Google Performance Max 'Store Visits' for Restaurants: Are the Numbers Real?

If your Performance Max report displays a Store Visits figure, that platform-reported number alone does not prove how many identified diners completed a visit, what they spent, or whether the ads caused the visit. Verify the campaign through a separate chain: trackable response, identified guest, completed dining visit, paid non-voided settled check, costs, and—when the decision warrants it—a controlled lift test.

That distinction is the heart of restaurant marketing ROI. A dashboard can report activity while leaving the owner’s real question unanswered: did this spend produce profitable business that would not have happened otherwise?

One agency reported losing eight restaurant clients because the clients no longer believed Google's estimated Store Visits. That experience shows a verification problem; it does not prove that any Google-reported figure was false.

A Store Visits figure is not the same record as a diner or a paid check

When that figure appears in your PMax report, it is a platform-reported result; it is not your guest ledger or your point-of-sale ledger. Keep each piece of evidence under the name of the event it actually records.

A reservation is one booking record. A seated party is one service event, covers are the diners in that party, a unique identified guest is one known person, a completed visit event is one finished dining occasion, and a paid check is one transaction. One party can contain several covers, a known guest can return more than once, and a different person can pay, so report those counts separately.

A hotel restaurant group once asked for location-based tracking specifically because it wanted to avoid a redemption step during service. That request describes the tradeoff exactly: less friction for the guest also means less direct proof for the owner.

Step 1: Write the claim before choosing the numbers

Start by writing the exact sentence you want the evidence to support. “Google reported Store Visits” needs only the Google report; “the campaign produced verified visits” needs identified completed dining visits; “the campaign returned profitable sales” needs settled checks, costs, and a defensible comparison.

Your written measurement plan needs these fields before the campaign starts:

• Campaign name, location, start date, and end date
• The Store Visits figure displayed in the PMax report, if present, retained without relabeling it
• One response path that carries the campaign source into a person-level record
• A fixed response-to-completed-visit window
• A fixed completed-visit-to-check-reconciliation window
• The staff action that verifies the identified party was seated and completed service independently of payment
• The fields used to link that arrival to a check
• Whether each rate counts unique people, seated parties, covers, completed visit events, or checks
• The written rule for voids, refunds, shared codes, cash checks, and a different person paying
• The comparison period or control used for any lift claim

The plan is complete when another manager can classify every record without guessing. If a source tag changes between the ad, booking page, and guest record, fix that join before spending; one Feast campaign investigation found that mismatched source names hid valid reservations until the rule was corrected.

Step 2: Give the ad a response you can follow through a completed visit

Use one trackable response that preserves identity and source from the ad through the visit. A dedicated form, reservation path, claim, or code can start that trail, but none of those actions proves that the guest was seated or completed service.

A click can be anonymous. A code can be shared. A reservation can be canceled, missed, or paid by somebody else. Keep those records as responses or reservations until staff confirm the party was seated and completed service. The difference matters enough that a reservation needs its own value calculation, rather than being counted as a diner on the day it is booked.

A restaurant marketer described monthly click reports as incomplete because they did not show how those clicks became revenue.

No response path captures every anonymous viewer, walk-in, cash check, or different payer. Put unmatched activity in a separate count instead of quietly assigning it to the campaign.

Step 3: Test the full response-to-check chain before launch

Run one test record all the way through a paid, non-voided settled check before the campaign starts. A response that reaches a form but disappears at the host stand is not an acceptance test.

1. Open the campaign destination and complete the intended response with a test identity.
2. Confirm the source, location, and campaign fields survive in the guest record.
3. If the path includes a reservation, confirm the booking while keeping the party size separate from the number of unique identified guests.
4. Use the planned host or service step to confirm that the identified guest's party was seated, then mark the visit complete only after service finishes.
5. Close a paid, non-voided test check and confirm the correct guest, location, amount, and campaign appear in reporting.
6. Only after observing that success, void or refund the test transaction and confirm it no longer enters settled sales.
7. Test one unmatched case and confirm the system leaves it unresolved rather than forcing a match.

For multi-location groups, the naming rule matters as much as the link. One corporate restaurant marketer flagged the difficulty of keeping offer names and point-of-sale keys consistent across dozens of stores. Write the approved campaign and POS labels once, then have each location use the same values.

Step 4: Use mature denominators that cannot produce impossible rates

Count a success only after every record in its group has received the same amount of time to succeed. Early completed visits stay outside the public rate until their full visit window has elapsed, even when the visit has already happened.

