The Number On Your Ads Dashboard Is Marking Its Own Homework

I have sat on both sides of this table. Agency side, where I built the reports that made the platforms look like heroes and the retainer look cheap. Client side, where I watched a leadership team renew a seven-figure budget off a return-on-ad-spend figure that Meta had calculated, reported and graded, all by itself. Nobody in that room had checked the working. Nobody in that room could have.

Here is the thing. The number on your ads dashboard is the platform marking its own homework. It is an attributed number, not an incremental one, and the gap between those two words is where Australian marketing budgets quietly bleed out.

That is the position I will put my name to. Meta and Google report their own performance, they have every commercial incentive to make that number look strong, and most businesses allocate capital on it with zero independent verification. The fix is not a better dashboard. The fix is turning things off and finding out what would have happened anyway.

Australian advertisers send roughly 70% of every digital dollar to Meta and Google, the same platforms that measure and report their own performance. There is no independent umpire checking whether that spend actually worked.

Attributed means the platform drew a line from an ad to a sale and took the credit. Incremental means the sale would not have happened without the ad. Those are different questions. The dashboard only answers the first one, and it answers it generously, because it is the one being paid.

Australian advertisers send roughly 70% of every digital dollar to Meta and Google. That is a heavier concentration than most comparable markets. Two platforms take the lion's share of the spend, report their own results and set the benchmark you judge them against. There is no independent umpire on the field.

This is not a small pool of money. The Australian internet advertising market just hit a record $4.9 billion in a single quarter, up 15.3% year on year, with search alone taking around $2.16 billion of it. More money, more pressure to move fast, more businesses optimising hard toward a figure the scorekeeper controls.

You do not have to take my word for it. Somebody already ran the test, at enormous scale, and paid for it in real revenue.

eBay turned paid search off. Not in a slide deck. In the real world, across 68 US markets, they went dark and measured what happened to sales. The study, published in 2014 by economists Blake, Nosko and Tadelis, found that brand keyword ads produced no measurable short-term benefit at all. When the paid ads stopped, almost all of that traffic simply came back through free channels. People who were going to buy anyway just clicked a different link to get there.

The dashboard never knew. It happily attributed every one of those sales to the ad, right up until the ad was gone and the sales stayed. The true return was a fraction of the non-experimental estimate the platform had been reporting.

Read that back with your own account in mind. A meaningful slice of what you are paying for is demand you already had. You are paying a toll to intercept your own customers on their way to the checkout, then thanking the toll booth for the traffic.

Cricket solved a version of this decades ago. Rain interrupts a match, the conditions change, and you are left staring at a scoreboard that no longer tells you who is actually winning. So they built the Duckworth-Lewis method. A system that isolates the variables and gives you the true result out of messy, incomplete conditions.

Marketing has the same problem and almost none of the discipline. Your sales are the product of a hundred things happening at once. Brand, season, word of mouth, a competitor stumbling, the ad. The platform looks at that mess, points at itself and says "that was me". No control. No isolation of variables. Just a confident claim with a commercial motive behind it.

Incrementality testing is the Duckworth-Lewis method for your budget. You hold a group back, you keep everything else the same, and you measure the actual delta between the version of the world with the ad and the version without it. That number is the truth. The dashboard number is a hope with good graphics.

Marketers know something is off, so a lot of them are reaching for marketing mix modelling. The two tools everyone is adopting are Google's Meridian and Meta's Robyn. Both are free. Both are open source. Both are built by the exact platforms being measured.

Think about what that means. You suspected the referee was biased, so you accepted a free whistle from the referee. The tool might be perfectly good. That is not the point. The point is who built the ruler. When the company whose spend you are trying to justify hands you the instrument that justifies it, you have not removed the conflict. You have dressed it up in open-source clothing and invited it back in.

We have scored and benchmarked the marketing health of Australian businesses across a wide range of industries, on six dimensions. There is a pattern I would stake the firm on.

The weakest dimension, again and again, is the data and tracking layer. Not the ads. Not the creative. The ability to answer a basic commercial question about your own money.

Most businesses cannot tell you what would have happened anyway, because they have never built the infrastructure to ask. They can run the campaign. The doing is fine. What they cannot do is separate the sales the ad caused from the sales the ad took credit for. They have a shopfront with a thousand people walking through it, a till that counts the money, and no idea which of those buyers they actually paid to bring in. That is not a reporting gap. That is flying blind with a full tank of fuel.

If you run marketing, or you sign off on it, here is where I would put my attention before I renewed another dollar.

Turn something off. Pick your biggest platform line and go dark in a few markets, or hold back a slice of your audience, for long enough to read the result. This is the whole game. If you switch it off and revenue drops, brilliant, now you know that spend is real and you scale it with confidence. If you switch it off and nothing changes, then the question is obvious. Why were you spending money there at all? Either answer makes you richer. There is no version of this test that loses.

Hold out a geo. You do not need to bet the account. Keep one comparable region running as your control and change nothing in it. Watch what the ad-supported regions do against it. That gap, region against region, is your incremental lift. It is the closest thing to a clean read you will get outside a lab.

Budget for testing, not just media. Most Australian marketing budgets do not have a testing line at all, which is exactly why they keep paying tolls they cannot see. eBay spent real revenue to learn the truth. Put a small reserve in your back pocket for holdouts before you spend the rest optimising toward a number you have never verified.

Stop treating the dashboard as ground truth. Treat it as a claim from an interested party. Useful, directional, worth reading. Not evidence. The moment you demote that number from fact to opinion, you start asking the right questions.

The money is already moving. Retailers and banks are building measurement platforms around the incrementality question, because they have no dashboard to protect and they have to prove causation, not report correlation. The advantage is going to the businesses willing to turn things off and find out.

The rest of the market is still opening the dashboard on a Monday, reading the number the platform wrote about itself, and calling it performance. That is not measurement. It is trust dressed up as data. Numbers on the board are the lifeblood. A number the scorekeeper made up about its own game is just noise with a nice chart on top.

If you want an outside read on where your own measurement setup actually stands, run a Lens scorecard and see how your Data and Tracking dimension compares to your industry.

What is the difference between attributed and incremental conversions?

Attributed means the platform drew a line from an ad to a sale and took the credit. Incremental means the sale would not have happened without the ad. A dashboard only ever answers the first question, and it is the platform being paid that answers it.

How can a business test incrementality without a data science team?

Turn a channel off in a few markets or hold back a slice of the audience, then wait long enough to read the result. Compare that group against a comparable control region that keeps running as normal. The gap between the two is your real, incremental lift.

Are Google's Meridian and Meta's Robyn independent measurement tools?

No. Both are marketing mix modelling tools built and open sourced by the same platforms whose spend they are used to justify. They may be genuinely useful. They do not remove the underlying conflict of interest.