Seven Meta Ads defaults every restaurant should switch off
Switch off seven Meta Ads defaults that can blur a restaurant's offer, audience, placement, destination, creative, and measurement.
For a clean local Facebook and Instagram test, switch off multi-advertiser ads, browser add-ons, automatic sitelinks, Audience Network, automatic recommendations, audience expansion you did not choose, and automatic translation. Keep one on only when it has a written job and a controlled test shows that it improves the restaurant outcome you actually care about—not merely Meta activity.
These are baseline Facebook ads settings for restaurants, not a claim that automation is always bad. The aim is to make one offer, one audience, one destination, and one result easy to trace before you give Meta more room to make choices. For campaign designs that use the same finish line, see restaurant advertising examples you can track from ad to check.
Meta changes labels and rolls out features unevenly. This checklist reflects a Feast account review from March 20, 2026; current Meta material checked September 2, 2026, uses names such as Advantage+ placements and Advantage+ creative and Advantage+ audience. If your account does not show the exact wording below, use the durable question beside it: what is Meta allowed to change without your approval?
Write down the campaign you mean to launch
The fastest way to catch a harmful default is to compare the saved draft with a one-page release card. Before opening Ads Manager, write down:
- the restaurant and location;
- one offer, including dates and exclusions;
- one call to action and its exact destination;
- the local area you intend to reach;
- the approved text, image or video, and language;
- the campaign code that will follow a response;
- what counts as a verified in-person completed visit; and
- how that visit will be reconciled to its own paid, non-voided settled check.
Give each location its own event source instead of teaching one pixel with unrelated traffic from several restaurant concepts. A restaurant group in Feast's records had ten brands sharing one Meta pixel; the ads were paused while the account and location-level tracking were repaired. That is an observed setup problem, not proof that a new pixel by itself improves sales. The fuller setup is covered in one offer, one pixel.
Now duplicate the campaign into a draft and use a practical build order: placements and audience first, then ad presentation and automated changes. Do not assume duplication preserved your choices: Feast's campaign team has seen Meta turn options back on when an ad is duplicated.
Switch off the seven defaults in build order
Each change below should leave a recorded value on the release card. Names and locations in Ads Manager can vary by objective, account, and rollout, so use your account's help or search and inspect any Advantage+ or enhancement controls your build exposes rather than hunting for an identical screen.
1. Remove Audience Network from the first placement test
Choose manual placements and leave Audience Network out of the clean first test. Record the Facebook and Instagram placements you deliberately kept.
Meta's current placement guide says Advantage+ placements can distribute across Facebook, Messenger, Instagram, and Audience Network, while manual placements let an advertiser select placements with checkboxes. Meta recommends Advantage+ placements; that is a platform recommendation, not evidence that every placement produces restaurant visits for your location. Meta for Business, checked September 2, 2026
Keep it on when: you are running a named placement test and have verified which placement-level fields your own account can export. If the export cannot isolate Audience Network, the test cannot tell you whether that placement helped. Whatever the export contains, trace eligible campaign-coded responses through to verified visits and settled checks; do not reject or keep a placement based only on low-cost clicks.
2. Remove audience expansion you did not choose
Keep the restaurant's location boundary. If your account shows the phrase “Similar Audiences,” remove that option for the baseline; otherwise remove an extra lookalike or Advantage+ audience suggestion unless it is the audience you meant to test. Record the location and every audience input that remains.
The wording is especially unstable here. Meta Blueprint's current training, checked September 2, 2026, teaches new, custom, lookalike, and Advantage+ audiences; it does not use “Similar Audiences” as the main public course label. Meta Blueprint That is why the durable rule is more useful than a screenshot: do not let an optional suggestion quietly replace the local audience named on your release card.
Keep it on when: the location still behaves as the required boundary, the added audience is written down as the test, and the resulting guests—not just responses—can be compared with the baseline.
3. Turn off multi-advertiser ads
Turn off multi-advertiser ads for the baseline, then preview the ad and record whether the restaurant appears by itself or inside a unit with ads from other businesses. Also check whether the image or video is resized or cropped. Those are the plain-English presentation changes you need to inspect; do not assume the option renders the same way in every account or placement.
