Client reporting: blended ROAS or platform ROAS in the monthly deck?

Clients keep asking for one ROAS number, and I go back and forth on which one goes in the deck. Platform ROAS from Google and Meta adds up to more revenue than the shop actually booked, because both platforms claim the same order. Blended ROAS (total revenue divided by total ad spend) avoids the double count, but it hides which channel is doing the work, so a weak campaign can sit behind a strong one.

Take a skincare brand spending 40k across Meta and Google. It reports 6x blended, while the platforms together claim 9x. For the budget split, the owner shouldn't rely on either number alone.

My view is to lead with blended as the headline and show the platform numbers underneath, labeled as claimed. How do you handle it when the client only reads the top line?

Keep blended as the headline, but don't treat the gap between 6x and 9x as a footnote. It's the most useful number in the deck. Roughly a third of the revenue the platforms claim is being counted more than once, and showing that makes it obvious why adding their ROAS together misleads.

I wouldn't use blended for the budget split, though. It moves when organic or repeat sales move, so a strong month for email can make paid look better than it is. For the split, look at each channel next to a simple test, like a geo holdout or pausing one campaign for two weeks, so the owner sees what actually disappears when the spend does.

A client who only reads the top line will be fine as long as the deck says in one sentence what blended can't answer.

Which decision is the client making with this number: the total budget, or how to split it between Meta and Google?

Agreed on the holdout. For skincare, blended is only as clean as the revenue line under it. Platforms report gross order value at checkout, so cancellations, returns and discounts inflate their numbers while the shop's net revenue comes in lower. I'd calculate blended from net revenue, then split it into new-customer and returning-customer revenue. Paid spend mostly buys the first group, so new-customer ROAS tells you whether the budget is actually growing the business or just collecting credit for orders that would have happened anyway.

Agreed on net revenue. For the new-customer split, I wouldn't take "new customer" from the ad platforms at all. Google and Meta each use their own lookback window, so the same buyer can show up as new in one and returning in the other. Pull the first-order flag from the shop's own customer records, matched on email or customer ID, and compare that count with the new-customer conversions the platforms claim. The gap tells you how much of the paid credit is going to people who had never bought before.

The first-order flag from the shop's own records is the right call. Before you trust the split, though, look at the email match rate. Guest checkouts and changed emails can push a chunk of repeat buyers into the "new" bucket, which quietly inflates paid's new-customer share.

Honestly, it depends on how many orders go through guest checkout. If the shop requires an account, email matching is close to complete and the split holds up. If guest orders are a big share, check the match rate month by month before you show the client a new-customer number.

Agreed, and the lookback on the shop side is the bigger trap. If the first-order flag only covers the reporting month, anyone who bought in March shows up as new in April, so paid's new-customer share climbs for no real reason. Pull order history back to the shop's first order, and match guest checkouts on phone as well as email, since phone catches a lot of people who used a different address at checkout.

Phone only helps once it's normalized. The same customer can show up as +90 5xx..., 05xx... or 5xx... depending on the checkout form, so a raw string match misses a lot of the repeat buyers you're trying to catch. Convert everything to E.164 before matching, and treat numbers shared by several customers, like a household line or a gift recipient's number, as weak matches. Then report how many matches came from email and how many came only from phone, because that shows how much the phone step actually changed the split.

I'd disagree with calling shared numbers weak matches. A weak match still merges two different people into one customer, and that's exactly the error that flips the split. If a household line ties a genuinely new buyer to an old customer, the new buyer gets counted as a repeat, and paid's new-customer share drops for no real reason.

The fix is to pull any number that shows up on more than one customer record out of the phone key entirely, then count what's left. You can still report the phone-only count, but only after hand-checking a sample of those pairs.