Restaurant retargeting: the 2026 numbers that break the myth

The 70.19% who leave without ordering: the number that defines what retargeting is for

Restaurant retargeting does not win new customers, it rescues the ones who already looked and left without ordering, and the number that justifies it is the 70.19% cart abandonment rate Baymard Institute reports in its 2025 review of 49 digital commerce studies. Picture the real funnel of your online menu: out of every ten people who add a dish to the order, seven close the tab before paying, almost always because of surprise delivery fees, forced registration or a menu that loads slowly. That group already declared purchase intent while hungry with a card in hand, so speaking to them again costs a fraction of what it takes to convince a stranger. That owner in Medellín who spent 1,860 USD in two months to bring in 43 orders at 43 USD each, with an average ticket of 27, did not have a retargeting problem: he had a reach campaign in disguise.

How much audience do you need before switching on the first ad?

You need a thousand identifiable people as the floor, and below that mark the platform abandons your audience and expands into generic interests, which is exactly where cost per order becomes indefensible.

Meta and Google ad accounts demand minimum thresholds to activate a custom audience, and a single-location restaurant that only tracks menu views usually takes four to eight weeks to gather them. The cheapest input for filling that tank is already produced daily: vertical video of the kitchen, of service, of the hand plating a dish. A fifteen-second view runs between 3 and 6 USD per thousand across most markets, while a click to the menu costs fifty or sixty cents. Sequence matters here: fill the tank first with cheap video, then flip the switch on the conversion ad. Reverse that order and you burn budget training a blind algorithm. Difference number one is not the platform, it is the event you measure, because a restaurant that sends the purchase with the real ticket value trains the algorithm with money, and one that only sends ViewContent trains it with curiosity.

The event you send decides which money trains the algorithm

Curiosity does not cover payroll, least of all when input costs sit 35% above 2019 levels in both food and labor, according to the National Restaurant Association in its 2024 profitability analysis. Under that cost pressure, every ad dollar has to arrive with cash attribution, not with vanity metrics. In practice that means firing Purchase with value and currency from your ordering platform, separating delivery orders from table reservations, and keeping tips outside the value so the return is not inflated. A dashboard showing 400 AddToCart events and 11 purchases is telling you where the money leaked, not how many people felt interested. Splitting the 7-day window from the 14-day one into two separate ad sets lowers cost per order without touching the creative, and segmenting at 180 days does the opposite even though the audience count looks bigger on screen. Somebody who opened your menu in February has already eaten out eighty times elsewhere before your ad reaches them in August; that impression costs the same and never converts.

Seven and fourteen-day windows: why 180 days makes each order more expensive

The logic is temperature, not volume: whoever abandoned a cart the day before yesterday responds to a reminder featuring the exact dish they left behind, while somebody who ordered five weeks ago responds to a fresh reason to return, maybe a seasonal dish or a slot with a shorter wait. Email works as a mirror here: Omnisend measured a 25.1% average open rate in 2023 and Stripo documents 26% higher opens with personalized messages, the same relevance mechanic in another channel. Budget by temperature and the order price drops on its own. I pushed too many owners to produce more ads when the problem sat at the destination, and that mistake cost months of somebody else's budget before I admitted it. If the link lands on a 4 MB PDF menu that forces pinch-zoom on a phone, no creative rescues the conversion: the user arrived with intent and you greeted them with a print file.

I got this wrong for years: the destination weighs more than the ad

The fix is boring and it works: an HTML menu, prices visible without opening anything, an order button above the fold, and the very dish the person abandoned as the first card on the page. At Masterestaurant we always measure retargeting against the destination, never against the ad alone, because the 33% of operators naming customer attraction and retention as their top challenge in 2026, per Toast, rarely have a traffic problem. They have a friction problem in the final thirty seconds of the order. Doubling spend on a broken funnel multiplies the loss by two, and the thread is worth following to the end because almost every owner does exactly that when the dashboard cools off. With 1,860 USD and 43 orders at a 27 USD ticket, the Medellín campaign returned 1,161 USD in gross sales, which at a 65% contribution margin leaves roughly 755 USD of real contribution against 1,860 in spend.

What happens if you double the budget without fixing the event or the destination?

Push the budget to 3,720 and the algorithm, forced to spend faster, goes out to buy pricier impressions in colder audiences: cost per order does not stay at 43, it climbs.

Fix the purchase event and serve the menu as HTML, though, and those same 1,860 USD usually move cost per order into the 8 to 14 USD range, which finally fits inside a 27 USD ticket. Order of operations rules: repair the meter, repair the destination, then add money. Paid retargeting performs better when it shares the work with owned channels, even though it looks like they compete for the same customer, and that tension resolves once you assign who does what. Paid finds the person who left without handing over data; email and text messaging work with whoever already handed it over. Omnisend reports a 25.1% average email open rate for 2023, Stripo documents 26% higher opens with personalization, and Tabular measured 25% more SMS engagement in food and beverage during 2025.

The cheap-channel paradox: email and SMS hold up your paid clicks

Those three numbers say the same thing: an owned message costs almost nothing and converts at a pace no auction matches. That is why the pixel is not the main asset, it is the bridge. Every order arriving through retargeting must leave with an email or a phone number captured, because from there the second order is one you bring in for free. Skip the capture and you rent your database from the platform forever. Tattoo 70.19%, 1,000 and 14. The first is the cart abandonment rate Baymard Institute documents in 2025, and your concrete action is to build today an exclusive ad set for AddToCart without Purchase in the last 72 hours, with the abandoned dish in the creative and the delivery fee stated in the opening line. The second is the floor of identifiable audience: publish vertical kitchen and service video until you cross a thousand people before switching on a single conversion campaign, and check the progress weekly in your audience panel.

The 3 numbers you should tattoo on yourself

The third is the maximum window in days for your warm set; split retargeting into 7 and 14 days with separate budgets and shut down anything past 30 unless a specific seasonal promotion demands it. Close with this: open your ordering platform this week and confirm the Purchase event travels with real value and currency. If it does not travel, everything else is decoration. Difference number one is not the platform, it is the event you measure. A restaurant sending Meta a purchase event with the real ticket value trains the algorithm with money; one sending only ViewContent trains it with curiosity, and curiosity does not cover payroll. The second difference is the window. Segmenting at 180 days sounds like more audience and behaves like more noise: someone who saw your menu in February has eaten elsewhere eighty times since. Split 7-day and 14-day windows into separate ad sets and you pay differently for each temperature, which lowers cost per order without touching the creative.

Where spending and investing actually part ways?

Third, and I got this wrong for years: I pushed owners to produce more ads when the real problem sat at the destination.

If the link lands on a 4 MB PDF menu that forces a phone call, delivery conversion collapses and you blame the ad. Two-tap checkout first, budget second. Fourth, retargeting amplifies whatever online reputation you already carry. A venue at 4.6 stars with 300 reviews converts the same impressions 30% to 40% better than one at 4.0, because guests open Google before they pay. Paying to remind people of a weakly reviewed restaurant is paying to be turned down.