Attribution is not measurement

Attribution is not measurement. Attribution assigns credit. It looks at the path a customer took and decides which ad, click or view gets the trophy. Measurement answers a different and harder question: what would have happened anyway? Would this customer have bought without the ad? Attribution cannot answer that. It was never designed to. For a UAE business spending AED 30,000 a month across Meta, Google and TikTok, confusing the two is expensive. You end up increasing budget on channels that claim credit for demand they did not create. Only measurement should move budget. This essay explains the difference, why platforms blur it, and what decision-grade measurement looks like at SME scale.

Key takeaways

• Attribution assigns credit to a touchpoint. Measurement estimates what would have happened anyway.
• Platforms grade their own homework. Last-click and view-through attribution inflate claimed revenue.
• Add up the revenue Meta, Google and TikTok claim and you will often get more than 100 percent of your actual sales.
• WhatsApp and phone conversions, the lifeblood of UAE SME sales, are largely invisible to platform attribution.
• Decision-grade measurement at SME scale means blended CAC, holdout thinking, geo splits and honest questions about any reported ROAS.

What attribution actually does

Attribution is a bookkeeping system. It records which touchpoints a customer passed on the way to a conversion and distributes credit among them according to a rule. Last click gives everything to the final touchpoint. First click gives everything to the opener. Data-driven models spread the credit across the path using statistical weights.

All of these rules share one structural limit. They observe correlation: this person saw the ad, then bought. They never observe the counterfactual, the technical term for what would have happened in a world where the ad did not run. A customer who searched your brand name, clicked your ad and bought would probably have clicked the organic result one position lower and bought anyway. Last-click credits the ad with the full sale. Nothing in that number tells you the ad caused anything.

Platforms grade their own homework

Every major ad platform measures its own performance using its own attribution windows, then presents the result as evidence you should spend more. Meta counts conversions for up to 7 days after a click and 1 day after a view by default. Google and TikTok run similar windows. View-through means the person did not even click. They scrolled past your ad, then bought something, possibly for a completely unrelated reason, and the platform claims a conversion.

Here is what this looks like in practice. A Dubai home and living brand spends AED 45,000 in a month. Meta reports AED 40,000 in attributed revenue. Google Ads reports AED 35,000. TikTok reports AED 12,000. The dashboards claim AED 87,000. The store's total revenue that month, all channels, all walk-ins included, was AED 55,000. The platforms are not lying exactly. Each one genuinely saw the customer. Many customers were counted twice or three times, and many of them were going to buy regardless. Summing platform dashboards is not measurement. It is arithmetic on overlapping claims.

Fixing your data plumbing helps but does not solve this. Server-side tracking and a clean CRM integration deduplicate events and recover signal lost to iOS restrictions. They make attribution more accurate. They do not make it measurement. If you are not sure whether your tracking is even recording reality, start with our guide to tracking marketing ROI for UAE businesses.

Measurement asks the harder question

Measurement, in the decision-grade sense, estimates incrementality: the portion of your results that would not have happened without the spend. The gold standard is the controlled experiment. You show the ads to one group, withhold them from a comparable group, and compare outcomes. The difference is what your ads actually caused.

This is not theory. In a widely cited field experiment, economists Thomas Blake, Chris Nosko and Steven Tadelis ran tests with eBay and published the results in 2014. eBay paused its paid ads on brand keywords, searches that already contained the word eBay, across a large share of US regions. Almost all of the traffic that had arrived through paid brand clicks simply came back through free organic links. The ads were harvesting clicks from people who were already on their way. When the researchers looked at non-brand keywords, they found weak average returns for a company as well known as eBay. The touchpoint that attribution credited with the sale was often not the cause of the sale at all.

Your business is not eBay, and paid search genuinely creates demand for brands nobody has heard of yet. That is precisely the point. The value of a channel varies enormously, and the only way to know your number is to test what happens when the channel is not there. The dashboard will not tell you. It cannot.

The UAE problem: half your conversions never appear

Everything above is made worse by how people actually buy in the UAE. The dominant conversion path for a huge share of SMEs looks like this: someone sees your Instagram ad, screenshots the product, sends it to a family WhatsApp group, then messages your business on WhatsApp or calls. The deal closes in a showroom in Al Quoz, a clinic in Jumeirah, or over a bank transfer after a voice note exchange in Arabic and English.

