Nearly three-quarters of restaurant traffic never comes through the door

Nearly 75% of all restaurant traffic is now off-premises, and for 58% of limited-service operators and 41% of full-service operators it is a larger share of sales than it was in 2019. That is not a pandemic residue working its way out of the numbers; it is the category's new shape, six years on. Most restaurant marketing still treats it as a secondary channel: the website is built around the room, the photography is built around the table, and the ordering surface is a link in the navigation. Meanwhile 65% of limited-service operators offer delivery and nearly six in ten Gen Z and millennial diners use takeout or a drive-thru at least weekly.

The dining room stopped being the denominator six years ago, and the website is still built as though a table is the thing being sold.

1. 01

   Off-premises is structural, and the operator data says so both ways

   58% of limited-service and 41% of full-service operators report it as a larger share of sales than 2019. Those two figures being different is the useful part: a full-service operator is running a genuinely dual business with two cost structures, and averaging them into one marketing plan produces a plan that fits neither.

2. 02

   The ordering surface is the storefront

   57% of adults have recently used mobile ordering, rising to 74% of millennials and 65% of Gen Z. For those cohorts the first impression of the restaurant is a menu screen with a photograph and a wait estimate, not a dining room. Every argument that would normally be made about a restaurant's frontage now applies to that screen.

3. 03

   Loyalty is a drive-thru mechanic before it is a dining-room one

   65% of drive-thru users say a loyalty programme influences where they order. That is a meaningfully different claim from loyalty influencing whether somebody returns eventually; it is loyalty operating at the moment of a repeated, low-consideration, in-car decision, which is the highest-frequency decision the category has.

4. 04

   Search converts at 8.05% here, and the lead is not the transaction

   A $2.05 click and a $30.57 cost per lead sit against a conversion rate well above the all-industry median, because the intent behind a restaurant search is unusually immediate. The trap is that the conversion event is usually a click to an ordering surface owned by somebody else, at which point the measurement stops and the commission starts.

5. 05

   The Meta figure quoted in this category is a median off a very thin sample

   The published social benchmark shows restaurants at a $3.16 cost per lead and an 18.25% conversion rate, roughly a tenth of the search cost. That median sits on as few as two campaigns in the underlying set. It is not a planning number. We cite it only with that caveat attached and never on its own, and any plan built on it should be treated as untested.

**~75% of restaurant traffic is off-premises** (takeout, drive-thru, delivery). 58% of limited-service and 41% of full-service operators say off-premises is a bigger share of sales than in 2019.

The model builds outward from the ordering surface rather than from the website. Phase one instruments where orders actually originate — profile, search, marketplace app, direct link — and separates first-party from intermediated volume, because those two have different margins and are usually reported as one. Phase two rebuilds the local surfaces that carry immediate intent: the profile, the menu, the hours, the pickup instructions. Phase three runs search against off-premises intent specifically, with the conversion event set at the order rather than at the click through to a marketplace, and social used for occasion and daypart demand rather than for lead capture.

1. 1

   Weeks 1–2

   Order-origin instrumentation

   Order volume separated by originating surface, first-party split from intermediated

   Owner: OmniFlow

2. 2

   Weeks 2–5

   Local surfaces

   Profile, menu, hours, pickup instructions and photography rebuilt for off-premises intent

   Owner: OmniFlow

3. 3

   Weeks 4–7

   First-party ordering path

   Direct ordering route made the default in every owned surface, conversion event set at the order

   Owner: OmniFlow + client ops

4. 4

   Weeks 6–10

   Search and social

   Search against off-premises and daypart intent; social used for occasion demand, not lead capture

   Owner: OmniFlow

5. 5

   Week 12

   Review

   Orders by origin, first-party share, cost per order rather than cost per lead

   Owner: OmniFlow

The model reports cost per order and the first-party share of orders, not cost per lead. At the published search benchmarks a restaurant pays $2.05 a click and $30.57 per recorded lead, and if that lead is a handoff to a marketplace the real cost of the order includes a commission the media report never sees. Separating first-party from intermediated volume is the first deliverable for that reason. The social figures in this category are not stable enough to plan against and are reported as observed results only. Every benchmark input is replaced by the operator's own order data from week three, reported per location rather than per brand.

Modelled. Inputs: National Restaurant Association, 2025 Off-Premises Restaurant Trends Report; WordStream by LocaliQ, Google Ads Benchmarks 2026 (13,474 US search campaigns, April 2025 – March 2026); WordStream by LocaliQ, Facebook Ads Benchmarks 2025 (data April 2024 – June 2025). Modelled outputs are not a forecast or a guarantee of results. Cited: National Restaurant Association *2025 Off-Premises Restaurant Trends*.

Every figure below is published research, not a client result. They are the inputs to the arithmetic above, listed so it can be checked rather than taken on trust. The bracketed number points to the full citation at the end of this page.

These are the surfaces this engagement is run and measured from, shown with representative figures built around the benchmarks cited on this page. Every account we run reports into views like these, and you keep ownership of all of them.

Each figure on this page, the system it is read from, and the definition and window it is measured over.

Not applicable — this is a modelled engagement. Its weakest input is that the off-premises figures are operator and consumer self-report rather than transaction data, and self-reported channel behaviour in food consistently overstates intention. The search benchmark is firmer but pools quick-service and full-service restaurants, whose economics differ more than almost any other pairing inside a single benchmark category. The instrumentation phase replaces both with the operator's own order data inside a fortnight.

A modelled capability, not a client account. Figures illustrate what the model produces and are labelled as modelled wherever they appear. See all our work

Full citations for everything cited on this page, with the sample and period each study covers, so you can go and read the original.