How to calculate cost per incremental customer (with worked examples)

To calculate cost per incremental customer, divide the ad spend you’re evaluating by the number of extra customers that spend caused. The arithmetic is one line. The hard part is the word “caused,” and that is what most of this guide is about.

It’s the most useful single number a multi-location brand can put on a channel. It cuts through platform reporting, it compares across channels that report in different currencies (leads, calls, store visits), and it plugs straight into the question finance asks: if we spend another $10,000, how many more customers do we get?

The cost per incremental customer formula

Cost per incremental customer = spend ÷ (actual customers − customers you’d have had without the spend).

“Customers” means whatever your locations count: guests, orders, new members, new patients, booked jobs. “Customers you’d have had” is the counterfactual, the part nobody can observe directly. You estimate it from markets where spend didn’t change, weighted so they tracked the test markets beforehand. Everything downstream depends on how good that estimate is.

Cost per incremental customer vs CPA and CAC

How to calculate it, step by step

1. Fix the window. The test period plus a tail of three to four weeks, since customers keep arriving after the media changes.
2. Pick the outcome. New customers if you can separate them, otherwise total transactions. Use counts, not revenue: average ticket moves for reasons that have nothing to do with ads.
3. Build the comparison. A weighted blend of steady-spend markets, fitted on 26 weeks or more of pre-period data.
4. Subtract. Actual customers in the test markets minus what the blend says they’d have done. That’s incremental customers, with a range.
5. Choose the right spend. Added spend for a heavy-up. The paused spend for a go-dark test. Include production and agency fees if you want the fully loaded number, and say so.
6. Divide, three times. Spend ÷ the central estimate, ÷ the high end, ÷ the low end. Report all three.

Worked example 1: a heavy-up for a dental group

Illustrative numbers. An 18-practice dental group adds $9,600 a month of Google Ads and Meta in 4 markets covering 7 practices, for 3 months. Each practice normally sees about 60 new patients a month. The other markets keep spend flat.

The $47 CPA isn’t a lie. Those were real calls and form fills. Most of them were people who would have booked anyway, or calls that never turned into a first visit. The new patients the money actually added cost almost four times as much.

Notice the range is lopsided: $127 to $294 around a central $178. That always happens, because you’re dividing by an uncertain number. Small denominators swing costs a lot. Don’t average the ends; report them as they are.

Worked example 2: a go-dark test for a pizza chain

Illustrative numbers. A 22-location pizza chain pauses non-brand paid search in 5 markets (6 stores) for 6 weeks. Those markets normally get $3,500 a week. The comparison blend says the 6 stores would have taken 52,000 orders with search on; they took 51,064.

• Orders search was driving: 52,000 − 51,064 = 936, or 1.8%.
• Spend that was paused: $3,500 × 6 = $21,000.
• Cost per incremental order: $21,000 ÷ 936 ≈ $22.40.

Here the numerator is the whole channel budget in those markets, so $22.40 is the average cost of an order that non-brand search brings in. That’s a different number from the heavy-up’s, which measured only the added dollars. Which leads to the most common analysis mistake in this whole area.

Marginal vs average cost per incremental customer

Ad response curves bend. The first $10,000 a month in a market reaches the people most ready to come in; the fourth $10,000 reaches people who mostly won’t. So the average cost across all your spend is lower than the cost of the next dollar, often by a lot.

At $30,000 a month this market looks fine on average ($31.58) and terrible at the margin ($66.67). If your ceiling is $57.60 (we work that number out in the next section), the last $10,000 should go somewhere else. Go-dark tests give you averages; heavy-ups and spend changes give you margins. Know which one you’re holding.

What is a good cost per incremental customer?

There’s no industry benchmark worth trusting, because it depends on what a customer is worth to you. Work out a ceiling instead: the first-year contribution margin of a new customer.

For the pizza chain (illustrative): a $32 average ticket, about 6 orders in a new customer’s first year, and a 30% contribution margin give $32 × 6 × 0.30 = $57.60. Any channel bringing in new customers for less than $57.60 pays back inside a year. The measured $22.40 is comfortably inside. A dental group’s ceiling will be in the hundreds; a gym’s depends heavily on churn.

Be careful with the pizza number, though: it is cost per incremental order, not per new customer. If half of those orders came from existing customers ordering more often, the cost per new customer is about twice as high. Separate new from returning customers whenever your POS lets you.

Using it to move budget between locations

The payoff comes when you calculate it per market and lay the results side by side. Markets whose whole range sits under your ceiling can take more. Markets whose range sits above it should give some back. Markets whose range straddles the ceiling need more data before you move anything.

Our guide to allocating ad budget across locations covers how big each move should be and how to check it worked.

Common mistakes when calculating cost per incremental customer

• Using platform conversions as the denominator. That’s CPA with a new name. The whole point is to count only customers the ads caused.
• Total spend over incremental customers from a heavy-up. If you added $10,000 to a $30,000 budget, divide the $10,000 by the extra customers it produced, not the $40,000.
• Dropping the range. “$178” and “$127 to $294” lead to different decisions when your ceiling is $250.
• Revenue instead of counts. Average ticket moves with menu prices, mix and weather. Count customers, then multiply by value separately.
• A window that’s too short. A dental patient may book five weeks after seeing an ad. Cut the window at week four and you’ll overstate the cost.
• Comparing channels measured differently. A go-dark average for search against a heavy-up margin for social is apples and oranges. Compare like with like.

Questions people ask about cost per incremental customer

One line of arithmetic, one honest comparison. Get those right and most budget arguments get much shorter.

Written by The Countlift team at First Byte in The Woodlands, Texas. Questions or corrections? Email hello@countlift.com.