For about a decade, performance marketing had a comfortable deal: click, cookie, conversion, credit. Every riyal could be traced to an outcome, and the dashboard told you what to do next.
That deal is over. App tracking permissions, browser restrictions on third-party cookies, and stricter consent expectations have each removed a piece of the chain. What is left is a measurement environment that looks broken if you keep asking it the old questions.
The advertisers doing well in the GCC right now are not the ones who found a workaround. They are the ones who changed the question.
What actually broke
Three things, and it helps to keep them separate:
- Cross-site identity. Platforms can no longer reliably follow a person from an ad to a purchase days later on another device. Conversions still happen; the platform just cannot always claim them.
- Reporting latency and modelling. Much of what a dashboard now reports is estimated rather than observed. That is not dishonest, but it means two platforms will both claim the same sale.
- Audience targeting precision. Small, tightly defined remarketing pools have thinned out. The platforms compensate with broad targeting and their own modelling.
Notice what did not break: people still see ads, still click, and still buy. Only the accounting changed.
Rebuild measurement from the business outwards
The instinct is to fix attribution. The better move is to stop depending on it for decisions it was never good at anyway.
Start with a number the platforms cannot touch
Total enquiries or orders, and total spend, in the same period. Divide one by the other. That blended cost per acquisition is unglamorous, ignores channel credit entirely, and is the only figure in your reporting that cannot be inflated by a platform’s self-attribution.
Track it weekly. If blended CPA is falling while you spend more, the programme is working — regardless of what any individual dashboard claims.
Then add server-side conversion tracking
Sending conversions from your own server, with hashed first-party identifiers where you have consent, restores a meaningful share of the signal that browser-side pixels lost. It is a real project — not a checkbox — but it is the highest-return technical work in paid media right now.
Two things to get right: consent must be genuine and recorded, and the data you send must match what your CRM believes. A conversion feed that disagrees with your own sales records will optimise toward the wrong customers.
Use holdouts when the number matters
If you need to know whether a channel is actually contributing, turn it off in one region for two weeks and watch total demand. Geo holdouts are blunt, slightly uncomfortable, and far more honest than any attribution model. In a market like the Kingdom, where regional differences are real, run the test across comparable cities rather than one against the rest.
Ask the buyer directly
The most underrated measurement tool is a single optional field on the enquiry form: How did you hear about us?
The answers are messy, self-reported, and biased toward whatever the person remembers last. They are also the only data source that captures the WhatsApp forward, the friend’s recommendation, and the Instagram story nobody clicked. Across a few hundred responses, patterns emerge that no platform reports.
We now put this on every client form. It regularly reveals channels the dashboards were undercounting badly.
Where the leverage moved
When targeting precision falls, creative carries the weight. The platforms’ models are now good at finding the right person if you give them something that clearly signals who it is for. Which means:
- Volume beats polish. Several concepts tested per month, each with variations, beats one expensive hero film per quarter.
- Vertical, sound-off, first-three-seconds. Still true, still ignored.
- Arabic and English creative, made separately. A subtitled English ad is not an Arabic ad. In the Gulf the difference in performance is usually large enough to be embarrassing.
- Say the offer plainly. Vague brand-mood creative performs badly when the model is relying on your creative to identify the audience.
- Kill on evidence, not on taste. The ad the team likes least is often the one that works.
Broad targeting, tight creative
The pattern that works now is close to the opposite of 2019. Give the platform a wide audience and let the creative do the qualifying. Layer on exclusions rather than narrow inclusions. Keep the account structure simple — fewer campaigns, more budget each, faster learning.
Landing pages are still where money is lost
None of the above matters if the click lands on a slow page that asks for eleven fields. In the Gulf, where a large share of traffic is mobile and often on a variable connection, the gap between a two-second and a five-second load is the difference between a viable and an unviable campaign.
Before raising a budget, check: does the page load fast on a real phone, does it say the same thing the ad said, does the form ask only for what you need, and is there a WhatsApp option? That last one converts far better than a form for a lot of Saudi audiences, and plenty of advertisers still do not offer it.
What good looks like now
- A blended CPA you watch weekly and trust more than any platform’s number.
- Server-side conversions feeding the platforms clean, consented data.
- A creative pipeline producing new concepts continuously, in both languages.
- Broad targeting, simple account structure, decisions made on two weeks of data rather than two days.
- An occasional holdout test when a real budget decision depends on the answer.
- A “how did you hear about us” field, read monthly.
It is less precise than the old model and considerably more honest. Advertisers who make the shift stop arguing with dashboards and start arguing about the business, which is the more useful argument.
This is how we run every account in our performance marketing service, usually alongside the creative production that feeds it. If your reporting stopped making sense, show us the account — we will tell you what we would change first.