Customer acquisition cost by channel: how to calculate it from closed sales

Customer acquisition cost by channel is a channel's cost divided by the new customers it produced, counted from closed sales with repeat customers left out. Rank channels by it, not cost per lead, because the two often disagree. Keep untraced customers in their own line, use fully loaded cost, and check payback: CAC divided by monthly gross profit per customer.

A pest control company buys leads from four places. Facebook ads bring them in at $15 each, half the price of search. On a cost-per-lead report Facebook is the best channel the company has, and direct mail, at $60 a lead, is the worst. Count the customers instead and the order flips: each new Facebook customer cost $240, and each direct mail customer $180. The cheapest leads were the most expensive customers.

This post is for businesses where a lead is a phone call or a form and a sale is an invoice in another system: home services, dentists, law firms, roofing and anyone else who closes by phone, in a showroom or at the kitchen table. Our post on customer lifetime value by channel covers the other side of the ratio, what each customer is worth.

The customer acquisition cost formula

Customer acquisition cost (CAC) by channel = the channel's cost in a period ÷ the new customers that channel produced from that period's leads. If search ads cost $9,000 last quarter and produced 60 new customers, search's CAC is $150. Both halves need care, and the bottom half is the one most reports get wrong.

The top of the formula is invoices you already have. The bottom is a count of people, and for an offline business it exists nowhere until you build it. The ad platform knows how many calls and forms it sent; the sales system knows who bought. Neither knows which buyers came from which ads, or who was new. A CAC built on the wrong count is a precise number for the wrong thing.

Count new customers from closed sales, not leads

A new customer is a person who bought for the first time in the period, found in the sales records and traced to a channel by matching their phone or email to its leads. A lead, a booked appointment or a platform conversion is not a customer; dividing by any of them gives a cost per inquiry under another name.

A lead count includes everything a sale count does not:

• Leads that never bought. Most of them, in a share that varies by channel, which is the whole problem.
• Duplicates. The same person calling twice and filling in a form, counted three times. Our post on duplicate leads covers how to fold them together.
• Spam, wrong numbers and people outside your service area. Cheap channels tend to carry more of these.
• Existing customers. Someone on a service plan calling to reschedule is a lead to the ad platform and an acquisition to nobody.

The ad platforms' cost-per-acquisition figures are cost per lead in this sense: spend divided by whatever you told the platform to count, which for an offline business is usually a call or a form.

Our post on cost per lead makes the full case against allocating budget on it. The fix is the same join it describes, with people as the denominator: export closed sales, match them to leads on phone and email, and count. Our post on marketing attribution for small business sets out that match for a business with no data team.

Leave repeat customers out

Customer acquisition cost counts customers acquired, so anyone who had bought from you before the period is excluded, even when their latest sale matched a lead from the channel. Leaving them in makes the channels that existing customers use to come back, usually search and mail, look far cheaper than they are.

Existing customers generate leads all the time. They search your company name, click the ad above it, and call. They answer a postcard mailed to a list that included them. Each is a real lead and a real sale, and none is an acquisition.

To find them, take the sales history from before the period and normalize the phone numbers and emails in it, so that (555) 010-2030 and +1 555 010 2030 are the same person; our post on phone normalization covers why this step fails most often. Any customer in the period whose phone or email appears in the earlier history is a repeat customer. The history needs to reach back at least as far as a normal gap between purchases: a year for quarterly pest control, several years for roofing or HVAC replacements.

What about customers who lapsed?

A customer who left two years ago and came back through a search ad is a judgment call. Pick a rule, such as no purchase in 24 months counts as new again, write it down, and apply it every quarter. A CAC that redefines "new" each quarter cannot show a trend.

A worked example: four channels, two rankings

A pest control company, leads from April to June 2026, sales closed through 31 July. The figures are illustrative, not benchmarks. Ranked by cost per lead and by customer acquisition cost, the four channels come out in nearly opposite orders, and only the second says what a customer cost.

The arithmetic. Cost per lead is spend ÷ leads: $9,000 ÷ 300 = $30, $4,800 ÷ 320 = $15, $6,000 ÷ 240 = $25, $5,400 ÷ 90 = $60. New customers are matched sales minus customers who had bought before: 75 − 15 = 60, 24 − 4 = 20, 30 − 0 = 30, 36 − 6 = 30. CAC is spend ÷ new customers: $9,000 ÷ 60 = $150, $4,800 ÷ 20 = $240, $6,000 ÷ 30 = $200, $5,400 ÷ 30 = $180. For the four together, $25,200 ÷ 140 = $180.

