Three ROAS numbers for the same campaign

Short answer: because each tool is answering a different question, and all of them are answering it honestly. The platform reports what its own ads touched. Your analytics reports last non-direct click. A third-party tool reports its own model. You do not resolve this by finding the true number. You resolve it by picking one number to make decisions on, and demoting the rest to steering.

The end state, in someone else’s words

An agency operator put it on X in July 2026, and it drew 54 likes: “You sign up for clarity. Six months later you have three ROAS numbers for the same campaign and no idea which one to trust.”

A practitioner had already tested the claim the expensive way, paying for four attribution platforms at the same time and publishing what each returned. The numbers landed between 2.4x and 2.9x for the same spend. His conclusion was not that one was right: “None were ‘accurate.’ Attribution is modeling, not truth.”

And underneath it all sits the owner who has to make a decision anyway. One posted that he was spending $9,000 a month split between Facebook and Google with zero idea which platform makes real profit after cost of goods and shipping; Facebook showed a decent return, Google an acceptable one, and neither told him what was actually left.

Three different people, three different vantage points, one problem. The tools multiplied and the certainty went down.

Why the numbers differ, specifically

The differences are not noise. Each one has a named cause.

• Different credit rules. The platform credits itself for a view that led to a purchase days later. Your analytics credits the last non-direct click. Given one customer who saw an ad, searched your brand, and bought, both will claim it, and both are following their own stated rule.
• Different windows. A seven-day click window and a one-day view window describe different populations of the same buyers.
• Different starting data. A platform sees only its own traffic. Your analytics sees everything that reached the site but loses whatever the browser blocked. Neither has the full picture, and the missing parts are different.
• Different maths. Platform return is calculated on revenue it can see against spend it knows. Blended return is total revenue against total spend, including the channels nobody is claiming.
• Double counting. Where two tools fire the same purchase with no shared order ID, one number is simply inflated. That one is a fault rather than a difference, and it is worth ruling out before you interpret anything.

Add them up and a 2.4x against a 2.9x is not a contradiction. It is five known mechanisms producing five defensible answers.

The number to run the business on

There is one number no platform can inflate and no browser can block: total revenue against total advertising spend, over a period long enough to mean something. Your bank knows the first. Your card statements know the second.

It is a blunt instrument. It cannot tell you which creative worked. It moves for reasons that have nothing to do with advertising, like a seasonal dip or a price change. But it is the only figure denominated in money you actually received, and it is the correct number to decide budget on.

The platform numbers keep a job, and it is a real job: deciding which ad, which audience, which placement inside a channel. A platform is well placed to compare two of its own ads against each other. It is poorly placed to tell you what share of your business it deserves credit for.

So the rule is: steer with platform numbers, decide with the blended number. Most of the confusion in that opening quote comes from asking one number to do both jobs.

The trap on the other side

Having said that, blended has a failure mode people run into fast.

If you spend nothing on brand advertising and your blended return looks fantastic, part of what you are seeing is demand you already had. Customers who would have found you anyway sit in the numerator. A blended figure that never moves when you change spend is telling you that most of the revenue is not being driven by the spend at all.

The practical test is not clever, but it works: change one channel’s budget meaningfully, hold everything else still, and watch the blended number over a period longer than your sales cycle. If nothing moves, that channel was not doing what its dashboard claimed.

One practitioner reduced the whole problem to four words: pick one and stay consistent. A number you trust and track weekly beats a more sophisticated number you re-litigate every month.

Where MIQAS comes in

MIQAS builds the blended number out of your own orders rather than out of platform claims, then puts the platform claims next to it.

Orders come from the source you already have: your store, your point of sale, your delivery-app exports. Spend comes from the connected ad accounts. The ROI view sets real revenue against real spend and sorts each channel into scale, optimise or cut, so the output is a decision rather than a figure to interpret.

The Conversion Check with the Ad Claims Audit tool, on Growth and above, exists for the reconciliation itself: what the platforms claimed against what your store recorded, with the gap shown rather than hidden. That is the number that ends the argument in a meeting, because it is sourced from the business, not from the seller.

None of this makes MIQAS a more accurate oracle than Meta about Meta. It is not trying to be. It is trying to give you one number that belongs to you, alongside the platform numbers, so you stop asking five sellers to agree.

What to do this week

• Write down your blended return for last month: total revenue over total ad spend. This is your baseline, and most people have never calculated it.
• Write each platform’s claimed revenue next to it. Add them up. If the sum of the claims exceeds your actual revenue, you now know by how much.
• Rule out double counting before interpreting anything, because a duplicate is a fault.
• Pick your decision number and say out loud which one it is, so next month’s meeting is about the business rather than about whose dashboard is right.

Three numbers for one campaign is not a sign that your setup is broken. It is a sign that you are asking three sellers to grade themselves. The fix is not a fourth opinion. It is one number of your own.