Restaurant SMS Marketing: The Messages That Earn Another Visit

Restaurant SMS marketing works when you text only people with a documented permission basis, give them one clear reason and one clear reply, then follow that reply through an identified in-person visit and a paid, non-voided settled check. Build the list rules and reply handling before you write the campaign, because delivery and clicks do not prove that a guest came in or that the send made money.

If your phone list came from several systems, start by deciding what you can actually do with the numbers you already have. The message is only the middle of the job. The useful chain is:

Eligible guest → delivered text → reply or other trackable action → identified arrival → completed in-person visit → paid, non-voided settled check → measured return visit

Each handoff has to work before the next one can tell you anything.

A delivered text is not the result. Another paid visit is.

The point of restaurant text marketing is not to collect sends, clicks, replies, or reservations. Those are early signs; the customer result is a completed visit, and the financial record is the settled check linked to that identified guest.

That distinction prevents a common reporting mistake. A guest can reply and never arrive. A reservation can no-show. A guest can arrive while another person pays. A check can be voided after it first appears. If all of those events become one number called “conversions,” you cannot tell whether the message, the reservation handoff, the host stand, or the check match failed.

Keep each event in its own column:

Event	What it proves	What it does not prove
Sent	Your system attempted the message	The phone received it
Delivered	The provider recorded delivery	The person read it or cared
Reply or link action	The person took the requested next step	They arrived
Reservation or claim	A place or offer was requested	The party showed up
Identified arrival	The guest was recognized in person	A paid sale settled
Paid, non-voided settled check	A sale was recorded and remained paid	The message caused an incremental sale
Later paid visit	The same identified guest returned	The first message alone caused the return

Directly matched sales are still not profit and are not automatically incremental. They are the strongest customer-level record in this chain. A like-for-like change in covers or sales is observed movement, while a forecast is only a modeled opportunity. Keep those labels separate.

Build the sendable audience before you build the message

Your first campaign audience should contain only unique people whose collection record, current permission, and suppression status support the planned promotional text. A phone number in a POS, reservation, Wi-Fi, ordering, or loyalty export proves that you possess the number; it does not prove why the person supplied it or what messages they agreed to receive.

Start with a frozen copy of the source data, then deduplicate by phone number without deleting the original records. For every unique person, retain the collection source and date, the seller named in the disclosure, the exact disclosure version, the affirmative action used to opt in, and any later opt-out. Put unknown-source records and incomplete permission records outside the promotional audience until qualified counsel approves a lawful use.

The U.S. review has three separate gates:

Federal rules. Promotional texts sent with technology covered by the FCC's autodialer rule require prior express written consent unless an exception applies. The federal definition depends on the actual system, so neither “bulk” nor “manual” settles the question by itself.
Federal and state Do Not Call and text rules. The FCC's Do Not Call requirements are a separate analysis, including a seller-specific suppression process. Federal law can leave room for more restrictive state rules, so have counsel check the recipient and sender states instead of treating federal review as nationwide clearance.
Your provider's rules. Confirm the selected provider's current registration, content, opt-out, sending, and record requirements. Provider registration or campaign acceptance does not prove legal consent, and a legally reviewed message is not guaranteed delivery.

This is an operating screen, not legal advice. Have qualified counsel approve the audience, sending technology, collection records, states involved, message, and suppression process before launch.

Your stop process has to understand more than STOP

Configure STOP, QUIT, END, REVOKE, OPT OUT, CANCEL, and UNSUBSCRIBE, but also send ordinary-language requests such as “please don't text me again” to a person or system that can suppress the number. Under the applicable FCC rule, a covered request must be honored as soon as reasonably possible and no later than ten business days; that outside limit is not a waiting period for more promotions. One additional confirmation text may be allowed only under the rule's conditions and cannot contain marketing.

Write the suppression to the restaurant's own record, not only the current campaign. Train anyone who answers the phone or reads replies to record a request immediately. Before every send, remove suppressed people first and count them separately from attempted messages.

Give the guest one reason to care and one easy next step

A useful restaurant text says who it is from, why this person should care now, what to do next, and how to stop future texts. One message should ask for one primary action that your team is prepared to fulfill.

The action can be a reply, reservation, claim, call, or order. Replies are especially useful when the decision needs a human answer, but “Reply YES” is not a plan unless somebody knows what YES triggers, which times are available, and when the conversation gets handed to a manager.

