I spent thirteen years on the other side of this table, running agencies for the Saudi market. I have written the kind of deck that puts “AI” on the cover. So when a proposal leans on that word, I usually know what it is covering for. Choose a performance marketing agency for its measurement discipline, its operating team, and its ability to connect advertising to a real business outcome. The “AI” in the pitch is not the reason to sign. In the GCC, the right partner must also understand market differences, bilingual execution, lead handling, and the distance between platform conversions and collected revenue.
Your job as the buyer is narrow. Test whether the agency can diagnose, execute, measure, and hand back knowledge you keep. This guide gives you the questions, and the evidence, to demand before you sign.
What you should actually be buying
A strong performance engagement should cover five responsibilities:
- Commercial diagnosis: clarify the offer, margin, capacity, sales process, and target outcome.
- Channel execution: plan and run campaigns, creative tests, landing pages, and audience strategy.
- Conversion infrastructure: protect tracking, CRM routing, lead response, and order-status feedback.
- Business measurement: reconcile ad-platform reporting with qualified leads, paid enrolments, delivered orders, or collected revenue.
- Decision cadence: explain what changed, what was learned, and what happens next.
Media buying without the other four may generate activity while leaving the commercial problem untouched.
What “AI-driven” should mean
Useful AI does the boring work faster. It drafts creative variants you still have to judge, chews through audience and query data, runs bid and budget automation inside the ad platforms, flags anomalies before a human would catch them, and routes leads so nobody waits two hours for a first reply. Research and quality-assurance checks ride along with it. None of that is the headline most decks are trying to sell you.
The line I watch for is who makes the decision. AI should never be the thing an agency hides behind. Ask which actions run on their own, what data sets them off, who signs off, and how you reverse one when it goes wrong. “Our AI optimises everything” is a slogan, not an operating model.
The GCC requirements many proposals miss
Market-specific economics
UAE, Saudi Arabia, Qatar, and neighbouring markets differ in audience scale, competition, purchase behaviour, and sales operations. A regional plan should show where one campaign structure can be shared and where it must diverge.
Budgets should follow demand and competition, not a flat regional split. In practice: UAE Search and Shopping run in one of the more expensive CPC environments in the region, so account structure and negative-keyword discipline matter more than raw spend. Saudi Arabia needs bilingual creative and Arabic-audience budget parity, not an English-first campaign with an Arabic afterthought. Qatar has a smaller top-of-funnel population, which usually rewards tighter geo-targeting and a longer testing runway before scaling spend. If a proposal runs UAE, KSA, and Qatar off one shared account structure without separating budget, currency, and creative by market, ask why. For what execution actually looks like market by market, see Google Ads for Dubai and the UAE, performance marketing for UAE businesses, Google Ads management in Saudi Arabia, and Google Ads for Qatar.
Arabic as an operating requirement
Arabic work needs independent copy review, not translation approval. Ask who writes and reviews it, which register they use, and how search terms and sales objections differ by market.
Lead quality and response
For service businesses, a cheap lead has little value if it is outside the market, cannot afford the offer, or waits hours for a reply. Require reporting on qualification, contact rate, appointment rate, and sales outcome where data is available.
Collected revenue
For e-commerce and enrolment businesses, orders and bookings are not automatically revenue. Returns, failed COD deliveries, cancellations, and unpaid reservations must feed back into reporting.
Tracking stack
Before any AI layer earns trust, the tracking stack underneath it has to be sound. Ask what the agency actually builds: server-side or Conversions API tagging alongside the on-page pixel, GA4 events mapped to real business actions rather than page views, a GTM container you can audit yourself, and a CRM or order-management connection that can push confirmed outcomes back into the ad platforms. Automated bidding inside Performance Max, Advantage+, or standard Search campaigns is only as good as the signal feeding it. A proposal that never mentions server-side tracking, CAPI, or a CRM connection is asking you to trust a black box built on top of an already-noisy one.
Offline revenue
Most of the revenue in a services or COD business happens off the ad platform: a phone call, a CRM stage change, a delivery confirmation days or weeks later. A competent agency can import that outcome back into Google Ads' offline conversion import or Meta's Conversions API, on a schedule that matches your sales cycle, so the bidding algorithm eventually learns from collected revenue rather than clicks. Ask which conversion action is being optimised for the offline-imported event, how often the upload runs, and what happens to the bid strategy when that pipeline breaks for a week, because it will eventually break.
The two-number measurement test
This is the test most agencies quietly hope you skip. The platform number is the one printed on the case study. The collected number is the one that pays your team. When those two drift apart, the story drifts with them, and the agency would rather you never went looking. Ask every agency to show:
- the platform-attributed result;
- the business-confirmed result;
- the reason for the gap;
- the decision made because of that gap.
In an anonymised COD commerce case, the platform view showed 4.1× gross ROAS while delivered ROAS was 1.9×, with a 33% COD return rate. Both numbers are useful; reporting only the first would support the wrong scaling decision.
