Why your ROAS is lying to you (and what to track instead)
Return on ad spend is the most quoted number in paid media and one of the least useful. Here is what it hides, and the three metrics we put on client dashboards instead.
Every paid media report opens with ROAS, and almost every one of them overstates the truth. Not because anyone is being dishonest, but because the number is computed inside the ad platform that has an interest in the answer.
The attribution window problem
A platform reporting a 6x return is counting conversions it believes it caused. Change the attribution window from 7 days to 28 and the same campaign suddenly returns 9x, with no change to the campaign itself. Nothing improved. The measuring stick got longer.
Worse, two platforms will each claim the same conversion. Add up the ROAS your channels report and you will frequently find they have collectively sold more than you actually sold.
It ignores what the customer is worth
ROAS is revenue divided by spend on the first transaction. A campaign delivering 3x from customers who reorder quarterly is worth several times one delivering 5x from people who buy once and vanish, but the dashboard ranks them the other way round.
If a metric cannot distinguish a loyal customer from a discount hunter, it cannot be the metric you optimise against.
What we track instead
- Blended CAC — total marketing spend divided by total new customers, across every channel. It cannot be double-counted because it never attributes.
- Contribution margin per cohort — what a month's customers actually returned after product, shipping and payment costs.
- Payback period — how many weeks until a cohort has repaid what it cost to acquire. This is the number your cash flow cares about.
Keep ROAS as a directional signal for deciding which creative to kill this week. Just do not let it decide the budget.
How to make the switch
Start with blended CAC, because it needs no new tooling — you already know total spend and total new customers. Track it weekly for a quarter alongside your platform ROAS. When the two disagree, and they will, you will quickly learn which one predicted the bank balance.
