How We Decide an Ad Is Dead: About $135 Median, Nothing Past $580 in Feast's Account
In Feast's own account, first tracked purchases appeared by $579. Learn how to separate immature, unresolved, and genuinely dead restaurant ads.
In Feast's own acquisition ad account, median spend to the first tracked purchase was about $135 when the account was reviewed on July 22, 2026. Among ad sets that went on to record at least five purchases, the first purchase appeared after $77 to $579 in spend; the source contained no first purchase after $580.
That is why Feast treated zero-purchase spend below about $300 as usually too early and roughly $300 to $600 as the decision window. This is an internal Feast account rule, not a universal restaurant benchmark, a guarantee that a later sale is impossible, or a Remy result. The unit is an ad set, not each video inside it. The source summary does not provide the number of ad sets, the full observation window, or its attribution settings, so use the numbers as a transparent account-specific example rather than a forecast for your restaurant. A broader set of restaurant ad examples and their customer-tracking steps can help you choose the right finish line.
A restaurant owner described the real fear after a nine-day test with zero sales: spending another $1,500 to $2,000 could put the total near $3,000 without reaching break-even. The answer is not to wait forever. It is to decide what counts as a sale, how much evidence the test needs, and how much you are prepared to lose before the first dollar is spent.
A dead ad is different from an immature or unresolved test
An ad is genuinely failing only after the promised customer action has had enough spend and time to happen, the outcome records are closed, and the result still misses your written economic limit. Before that, you may have an immature test or a measurement problem.
| State | What you actually know | What to do |
|---|---|---|
| Immature | The ad set has not reached its written spend checkpoint, or the visit and check windows are still open | Keep the original test intact until the next checkpoint; do not count an early success yet |
| Unresolved | People responded, but arrival or check outcomes are still open, unknown, or cannot be joined to the response | Repair the customer path and close the records; do not judge the ad |
| Failing at one step | The test is mature and one step consistently breaks, such as response, arrival, check matching, or contribution | Repair that step and run one clean retest |
| Dead under your rule | The mature test reached its prewritten loss ceiling and still produced no acceptable customer result or contribution | Stop it rather than moving the line after seeing the result |
The Feast account pattern used cumulative media spend to a first tracked purchase. It did not measure full restaurant acquisition cost, an in-person visit, incremental revenue, contribution, or profit. Those are different questions, and your stopping rule needs the event that matters to your restaurant.
Decide what a sale means before the ad starts
For a restaurant campaign, a click, form submission, claimed offer, and reservation are responses, not completed visits or paid sales. A useful customer path is: ad, identifiable response, verified in-person arrival, and a linked paid, non-voided settled check.
Write the test card before launch:
- Test unit: one ad set with a fixed offer, audience, destination, and measurement window.
- First customer event: the exact response you can join to one eligible person.
- Restaurant result: a verified in-person visit, recorded separately from the response.
- Sales result: a paid, non-voided settled check linked after the visit.
- Spend checkpoint: the cumulative amount at which you will review the test.
- Loss ceiling: the complete campaign investment you are willing to risk before the result becomes unacceptable.
- One permitted repair: the first broken step you will change before one final clean retest.
A calendar date can be a useful checkpoint, but it cannot replace spend or a fully elapsed customer window. What Day 14 can and cannot tell you depends on whether the ad actually spent and whether the promised customer outcome had time to close.
Prove the response-to-check path before paying for a real test
Run one acceptance test from the response through the settled check before launch. Create a known response, record the test person's in-person arrival, and confirm that the correct paid, non-voided check appears under the campaign with the expected amount. Then void or refund that test transaction and confirm its check count and sales dollars disappear from the settled result.
Once the campaign is live, keep the two rates separate:
- Response-to-visit conversion: the denominator is eligible identified responders whose full fixed visit window elapsed and whose visit outcome was reconciled and known. The numerator is the subset of those same people who completed a verified in-person visit.
- Check-match coverage: the denominator is eligible verified completed visits whose full fixed check-reconciliation window elapsed and whose matched-or-unmatched outcome is known. The numerator is the subset with a linked paid, non-voided settled check.
Hold early successes until their window closes. Report open, immature, ineligible, and unresolved records separately at both stages; keep known unmatched checks separate too. That prevents a fast responder from entering the success count while a slower non-visitor remains outside the denominator.
Directly matched sales are attributed sales, not proof that the ad caused them. A like-for-like change in covers or net sales is an observed change. A valid control or holdout can support an incremental estimate. Profit requires the rest of the restaurant's costs as well.
Your loss ceiling needs every campaign cost once
Media spend alone can tell you when Feast's own ad sets first recorded a purchase, but it cannot tell you whether a restaurant campaign made money. Count the complete investment before comparing the result with your stopping rule.
Use one campaign ledger for paid media, creative production, applicable influencer compensation, separately priced commercial rights, any separate influencer-account advertising permission, an optional bonus only when the content actually runs in paid advertising, allocated software or agency fees, and paid campaign-management labor. Allocate shared charges using one written rule, such as the test's share of active campaign days. Keep owner time as a separate valued-time line unless it is a cash expense.
Use a different, non-overlapping transaction ledger for the ordinary variable food, beverage, packaging, processing, fulfillment, and incremental service-labor costs of the matched checks. Put the actual incremental cost of any offer or reward in that transaction ledger once. If net settled sales already reflect the discount, do not subtract the discount's face value again.
Then calculate two different answers:
Campaign cost per verified first-time guest = complete campaign investment / unique eligible identified first-time guests with a verified completed visit in the mature, reconciled-known group
Directly matched contribution = matched net settled sales after discounts, refunds, and voids − transaction variable costs − non-overlapping campaign investment
The first is an acquisition-cost ratio. The second is a directly matched remainder, not causal profit. If you cannot connect the response to a verified visit and settled check, leave those answers unknown rather than filling the gap with platform-reported clicks.
Kill or repair the first broken step
The right decision follows the first point where a mature test fails, not whichever number looks worst on the ad screen.
- The ad set did not get the planned spend or delivery. The offer has not had the test you wrote. Check setup and eligibility before judging the video.
- The platform reported delivery, but the mature test produced no identifiable responses. Repair the opening, offer, audience, or destination. Rising costs do not prove ad fatigue is the only explanation.
- People responded but did not arrive. Repair the offer instructions, timing, reminders, location details, or restaurant capacity.
- Verified visitors did not link to settled checks. Repair identity capture and checkout reconciliation before blaming the ad.
- Matched checks appeared but the full-cost remainder was unacceptable. Stop or change the economics. More clicks cannot rescue a contribution problem.
- The result cleared the written limit. Keep the working version intact. Raise spend in controlled steps and repeat the same full-cost check rather than assuming the result will hold forever.
Change one important thing at a time. If you change the offer, audience, video, destination, and budget together, the next result cannot tell you which repair mattered.
More influencer footage only solves a creative problem
If the offer, customer path, and economics work but you cannot keep producing new videos, build the next set yourself or use local influencers under written terms. More footage will not repair a broken visit path or turn attributed sales into incremental profit.
Remy manages the influencer-program work: finding influencers, booking, texts, reminders, follow-up, and file collection. You keep the decisions about whom to approve, what to serve, which files are useful, and whether any video deserves a customer-ad test.
Remy is $150 per location per month, with about $5 a day planned separately for influencer-recruitment ads. That recruitment spend is not the customer-ad budget discussed above.
If repeated influencer coordination is the part keeping you from making the next test, Learn more about Remy. Blog readers can get the first month for $100 through that link; the standard price is stated above.