By cost per lead the order is Facebook, marketplace, search, mail. By CAC it is search, mail, marketplace, Facebook. The channel at the top of one list is at the bottom of the other, and the mailer moves from last to second.

The two numbers are tied together by one more: the share of a channel's leads that became new customers. CAC = cost per lead ÷ that rate. That share is the lead conversion rate, and our post on lead conversion rate by source covers measuring it without flattering it. Search turned 60 of 300 leads into customers, 20%, so $30 ÷ 0.20 = $150. Facebook turned 20 of 320, 6.25%, so $15 ÷ 0.0625 = $240. The marketplace 12.5%, so $25 ÷ 0.125 = $200. Mail a third, so $60 ÷ (1 ÷ 3) = $180. A lead four times cheaper that closes five times less often is not a bargain; our post on lead marketplaces explains why shared leads close as they do.

Leave the repeat customers in, as a report built straight from matched sales would, and search becomes $9,000 ÷ 75 = $120 and mail $5,400 ÷ 36 = $150: 20% and 17% cheaper than they really were. Facebook ($4,800 ÷ 24 = $200) and the marketplace barely move. The lead looks wider than it is, padded by people already paying every quarter.

Paid-media CAC vs fully loaded CAC

Paid-media CAC divides ad spend alone by new customers. Fully loaded CAC adds every other cost of acquiring them: agency fees, tools, creative, a channel's staff time and, company-wide, the people and systems that handle leads. Paid-media CAC is quick to update; fully loaded CAC is the one to compare with what a customer is worth.

That was written for software companies with sales teams, but the principle carries across: a customer cost what it took to get them, not what the platform billed. Our post on marketing ROI lists which costs belong to a channel and costs staff time at the loaded rate, $46.89 an hour from BLS figures, rather than the wage. In the example: an agency fee on search, management and creative on Facebook, the office manager's hours disputing bad marketplace leads, and the mailer's design.

The arithmetic. Fully loaded cost is ad spend plus other channel costs: $9,000 + $1,350 = $10,350, $4,800 + $900 = $5,700, $6,000 + $750 = $6,750, $5,400 + $600 = $6,000. Divided by new customers: $10,350 ÷ 60 = $172.50, $5,700 ÷ 20 = $285, $6,750 ÷ 30 = $225, $6,000 ÷ 30 = $200. Sixteen hours at $46.89 is $750.24, rounded. Together, $28,800 ÷ 140 = $205.71.

The order holds in this example; the sizes do not. Facebook's customers cost $45 more each once its management fee is in, and the marketplace, which looked like it charged only for leads, costs $25 more per customer in somebody's afternoons. Report both columns, label which is which, and never put a paid-media CAC next to a lifetime value and call the gap profit.

Blended CAC vs per-channel CAC

Blended CAC divides total marketing cost by every new customer in the period, traced or not. Per-channel CAC divides each channel's cost by the new customers matched to that channel's leads. Blended is usually lower, because referrals and walk-ins count as customers the marketing paid for, and it cannot say which channel to cut.

In the example the company gained 200 new customers in the quarter. The four channels' leads account for 140; the other 60 matched no lead at all: neighbors' referrals, people who saw a truck, customers whose call was never logged. The company also pays $4,200 a quarter for costs no single channel owns: a website and SEO retainer at $1,000 a month and $400 a month for call tracking and the CRM.

The arithmetic: $25,200 ÷ 200 = $126. $28,800 + $4,200 = $33,000, and $33,000 ÷ 200 = $165. The same quarter's acquisition cost is $126 or $205.71 depending on which line someone quotes, and both answer real questions. Blended fully loaded CAC is the company-wide health check; per-channel CAC is the allocation tool. Shared costs like the website belong in the blended figure only. Our post on blended ROAS explains why a blended ratio flatters every channel at once, and the same argument applies to blended CAC.

How Direct / Unknown customers change CAC

New customers who match no lead, usually reported as Direct / Unknown, are left out of every channel's count. That keeps per-channel CAC honest about what each channel can be shown to have produced. Spreading them across channels in proportion lowers every CAC by the same factor and turns per-channel CAC back into blended CAC.

Try it. Share the 60 unknown customers among the four channels in proportion to their matched customers and search gets 60 × 60 ÷ 140 = 25.7 more, for 85.7 in all. Its CAC falls from $150 to $9,000 ÷ 85.7 = $105. Every other channel falls by the same 30%: Facebook to $168, the marketplace to $140, mail to $126, and the four together to $126, which is the blended figure. Nothing new was learned about any channel; the referrals were simply billed to the ads.