Keep it on when: the preview or live test shows an acceptable multi-business presentation, the restaurant's text and creative remain intact, and you deliberately test that version against the same offer, window, audience, and finish line. A lower platform cost per response is not enough if fewer eligible responses become completed visits.
4. Set browser add-ons to None
Set browser add-ons to None if your account exposes that control. Then use the preview to verify that no extra button, contact option, or destination created a second response path; the saved ad should have the one action on the release card.
Keep one on when: that add-on is the campaign's intended action and it carries the campaign identity into the restaurant's response, visit, and check records. If a phone button creates calls that cannot be tied back to a visit, it may be useful for guests while remaining unusable for this measurement test.
5. Delete automatic sitelinks
Remove sitelinks that send people away from the offer page. Then open the ad preview and confirm that the main button, image, headline, and any visible link all lead where the release card says they should.
Keep sitelinks on when: the campaign intentionally offers more than one path, every link is reviewed, and each path retains its own campaign identity. If you cannot tell which path produced the response, the extra convenience weakened the test.
6. Turn off automatic translation until a person checks it
Turn off automatic translation for the baseline unless a fluent person has approved the translated offer, dates, exclusions, location, and button text. The translated landing page must make the same promise as the ad.
Meta said in March 2026 that voiceover translation was becoming available to more advertisers and that a simplified Ads Manager flow could translate creative elements in one step. That confirms the direction of the tool, not that every account has the same translation controls or language coverage. About Meta, March 2026
Keep it on when: each served language has approved copy, matching destination content, and someone assigned to check the live rendering.
7. Decline automatic recommendations as a bundle
Do not apply Meta's suggestions in one batch. Read each recommendation, compare it with the release card, and accept only the change you can name and test.
This is not a rule to ignore everything Meta recommends. It is a rule against changing placement, audience, creative, destination, and budget at once, then having no idea which change mattered.
Keep a recommendation when: it preserves the campaign's intended offer and finish line, or you make it the only planned change in the next test.
Prove the complete path before spending normally
The settings are done only when the assembled campaign passes one positive test and one cleanup test. Use the actual ad preview or live test path available in your account, then:
- Confirm every visible placement shows the approved offer, language, button, and destination.
- Submit one test response and confirm its campaign code appears in the restaurant's response record.
- Carry that same record through a marked test visit and its own paid, non-voided settled check. Confirm the check appears under the correct campaign and amount.
- Create a second marked test check, confirm that its campaign and amount appear while it is paid, non-voided, and settled, then void or refund it and confirm the reversed amount is excluded from net paid, non-voided settled sales less recorded refunds.
- Reopen the saved ad and compare every setting with the release card once more.
If a setting has returned, correct it and repeat the preview. If the response loses its campaign code, repair the destination or form before launch. If the visit cannot be reconciled to its own check, fix that join before claiming the campaign produced restaurant sales.
Judge a setting by visits and checks, not a low-cost click
Record each activity field exactly as it appears in your own Meta account, and confirm what that field means for the campaign objective you chose. Those platform records can help diagnose delivery, but they do not prove a person visited the restaurant. A completed visit is not a reservation or claim, and directly matched settled sales are not causal incrementality, contribution, or profit.
Use a fixed arrival window. The response-to-visit denominator is the eligible responses whose full window elapsed and whose visit outcome was reconciled and known. The numerator is the successful subset of those exact records whose outcome was a verified completed visit. Hold early successes until their cohort matures, and report open, immature, ineligible, and unresolved records separately.
When comparing one setting on versus off, keep the offer, audience, creative, budget, and measurement window the same. Include media, creative production, separately paid influencer compensation and rights, the actual incremental cost of the offer, and consistently allocated software, agency, and additional paid-labor costs. Do not count the same labor, meal, or shared charge twice. A setting earns its way back on only when the evidence reaches the finish line you chose before the test.
Fresh creative is a separate decision from hidden settings. When response costs suddenly worsen after a stable run, use the creative-fatigue check before changing five campaign controls at once.
Hand off the settings without handing off control
An agency or restaurant marketer should return the release card, saved settings, live previews, campaign IDs, response records, and check-reconciliation report—not just a screenshot of clicks. That gives the owner a repeatable launch standard and a record of what changed.
Feast is the full restaurant performance-marketing service that runs Meta campaigns end to end; Remy handles influencer-program operations and is not the product being offered here. If you want the campaign built and managed against this standard: