You might not expect it, but across multiple businesses I've worked with, ROAS targets are set based on what they think the average order value is. Not based on what a customer delivers in total. If you shift the philosophy to account for what a customer can bring in on average — beyond that first purchase — you can afford a higher acquisition cost.
A customer who buys three times is worth three times more than their first order suggests. If you set your acquisition budget based on that first purchase alone, you're structurally underpaying for customers who will stay with you for years. And leaving acquisition opportunities on the table that you could actually afford.
This article is about how to understand the real value of a customer, why most businesses measure it incorrectly, and what that means for your ROAS targets and acquisition budget.
The calculation most businesses make
The standard approach goes like this. You know your average order value. You know the margin on it. You determine an acceptable acquisition cost based on that first transaction. And you set your Google Ads or Meta campaigns to a ROAS target that reflects it.
Some businesses go a step further. They look at the number of returning customers and try to build a multiplier from that. If 40 percent of our customers come back, we can pay more for a new one. That's the right way of thinking. But how that 40 percent is calculated changes everything.
Repeat share is not the same as repeat rate
This is where it most often goes wrong in practice. Repeat share is the proportion of your total orders that comes from returning customers. If you have 100 orders today and 40 come from people who have bought before, your repeat share is 40 percent.
Repeat rate is fundamentally different. It tells you how many of the customers you acquired in a given period actually came back. Of all the new customers you won three years ago, how many have placed another order since?
"Repeat share tells you how dependent today's revenue is on returning customers. Repeat rate tells you how loyal your past customers actually were."
The difference sounds subtle. The consequences are not. Picture this: 5 returning customers versus 10 new ones. Repeat share: 33 percent. The following year: 6 returning customers versus 20 new ones. Repeat share drops to 23 percent. Internal alarm about declining loyalty. While in absolute terms you have more returning customers than ever and acquisition is exploding. A declining repeat share alongside rising absolute volumes isn't a problem. It's exactly what you want to see.
| Year 1 | Year 2 | |
|---|---|---|
| Repeaters | 5 | 6 ↑ |
| New customers | 10 | 20 |
| Repeat share | 33% | 23% ↓ |
What determines the real value of a customer
The basic formula for customer lifetime value is straightforward: average order value, multiplied by the number of purchases per year, multiplied by the number of years a customer remains active. A customer who spends an average of €120, buys twice a year, and stays for three years is worth €720. Not €120.
That sounds logical. But most ROAS targets are built on that €120, not on that €720. The result is that you're structurally underpaying to win that customer, while they're actually worth six times more than you see at that first moment.
Where it gets really interesting: you can flip that logic around. If you know how many customers you won last year, and you know the repeat rate of comparable cohorts from the past, you can predict how many of those customers will buy again this year. Multiply that by your average order value and you have a grounded revenue forecast based on your existing customer base, without acquiring a single new customer.
The difference from using a global average repeat rate is that you look per cohort. Customers you won three years ago behave differently from customers you won six months ago. Whoever doesn't make that distinction is steering on an average that doesn't accurately reflect any single segment.
"Your existing customer base is a predictable revenue source. The question is whether you have the data to make that prediction."
What getting CLTV right changes
The implications of a correct CLTV are clear. The acquisition cost for a new customer can be higher. If you know a customer buys an average of three times over two years, you can pay more for that first purchase than someone who only looks at the margin on that first order. That's not a higher cost. It's a better investment.
Your ROAS target needs to reflect that. A ROAS target based on the first purchase alone is structurally too conservative for customers with a high repeat rate.
At one client where I worked through this, the ROAS target could be adjusted from over 700 percent to around 400 percent within an attribution model with a 60-day window. That sounds like a reduction. In reality it was an expansion: more could be invested in acquisition than the business had thought possible for years.
And if you know you're building loyalty and customer value is growing over time, you can scale the acquisition cost further. A rising repeat rate justifies a higher investment in new customers, because you know that investment pays itself back across multiple purchases.
"The most expensive mistake in acquisition is underpaying for a customer who stays with you for years."
In closing
The ROAS you measure is almost always the ROAS on the first purchase. That's the number your reporting shows, the number your targets are built on, and the number that determines how much you're willing to pay for a new customer. But the real ROAS — measured across the full lifecycle of a customer — is structurally higher for everyone who comes back.
That has a concrete implication. If you know your customers buy an average of three times over two years, then the ROAS on that first purchase is not the benchmark for your acquisition budget. You can set your ROAS target lower. Or your CPA higher. Not as a gamble, but as the logical conclusion of what your customers actually deliver.
The question is not whether you're allowed to raise your acquisition cost. The question is whether you have the data to justify it.