So treat matched CAC as an upper bound and say what the unknown share probably is. If most of the 60 are referrals, the matched figures are close to the truth. If many are leads nobody recorded, such as calls to an untracked number, true per-channel CAC is lower, and the fix is better records, not allocation. Our posts on unattributed revenue and match rates cover how to tell which, and how to shrink the bucket with evidence.

CAC payback period

CAC payback period = customer acquisition cost ÷ monthly gross profit per new customer, where monthly gross profit is monthly revenue per customer times gross margin. It answers the cash question a CAC alone hides: how many months a new customer must stay before they have paid for their own acquisition.

A local service business rarely has 20 months of cash to wait. In the example, new customers sign up for recurring service; Facebook's came in on a discounted first-year price, so they pay less a month. Margin is 50% throughout, to keep the arithmetic readable.

The arithmetic. Monthly gross profit is revenue × 50%: $27.50, $22.50, $25.00, $27.50. Payback is CAC ÷ that: $172.50 ÷ $27.50 = 6.27, $285 ÷ $22.50 = 12.67, $225 ÷ $25.00 = 9.00, $200 ÷ $27.50 = 7.27 months.

Payback assumes the customer is still there. If discount customers tend to cancel when the first year ends, a Facebook customer who needs 12.7 months to pay back often never does. Payback on revenue would have said 6.3 months ($285 ÷ $45), which is how a losing channel looks like a good one. Whether customers stay long enough is the lifetime value question, including the LTV-to-CAC ratio, that our customer lifetime value post answers.

Time lag: count customers by the month of their lead

A channel's cost lands when the ads run; its customers close days or weeks later. Dividing a month's spend by the customers who closed that month overstates CAC for any channel that is new, growing or seasonal. Count each new customer against the month of their earliest lead, and read the figure only after most of those leads have closed.

In the example, most of the Facebook budget ran in June. On 30 June, only 12 of the 20 customers it would produce had bought, and its CAC read $4,800 ÷ 12 = $400, the worst in the business by a distance. By 31 July all 20 had closed and it read $240. Nothing changed in July except that the report waited. That is why the worked example allows a month after the last lead.

Ad platforms stop counting after a conversion window, 30 days by default in Google Ads. Your sales records have no window, which is the point of counting from them, but the lead export has one in effect: the dates it covers. A customer who first called in March cannot be traced to that call if the lead file starts in April; they land in Direct / Unknown, or on whatever lead they made next. Export leads from well before the period, measure the typical gap from first lead to first sale in the pairs you matched, and wait at least that long before reading a quarter. Our post on long sales cycles covers reporting periods that have not finished closing.

CloseRev does not calculate customer acquisition cost, and it does not know your costs beyond the ad spend you give it. It reads two CSV files you export, closed sales and leads, matches each sale to leads on normalized phone and email, and credits it to the source of the earliest matched lead. Only high-confidence matches count automatically; medium- and low-confidence ones stay reviewable, and unmatched sales stay in Direct / Unknown. If you enter monthly spend per channel, the report shows ROAS for those channels. Excluding repeat customers and dividing fully loaded cost by new customers stay spreadsheet steps.

Where to start

Take one quarter of leads, the sales that followed them, and the sales history before them. Match, remove repeat customers, and divide each channel's cost by its new customers, first on ad spend and then fully loaded. Put cost per lead beside it. Where the two rankings disagree is the budget conversation worth having.

1. Export a quarter of leads with a source column, and the closed sales from that quarter plus a month after it, with date, amount, phone and email.
2. Export the sales history from before the quarter, at least one normal buying cycle deep, to identify existing customers.
3. Match sales to leads on normalized phone and email, credit each to its earliest lead, and keep unmatched sales in Direct / Unknown.
4. Drop every matched customer who appears in the earlier history, under a written rule for lapsed customers.
5. Divide each channel's ad spend by its new customers, then do it again with fully loaded cost, and compute blended CAC on all new customers separately.
6. Divide each fully loaded CAC by that channel's monthly gross profit per customer to get payback.
7. Repeat every quarter with the same rules, so the trend means something.

Cost per lead tells you what a channel charges for an inquiry. Customer acquisition cost tells you what it charged for a customer. Count the customers in your own sales records, leave out the ones you already had, add the costs the invoice does not show, and the cheapest channel on the dashboard often turns out to be the most expensive.