How Long Should a Restaurant Ad Run? What Day 14 Can—and Can't—Tell You
Day 14 is a restaurant ad checkpoint, not an automatic stop date. Follow each response through a paid check before deciding.
For the Feast-managed restaurant ad test documented here—about $20 per day for 14 days—Day 14 is a checkpoint, not an industry rule or an automatic stop date. By then, you can often find the first broken step between ad delivery and a paid check, but you cannot judge an event, reservation, or offer whose visit window is still open.
Feast's internal campaign practice, documented in July 2026, uses a two-week test stage before deciding whether to put more money behind an ad. That timing does not guarantee a result, and it does not apply to Feast or Remy's own customer-acquisition ads. The useful question is whether your restaurant has enough mature evidence to keep, repair, stop, or extend the test.
That is also why the strongest restaurant advertising examples connect ads to visits, not just to clicks.
| What Day 14 may tell you | What it cannot tell you by itself |
|---|---|
| Whether the ad received delivery for the money spent | Whether the test spent enough for a fair decision |
| Whether people clicked or took the trackable action | Whether a click, signup, claim, or reservation became a visit |
| Whether a mature response could be matched to an identified guest, completed visit, and paid check | What will happen to responses whose booking or redemption window remains open |
| Whether directly matched net settled sales covered a stated set of costs | Whether the ad caused incremental sales or produced profit |
| Which step failed first | Whether changing several parts at once would fix it |
Start the clock when the ad starts delivering, not when someone builds it
Start the test clock when the approved campaign begins delivering. A campaign sitting in draft, waiting for approval, paused, or unable to spend has not completed the test just because the calendar moved.
Write down the launch timestamp, intended audience, offer, trackable action, visit deadline, test-spend limit, and last date a response can still produce a valid visit. If the campaign starts late or barely delivers, keep the original result labeled incomplete rather than quietly moving the start date.
The offer controls when the result matures. An ad for tonight's walk-in special may produce known outcomes quickly. An event ad with future reservations cannot be judged on visits before the event happens, and a pass with an open redemption window cannot have a final visit rate while guests can still use it.
Set the decision date from the last promised guest action, not from impatience. Day 14 can still expose a broken ad or form, but the final visit and check decision waits until every eligible person has had the full stated window and each outcome is reconciled.
Decide what “worked” means before the first dollar goes out
A restaurant ad worked only when it reached the finish line you chose before launch. If you can estimate incrementality defensibly, the business finish line is incremental contribution greater than the full campaign cost; otherwise, report directly matched checks as an observed result and do not call them incremental profit.
Keep the math on two separate ledgers:
- Directly matched net settled sales are the paid, non-voided check sales after refunds linked from an identified response through a verified visit. They are stronger evidence than a click, but the match does not prove the ad caused the sale.
- Contribution associated with those matched checks is matched net settled sales minus the food, packaging, processing, fulfillment, and incremental labor costs required to produce those checks.
- Full campaign cost includes media, creative production, influencer compensation, usage rights, software or agency fees, and incentive fulfillment when each applies. Allocate any shared charge by a written rule and count every cost once.
- Incremental profit estimate starts with sales you can defensibly estimate would not have happened without the ad, subtracts the variable check costs, then subtracts the full campaign cost.
If a discount is already reflected in net settled sales, do not subtract its face value again. Count the actual incremental food or fulfillment cost of an incentive once, and keep a face-value view separate if you use one for planning.
Before launch, test the full path yourself. Take the real ad action, create an identified guest record, verify the completed in-person visit, pay a non-voided check, and confirm that check first appears in the campaign's matched-sales report. Then void or refund the test check and confirm that it is removed.
Follow the first broken step instead of staring at one return number
The first failing step tells you what to repair. Platform activity, campaign responses, claims, visits, matched checks, observed lift, incremental sales, contribution, and profit are different levels of evidence; compressing them into one return number hides the problem.
| Step | Count only this | If it fails first |
|---|---|---|
| Delivery | Spend, impressions, and whether the intended local audience actually received the ad | Check campaign status, audience, placement, budget, and delivery before judging the offer |
| Click or response | The exact click, reply, form submission, or other action named before launch | Repair the opening, food, offer, or call to action; delivery alone is not interest |
| Signup | A unique identified person who completed the required signup | Test the page, fields, message, and handoff; do not count page visits as signups |
| Claim or reservation | A unique valid claim or confirmed reservation | Check confirmation and reminders; neither state proves arrival |
| Verified visit | A unique identified person whose arrival was recorded in person | Check the deadline, staff handoff, guest instructions, and no-show follow-up |
| Paid settled check | A verified visit connected to a paid, non-voided check | Repair the identity-to-POS match; an unmatched check cannot be credited to the campaign |
| Directly matched net settled sales | Net settled check sales after voids and refunds | Compare the observed sales with their variable costs and the full campaign cost, without calling the result causal |
| Incremental contribution | A defensible incremental-sales estimate minus applicable variable and campaign costs | Keep spending only when the estimate clears the restaurant's written requirement |
Use the same eligible group in each rate. For visit conversion, the denominator is unique identified people whose entire visit window has elapsed and whose outcome is known; the numerator is the subset of those same people who completed a verified in-person visit. Keep open, immature, unmatched, and unknown records as separate counts.
Early arrivals do not enter that rate before their cohort matures. Excluding both early successes and open outcomes until the same window closes prevents a numerator from getting ahead of its denominator.
Low spend can make Day 14 too early, but it cannot make failure disappear
Spend changes how much evidence the checkpoint contains. Compare actual spend and delivery with the test limit you wrote before launch; there is no universal restaurant budget that makes every ad conclusive on the same date.
If the ad barely spent, say the test is underdelivered. You may extend it with a new fixed end date, but record the original spend and result so an extension does not become an excuse to leave an ad running forever. Build the limit backward from check contribution rather than borrowing a generic daily budget from another restaurant's marketing budget.
If the ad spent as planned and the mature chain stops at clicks, repair the trackable action. If signups arrived but visits did not, repair confirmation, timing, and the staff handoff. If verified visits produced paid checks but the full costs overwhelm the contribution, changing the headline will not fix the economics.
Audience waste belongs in the same review. When an ad meant to acquire new guests keeps reaching known regulars, the campaign can look active while buying visits you likely already had; a member exclusion audience can make the next test cleaner.
Keep, repair, stop, or extend the ad for one stated reason
Keep the ad at its current spend when the eligible outcome window has closed, the guest-to-check path works, and the result clears the requirement you wrote before launch. Increase spending only in a separate decision, because a larger budget can change the audience and the result.
Repair the ad when one stage clearly fails while the earlier stages work. Change that stage, start a new labeled test, and leave the old result intact so you know what actually changed.
Stop the ad when it reaches the written spend limit, the full guest-action window has elapsed, outcomes are reconciled, and the mature evidence misses the stated business requirement. Stop sooner when the offer is wrong, tracking is broken beyond repair, or the restaurant should not honor the promise in the ad.
Extend only when the result is not mature or the campaign did not receive the delivery needed for the planned test. Write down what remains open and the new decision date; “needs more time” without either item is not a decision.
Day 14 should end with a decision, not a dashboard screenshot
Your checkpoint should say what happened, where the chain first broke, what remains unresolved, and what you will do next. It should never turn clicks into guests, reservations into visits, attributed net settled sales into incremental revenue, or revenue into profit.
Feast can run restaurant paid ads and connect the campaign response to identified guests, completed visits, and supported POS check data when those records exist. You still decide the offer, the acceptable economics, and whether the evidence earns more spend.