In the FIT Institute education engagement, 121,330 AED in ad spend produced approximately 912,550 AED in collected revenue, about 7.5× clean ROAS. Because the outcome is collected educational revenue rather than a physical order exposed to returns, the clean number is the correct headline. See the FIT case study.
The Performance Max, Meta, and CRM feedback loop
Performance Max and Meta's Advantage+ campaigns both hand more of the targeting and bidding decision to the platform's own model. That trade only pays off if the platform is optimising toward the right signal. Ask the agency to walk through the loop end to end: which conversion action feeds PMax, which feeds Advantage+, how CRM stage changes such as qualified lead, appointment held, or closed-won map back to those conversion actions, and how often the loop actually closes in practice. An agency that runs PMax and Advantage+ off last-click platform conversions with no CRM feedback is letting the algorithm optimise for the wrong outcome: cheap leads and cheap add-to-carts, not qualified pipeline or delivered revenue. For a closer look at where PMax reporting and real ROAS diverge across GCC markets, see Google Ads and Performance Max ROAS: the real GCC number.
How to audit an agency proposal in one working session
Put the proposal, media plan, sample report, and contract on one call with the people who will actually run the account. Then work through four passes:
- Outcome pass: underline every promised result. Mark whether it is a platform metric, sales metric, or collected-revenue metric.
- Ownership pass: circle every account, audience, pixel, landing page, creative file, automation, and report. Write who owns it after termination.
- Operating pass: ask who does the weekly work, what AI touches, what requires approval, and what happens when the automation is wrong.
- Measurement pass: require a sample two-number report using your current funnel, even if the numbers are placeholders.
End the session with three lists: approved, unclear, and unacceptable. If the "unclear" list is longer than the approved list, do not sign yet. The goal is not to catch the agency out; it is to find whether they can operate in the real business, not just inside the ad account.
How to request a two-number report
Send this request before signing:
For each campaign, please report:
1. platform-attributed conversions and revenue;
2. CRM/order/enrolment-confirmed conversions and revenue;
3. the gap between the two;
4. the operational reason for the gap when known;
5. the decision made because of the confirmed number.For COD, the confirmed number may be delivered or collected revenue. For education, it may be paid enrolments or collected tuition. For services, it may be qualified opportunities, appointments held, or closed revenue. The exact event changes by business; the discipline does not.
A 12-question agency scorecard
Ask these questions in writing and score each answer from 0 to 2.
- What business outcome will you optimise?
- Which source will confirm that outcome?
- How will you report platform and confirmed results?
- Who will work on the account each week?
- What is automated, and what requires human approval?
- How do you test creative without diluting the brand?
- How do Arabic and English workflows differ?
- How will lead quality or delivery status return to the ad system?
- What must our internal team provide?
- Who owns accounts, data, audiences, creative, and workflows?
- What happens in the first 30, 60, and 90 days?
- How can either side exit without losing access or operational knowledge?
Interpretation:
- 20–24: strong operational clarity; verify references and contract terms.
- 14–19: potentially suitable, but close the weak areas before signing.
- 0–13: the proposal depends too heavily on activity, tools, or vague promises.
Red flags in the proposal
- guaranteed ROAS without access to margins, sales data, or fulfilment;
- case studies with spend and revenue but no definition of revenue;
- an impressive senior pitch followed by an unnamed delivery team;
- dashboards that cannot connect to CRM or order status;
- bundled software you cannot retain after termination;
- Arabic treated as a translation step;
- long lock-ins before tracking and baseline quality are proven;
- “AI” used as a reason to avoid explaining decisions.
What a fair first phase looks like
The first phase should reduce uncertainty, not maximise spend. A sensible sequence is:
- audit tracking, economics, offer, and sales or fulfilment flow;
- define the confirmed conversion event;
- establish a baseline;
- launch a limited set of tests;
- reconcile outcomes;
- scale only what survives reconciliation.
If the agency wants to increase budget before it can explain the current gap, pause.
Agency, consultant, or in-house?
An agency is strongest when you need sustained execution bandwidth across channels and creative. A consultant is stronger when the main constraint is senior diagnosis, system design, or an asset your team will own. In-house is stronger when marketing is core, volume is stable, and you can support leadership and specialist roles.
For the broader comparison, read AI consultant vs agency vs in-house. For the service methodology, see AI performance marketing.
The 30-day paid media audit
If you already have live campaigns and are not ready to switch agencies, a 30-day audit is the lower-commitment way to get the same picture. In four weeks: audit tracking and the platform-to-collected gap on your current spend, map how PMax, Advantage+, and any CRM feedback loop are actually wired, flag the market-specific gaps if you run across UAE, Saudi Arabia, and Qatar, and hand back the 12-question scorecard above with your current agency's answers filled in. You keep the findings regardless of what you decide next.
Next step
If you are reviewing an agency proposal, running live campaigns with an unclear revenue picture, or replacing a weak reporting model, book a paid media revenue audit. We can identify the measurement gap, the tracking gap, and the operating model you should buy — market by market across the GCC.