An ad budget, a booking fee and a discount are not comparable until you put them against the extra customers each route caused. Ads spend money on clicks whether or not anyone visits. Booking networks charge for diners who book through them. A discount reduces contribution on every redemption, including those by regulars. The useful comparison is cost per attributed visit against full-price contribution—then a test of how many of those visits were genuinely additional.

What does each route actually charge for?

Google Business Profile is free to claim and can show an offer, though making the profile useful still takes staff time. Meta and Google Ads use auctions, so there is no dependable UAE cost-per-click figure to borrow for your venue. OpenTable's UAE pricing includes a monthly subscription and a fee for each seated diner who comes through its network; bookings through the restaurant's own website have no per-diner fee. Delivery platforms may also supply ordering, payment and fulfilment, so their charge is not simply an advertising fee. Contract terms, not an industry average, belong in your calculation.

A bank-programme discount is different again. It can bring an offer to eligible cardholders, but who funds it is deal-specific. Never book an assumed bank contribution in the merchant's economics until the agreement confirms it.

How many new visits would cover the cost?

Take an illustrative Dubai restaurant with 100 attributed two-person dine-in visits. Each visit brings AED 200 before VAT and AED 120 contribution before the discovery cost. Assume its ad budget is AED 2,000 or its merchant-funded bank offer is 20% off. For the booking-network example, use OpenTable's published UAE network fee for 200 seated diners, converted and rounded to AED. The monthly subscription is excluded and must be added by a merchant using it.

The underlying threshold is route cost ÷ AED 120 contribution per visit; the table rounds up to a whole visit because a fraction of a visit cannot pay the bill. A 20% discount on an AED 200 bill costs AED 40 on every redemption. If only 20 of the 100 bank-offer visits were additional, they would add AED 2,400 of full-price-equivalent contribution against AED 4,000 in markdowns: AED 1,600 lost. The ad and booking rows cannot be declared better merely because their thresholds are lower. They reach different customers, include different services and may have fixed fees.

Why isn't an attributed customer necessarily new?

A customer may click an ad for a restaurant she already planned to visit, or book a familiar venue through a platform. The booking and the click are real; the incremental visit may not be. In a US field experiment at eBay, observed search-ad returns differed sharply from experimental results. A study of Facebook experiments likewise found that observational estimates often missed experimental lift. Neither study provides a UAE restaurant conversion rate. They explain why channel dashboards cannot settle causation on their own.

Turn one route off in comparable weeks or areas, and compare total visits and contribution, not its reported conversions. A small venue's result will be noisy; run the test long enough to cover ordinary weekly variation and report a range rather than a precise ROI.

Decision rule: For each route, calculate the share of attributed visits that must be additional. Keep paying only when a holdout suggests the actual share is higher after subscriptions, fulfilment, discount funding and staff costs are included. Do not add percentages from different channels together and call them new customers.