The Best Restaurant Promotions in Our Data: Lessons From 25 Campaigns Above 3x ROAS
See what 25 restaurant promotions at or above 3x Feast-recorded ROAS teach about offers, costs, visits, checks, and attributed revenue.
The best restaurant promotions give people a clear reason to respond, use an item or experience they already want, cost the restaurant less than their face value suggests, and stay traceable through a completed visit and settled check. In a Feast census pulled August 3, 2026, 25 campaigns across 15 restaurant organizations and 17 storefronts reached at least 3x Feast-recorded ROAS; about $34,000 in linked ad spend mapped to about $318,000 in Feast-attributed gross revenue.
Those 25 came from about 80 campaigns with measurable linked spend, so about 31% reached the 3x mark. That is useful evidence for testing restaurant promotions, not a promise that one offer or one restaurant will produce the same result.
What the 25-campaign result actually measures
The census measures Feast-attributed gross revenue divided by linked Facebook ad spend. It does not measure profit, contribution, or the sales that would have happened without the campaign.
| Census detail | What was recorded |
|---|---|
| Pull date | August 3, 2026 |
| Campaigns at or above 3x | 25 |
| Restaurant organizations | 15 |
| Storefronts | 17 |
| Campaigns with measurable linked spend | About 80 |
| Share at or above 3x | About 31% |
| Linked ad spend among the 25 | About $34,000 |
| Feast-attributed gross revenue among the 25 | About $318,000 |
| Blended recorded ROAS among the 25 | About 9.4x |
The revenue includes Feast attribution marked as indirect visits as well as gross check amounts that include tax and tip. A campaign enters the winning group when its attributed gross revenue is at least three times its linked ad spend. That comparison leaves out creative, software, staff time, and the actual cost of fulfilling the promotion, so even a recorded 3x result is not a profit claim.
The strongest examples do not prove that one offer type always wins
The named high-ROAS examples include free items, a buy-one-get-one offer, a seasonal event, and premium entrées. They show that several promotion shapes can work; they do not classify all 25 campaigns or prove that every winner used a giveaway.
| Anonymized example from the census | Feast-recorded ROAS | What it can tell you |
|---|---|---|
| Free fountain drink | 23.9x | A familiar add-on can be the offer. |
| Free noodles across three storefronts | 22.6x | A simple menu item can travel across locations. |
| Holiday pop-up | 20.5x | A timely experience can work without being a standard coupon. |
| Buy one, get one first drink | 12.5x | A purchase condition can still produce a strong recorded result. |
| Premium steak offer | 4.9x | A higher-ticket promotion can clear the same threshold. |
These are examples, not a ranking of tactics. The census does not support “free always beats a discount,” and it does not show that percentage discounts lost. The right comparison is the one you can run at your restaurant: response, verified visits, matched checks, and the actual cost of the food or experience you promised.
Start with something guests already want, then check what it costs you
A strong offer usually begins with a popular, familiar item whose perceived value is high relative to its incremental cost. In one Feast sales call, menu popularity was used as a practical proxy for perceived value because guests already recognize and order the item.
One burger restaurant owner explained why fries fit that job:
“It is our number-one selling item by three to one. So we're happy to give away free fries.”
— Restaurant owner, August 24, 2026
The owner paired the fries with a purchase and noted that fries and a higher-priced potato side had nearly the same food cost. That does not make free fries the answer for every menu. It shows the work: compare what guests want with the ingredient, packaging, processing, and added labor cost you will actually incur, then decide whether a free item, fixed-dollar offer, percentage discount, bundle, or access-based perk gives you enough room.
Write down the real terms before the ad goes live: eligible guest, included item, purchase requirement, exclusions, location, start and end dates, redemption limit, and exactly how staff will ring it. A vague “free entrée” creates a different cost and service problem from “one specified item with a qualifying purchase.”
A premium restaurant can lead with the experience and put the offer last
A promotion does not have to make the restaurant look like a coupon book. For a higher-end restaurant, lead with the room, hospitality, dish, chef, or occasion; reveal the trackable reason to act only after the guest understands why the visit is worth making.
In a February 5, 2026 fine-dining discussion, a Feast operator advised:
“Lead the marketing with the experience, hospitality, and product features, and only introduce the offer at the end as a ‘no-brainer’ to encourage trial.”
That is an offer-placement recommendation, not proof of a campaign result. Restaurants that will not discount can also test a claimable privilege such as early access to limited reservations, provided the capacity and terms are real and the response can still be connected to an arrival and check. If brand perception is the concern, use the fuller guide to promoting a restaurant without cheapening it.