Platform attribution sees none of this. It sees a click with no conversion attached and reports the ad as a failure, or it credits a branded Google search at the end of the chain and reports search as a hero. For restaurants, clinics, real estate brokerages, business setup consultancies and most B2B services in Dubai, the phone call or the WhatsApp thread is the conversion. Last-click attribution records the last tap on a screen, not the conversation that actually sold.

The practical consequence is that your reported ROAS is wrong in both directions at once. It overstates credit on channels that close the loop online and understates channels that start conversations which close offline. Budget moved on that number is budget moved on fiction. This is where a large share of the roughly 25 percent of marketing budget that companies waste tends to hide.

What decision-grade measurement looks like for an SME

You do not need an econometrics team. You need disciplined arithmetic, a small number of honest metrics, and the occasional deliberate experiment. This is the measurement layer we build in our growth intelligence work, and you can run a basic version yourself.

• Blended CAC. Total sales and marketing spend for the month divided by total new customers acquired that month. No channel splitting, no attribution windows, just money out and customers in. If platforms report improving performance while blended CAC stays flat, the platforms are taking credit, not creating demand.
• Blended ROAS, sometimes called MER. Total revenue divided by total ad spend. For an AED 45,000 month producing AED 180,000 in revenue, that is 4.0x, regardless of what any single dashboard claims.
• Holdout thinking. Pause a channel, a campaign type or a keyword set for a defined period and watch total revenue and total new customers, not the dashboard. If nothing moves, you have learned something worth more than any report.
• Geo splits. Run a channel in one emirate and hold it out in a comparable one for four to six weeks, then compare revenue movement. Crude, imperfect, and still far more truthful than a 7-day click window.
• Seasonality discipline. Ramadan, Eid, the summer exodus and Dubai's event calendar distort every comparison. Compare like periods with like periods, or any experiment you run will measure the calendar instead of the channel.

None of this is glamorous. It is slow, repetitive and occasionally uncomfortable, because it usually reveals that one channel everyone loves is doing less than its dashboard says. That is the job. Measurement exists to make the next budget decision cheaper, not to make the last one look clever.

Honest questions to ask of any reported ROAS

Whether the report comes from a platform, an agency or your own team, ask these before you act on it:

1. What attribution window produced this number, and how many of these conversions are view-through rather than click-through?
2. Is this revenue deduplicated against other channels and against sales that would have come organically?
3. What does blended CAC and blended ROAS say over the same period?
4. What is the evidence these customers would not have converted anyway?
5. What happened the last time this channel was paused or reduced?
6. Which conversions in this report are WhatsApp chats, phone calls or walk-ins, and how were they counted?

A report that cannot answer these questions is not measurement. It is bookkeeping with a marketing budget attached.

Frequently asked questions

What is the difference between attribution and measurement in marketing?

Attribution assigns credit for a conversion to a touchpoint using rules like last click. Measurement estimates causation: what would have happened without the spend. Attribution is bookkeeping. Measurement, through methods like incrementality testing and holdouts, is what should guide budget decisions.

Why does my total platform-reported revenue exceed my actual revenue?

Each platform attributes conversions independently using its own click and view windows. A customer who saw a TikTok ad, clicked a Meta ad and searched on Google gets counted by all three. Add the dashboards and the claimed total routinely exceeds your real revenue by 30 to 60 percent.

Is last-click attribution ever useful?

Yes, as a diagnostic. It shows which touchpoints close customers who already decided to buy, which helps with creative and landing page decisions. It is not useful for budget allocation, because it systematically overcredits bottom-funnel channels and brand search while ignoring what created the demand.

How can a small business measure incrementality without a big budget?

Track blended CAC and blended ROAS monthly, run simple holdout tests by pausing one channel for four to six weeks, and use geo splits where feasible. None of this requires expensive software. It requires the discipline to compare what happened against what would have happened anyway.

Does iOS tracking loss make attribution worse for UAE businesses?

Yes. Apple privacy changes removed a large share of the signals Meta and others relied on, so platforms now model and estimate more conversions than they observe. That makes reported numbers even less trustworthy as proof of causation, and makes blended metrics and holdout tests more important, not less.

The bottom line

Attribution tells you which touchpoint got the credit. Measurement tells you what would have happened anyway. The first is a story each platform tells about itself, graded by itself. The second is the only basis on which a budget should move. Get your blended CAC on one spreadsheet, ask the six honest questions of every ROAS report you receive, and run one holdout test this quarter. Boring, yes. Also the difference between spending and investing.

Want to know which of your channels actually create demand? Book a diagnostic call. No pitch deck, no pressure. We look at your numbers and tell you plainly what is measurement and what is bookkeeping.