Verified completed-visit rate

Unique eligible identified responders whose full visit window elapsed, whose visit outcome was reconciled and known, and whose party completed a verified in-person dining visit ÷ those same unique eligible identified responders whose full visit window elapsed and whose visit outcome was reconciled and known

The numerator is the successful subset of that exact person-based denominator. Report reservations, seated parties, covers, completed visit events, and checks separately; report open, immature, ineligible, and unresolved response or visit records separately.

Check-match coverage

Eligible verified completed-visit events whose full check-reconciliation window elapsed, whose match outcome was reconciled and known, and whose paid non-voided settled check was linked ÷ those same eligible verified completed-visit events whose full check-reconciliation window elapsed and whose matched-or-unmatched outcome was reconciled and known

The numerator is the successful subset of that exact visit-event denominator. Report open, immature, ineligible, and unresolved check-reconciliation records separately. Define the completed visit before matching the check; otherwise the denominator already assumes the success the rate is supposed to measure.

A prior restaurant analysis showed how badly the wrong unit can distort a rate: staff recorded visits through a discount key but did not preserve enough identity to join most of those visits to guests. The fix is not a better-looking percentage. It is a response-to-recipient join and a complete status count.

Step 5: Reconcile sales and costs without calling either one profit

Directly matched sales are the net settled sales on paid, non-voided checks linked to eligible verified visits under your written rule. They are observed matched sales, not proof that the ads caused every dollar.

Keep the two calculations separate:

Directly matched sales ROAS = directly matched net settled sales ÷ ad spend

That ratio answers a media question. It uses ad spend as its denominator, so it does not include the rest of the campaign or the cost of serving the matched checks. A “good” result still depends on your margin, which is why restaurant ad ROAS needs a break-even calculation.

Observed contribution after marketing = directly matched net settled sales − matched-check serving costs − marketing investment

Matched-check serving costs include the actual food and beverage, packaging, processing, fulfillment, and applicable incremental service labor for those checks. Count the actual fulfillment cost of any offer or reward in this ledger, using those same cost components. Marketing investment includes media, creative production, influencer compensation and usage rights when applicable, software, platform, or agency fees allocated under one written basis, and paid campaign-management labor not already included in those fees.

Count every cost once. Offer or reward fulfillment belongs only in matched-check serving costs, not marketing investment. If net settled sales already reflect a discount, do not subtract the discount’s face value again. Paid campaign-management labor belongs only in marketing investment; incremental kitchen or service labor belongs only in matched-check serving costs. If an agency fee already includes campaign management, do not add that labor again. Keep owner-management time as a separate planning cost unless you intentionally monetize it.

Step 6: Compare the PMax figure with your independent records

Compare the reports without forcing them to agree. They may count different events, populations, and time windows, so the gap is a question to investigate rather than a number to average away.

An agency or owner may still choose location-based reporting because a guest-facing claim step can add friction. That is a valid operating choice. It simply sets a lower ceiling on what the report can prove: platform activity instead of identified arrivals and reconciled sales.

Step 7: Use a control when the decision depends on incrementality

Use a comparable holdout when you need to know what changed because of the ads, not merely what happened while they ran. A clean comparison keeps the location, weekday mix, offer, measurement windows, and record definitions as consistent as the business allows.

A before-and-after increase is observed lift. It can be useful, but weather, events, seasonality, capacity, price changes, and other marketing may have moved at the same time. A planned control or holdout gives you a stronger estimate of incrementality; neither Store Visits nor directly matched checks alone removes self-selection.

This is also where privacy becomes an operating requirement. Collect only the identity and transaction fields needed for the written match, limit access, and retain aggregate reporting when person-level detail is no longer needed. A restaurant marketer in the source research raised POS-data access as a trust barrier before any measurement benefit could matter.

Decide whether to keep, repair, stop, or run a cleaner test

Keep the campaign when mature verified visits, check matches, and observed contribution meet the threshold you wrote before launch—and the unresolved count is small enough that it cannot reverse the decision. Treat the result as directly matched or observed unless a valid control supports an incremental claim.

Repair the campaign when responses arrive but completed visits do not, when staff fail to verify that service finished, when checks will not reconcile, or when the offer produces visits with weak contribution. Stop when a complete, mature chain shows unacceptable economics. Run a cleaner test when a broken join, capacity problem, one-off event, or changing offer makes the evidence insufficient.

The final answer is not whether a Store Visits figure in your PMax report is fake. It is whether you can make the money decision without asking that number to prove more than it records.

The Restaurant Marketing Audit is $27 one time for a limited time, with no subscription; it normally costs $47 per month or $500 per year. It includes lifetime restaurant performance reporting, Revenue Simulator access, and one strategy session. If you want to audit whether your PMax evidence reaches a paid, non-voided settled check, get the Restaurant Marketing Audit.