shop, a restaurant), you can accept a cost up to the margin from their first 12 months, as long as you measure those return visits in your own data rather than hoping for them. And look after the customers you have: a Harvard Business Review article notes that acquiring a customer costs 5 to 25 times more than keeping one. That's a wide range, but the direction is clear.

The cheapest lever: close more of your enquiries

When the cost climbs, the reflex is to hunt for cheaper clicks. Let's rerun the Google Ads numbers with three levers, one at a time, everything else unchanged.

Lever	Cost per Google Ads customer	Drop
Starting point	$783	None
Clicks 20% cheaper ($1,920 in ads)	$703	10%
Three extra questions on the first call rule out 8 visits	$683	13%
8 jobs from 24 visits instead of 6	$588	25%

The math: ($1,920 + $500 + $1,800) ÷ 6 ≈ $703; ($2,400 + $500 + 16 × $75) ÷ 6 ≈ $683, assuming the call was happening anyway and the 8 visits you skip would never have turned into jobs; $4,700 ÷ 8 ≈ $588. Raising your close rate (jobs ÷ visits) from 25% to 33% lowers the cost nearly two and a half times as much as clicks 20% cheaper ($195 versus $80). And those two jobs add 2 × $1,225 = $2,450 in margin.

Clicks 20% cheaper take weeks of fine-tuning or an agency. Closing one visit in three instead of one in four often comes down to simple habits: calling back the same day, offering two options instead of one, following up once on every quote.

Watch the seasons, too. The same painter in January: $600 in ads, 2 hours of management ($100), 6 enquiries, 4 visits ($300), 1 new customer. Cost: $1,000. Comparing January to May tells you nothing; compare January to last January and judge each channel over a rolling 12 months. Cutting a channel in January because it "costs too much" can starve you of March's enquiries.

Your plan for the next four weeks

1. Week 1: pull together 12 months of sales and marketing spending (card statements, invoices, gift cards). Ask your accountant to help you separate direct costs so you get your gross margin: this is a management calculation, not a tax one.
2. Week 2: list the period's new customers with their source. Cross off anyone who had bought from you before.
3. Week 3: estimate hours per channel from your calendar and email, then multiply by your rate.
4. Week 4: fill in the four-number sheet for each channel and run the first-job test. Before you touch the budget, look at your close rate.

From then on, add a "Source" column to your customer list and ask on every first call: "How did you hear about us?" Include "ChatGPT or another AI" as an answer. ChatGPT automatically adds utm_source=chatgpt.com to links to your site, so those visits show up in your web analytics, but a customer who reads the answer and then picks up the phone leaves no trace. No source, no cost per channel.

One last limit. With two customers a quarter, losing a single one doubles the cost: below about ten new customers per channel, read a trend over six to twelve months, not a verdict. And an Instagram post can nudge someone who later finds you on Google: before you drop a channel that fails the test, try making it less hungry for your hours.