Make the promotion ring the same way on every shift
A trackable promotion needs one response record and one consistent register action. If the ad, claim, arrival, and check cannot be joined, a strong offer can look unprofitable simply because the evidence broke at the register.
A multi-location restaurant team was told to use either one consistent discount code or a dedicated menu item for the offer so staff did not ring the same promotion several ways. Before launch, test the complete path:
- Submit one real test response through the live ad or landing page and confirm its campaign and offer identifiers are retained.
- Confirm the response is eligible, then let the test record follow the same claim, booking, or waitlist path a guest would see.
- Record a verified in-person completed visit. A click, claim, reservation, or prepaid purchase alone is not a visit.
- Link that visit to its own paid, non-voided settled POS check and verify the correct amount and promotion code.
- Void or refund the test check and confirm the reversed dollars leave sales reporting rather than remaining as a win.
Stop the launch if any handoff fails. Fixing it afterward cannot recreate a guest identity, visit record, or check link that was never captured.
Judge the offer only after every guest has had time to use it
Promotion rates should use eligible records whose full response or arrival window has elapsed and whose outcome has been reconciled and is known. Keep immature, ineligible, open, and unresolved records outside the denominator and report them separately.
Use one promotion-response record per unique person and offer after removing duplicate submissions. Then keep these numbers separate:
| Result | Exact definition |
|---|---|
| Response | An eligible person completed the stated claim, booking, waitlist, or purchase action. It is not a visit. |
| Verified-visit rate | Of the eligible response records whose full arrival window elapsed and whose visit outcome is reconciled and known, the share with a verified in-person completed visit. |
| Matched-check rate | Of those verified completed-visit records whose check-reconciliation window elapsed and whose check outcome is known, the share linked to their own paid, non-voided settled check. |
| Directly matched net settled sales | Paid, non-voided settled sales linked to those records, less recorded refunds. This is attributed sales, not causal lift. |
| Observed lift | The change in comparable covers, checks, new guests, or sales during the promotion versus a like-for-like baseline. A before-and-after difference alone is not incrementality. |
| Incremental result | The difference supported by a credible holdout or other defensible test design. Do not relabel ordinary attribution or observed lift as incremental. |
Prewrite the arrival window, check-reconciliation window, minimum mature sample, and keep-or-stop target before spending. There is no universal offer, budget, duration, or ROAS threshold for every restaurant. If you need a disciplined way to run more than one offer without pretending every test should win, use the three-offer testing guide.
ROAS alone does not tell you whether the promotion paid
The census ROAS uses linked ad spend as its cost denominator, so your restaurant still needs a complete ledger. Keep campaign costs and check costs visible rather than hiding them in one miscellaneous line.
Non-check campaign costs are media, creative production, influencer compensation if used, usage rights if purchased, and the campaign's allocated share of software or agency fees. Write down the allocation basis—for example, active campaign days or campaign spend—and use it consistently.
Matched-check variable costs are the food and beverage served, the actual incremental cost of the promoted item, packaging, payment processing, fulfillment, and added hourly labor tied to those checks. If the free item's ingredients or labor are already in the matched-check cost, do not count them again as a separate promotion charge. The offer's menu price is not its cost.
For a conservative campaign contribution remainder, subtract both ledgers from directly matched net settled sales. That remainder is still not incremental contribution, restaurant-wide profit, or proof that the campaign caused every matched check.
Keep, repair, or stop based on the broken step
Keep an offer only when enough mature, reconciled records show acceptable verified visits, matched checks, and full-cost economics under the target you wrote before launch. Otherwise, repair the specific break instead of swapping promotions blindly.
- Responses are weak: revisit the item, terms, creative, or audience.
- Responses are healthy but visits are weak: inspect expiration, booking friction, reminders, location details, and service capacity.
- Visits are verified but checks do not match: fix the identity join, register button, staff instruction, or POS import before judging sales.
- Matched sales look healthy but the contribution remainder is weak: change the fulfillment cost, purchase condition, menu mix, or paid spend.
- Records are still immature or unresolved: wait or reconcile them. The honest answer is not ready yet.
The 25 campaigns show that restaurant promotions can clear a meaningful recorded-sales threshold in different forms. The defensible lesson is not a secret giveaway. Build your next test around one understandable reason to act, a cost your restaurant can carry, a clean path from response to visit to check, and a decision made from mature records.
Feast produced the campaign and attribution records behind this census. If you want the same kind of connection between promotion activity and attributed restaurant sales, book a demo of Feast.