Use a real reason to visit rather than sending a generic “come see us” blast. That reason might be a quieter Wednesday seating, a returning dish, a live event, early access to limited reservations, or a reward the guest already earned. There is no universal offer, send day, expiration window, or monthly frequency for every restaurant. Match the message to the permission given, the daypart you can serve, and the amount of demand you can handle.

These examples use a fictional restaurant. They are campaign copy, not consent language; send them only to an audience that passed the legal and provider review above.

For a slow midweek seating

Green Fork Grill: Wednesday patio tables are open, and the short-rib special ends at 8. Reply TABLE for available times. Reply QUESTION if you need help. Reply STOP to opt out.

For a guest who has not returned

Green Fork Grill: We haven't seen you in a while. Reply RETURN and we'll send this week's quieter dinner times. Reply QUESTION if you need help. Reply STOP to opt out.

After a missed reservation, when the guest's permission and your rules allow it

Green Fork Grill: Sorry we missed you tonight. Want a new reservation this week? Reply YES for times. Reply QUESTION if you need help. Reply STOP to opt out.

The complete collection disclosure belongs at the point where the guest opts in, not squeezed into a later promotion. When written consent is required, the FCC definition calls for a signed writing that clearly authorizes the named seller to send the covered advertising or telemarketing messages to the specified number and says consent is not a condition of purchase. An electronic action can qualify when applicable law recognizes it, but counsel should approve the actual form and retained record.

For more copy patterns, see these restaurant promotional message examples. If you want to study a very short reply prompt, the useful lesson in the seven-word text is not that seven words always win. It is that a reply creates work that someone must finish.

A reply that sits unanswered is a broken campaign

Before you send, assign an owner and a response deadline for every reply type. Decide who handles a normal request, who can change a reservation, who can approve an exception, and who records the final outcome.

The response map can stay simple:

Expected keyword: send the promised choices or link, then record whether the person selected one.
Question: answer from an approved response library or send it to the person who can decide.
Natural-language opt-out: suppress immediately and send only a permitted nonpromotional confirmation, if your reviewed process uses one.
Complaint or sensitive issue: stop automation and send it to a manager.
No response: leave the record as no response. Do not turn silence into a claim, reservation, or visit.

Feast's internal operating notes describe replies to a “why didn't you show up?” message as a revenue opportunity that disappears when nobody follows up. The supported lesson is that replies require handling; it is not proof that every missed-visit text brings a guest back.

Automate the repetitive parts only after the decisions are clear. Software can recognize a keyword, send approved choices, timestamp the reply, and alert a manager. You still decide the reason to visit, the capacity available, the exceptions worth making, and whether the economics justify another send.

Test the full path before a real guest finds the broken step

An integrated test should prove delivery, reply handling, suppression, in-person identification, and settled-check reporting in that order. A delivered sample text is not enough.

Use authorized internal test recipients and follow the same path a guest will use:

Confirm the test recipient is in the eligible audience and not suppressed.
Send the approved message and observe its status move from sent or pending to delivered. Test a failed number separately if your provider supplies a safe test method.
Reply with the main keyword and verify that the promised next step arrives, the response is recorded, and a manager can handle an exception.
Use another authorized test record to send STOP and a natural-language opt-out. Verify both records enter suppression and cannot receive a later marketing test.
Have the primary test recipient complete the identification and in-person arrival steps at the restaurant. Open and pay a positive check through the same POS path a guest will use.
While the check is paid and non-voided, verify that it first appears in the matched campaign-sales report for that test recipient. Record the amount and match time.
Following your POS and accounting policy, void or refund the test transaction only after that successful observation. Verify the reversed transaction is then excluded from paid campaign sales.

Do not publish the campaign if the paid, non-voided check never appears. Fix the guest identity, POS connection, location mapping, or report rules, then rerun the full test. A reply-to-check system that works only in a spreadsheet after the fact will fail as soon as service gets busy.

Count every person once and let each window finish

Your scorecard should use unique recipients for message rates, unique identified guests for visit rates, and dollars for sales. Do not mix texts, phone-number records, parties, covers, checks, and people in one conversion rate.

Record these counts for each campaign:

Part of the job	Counts to keep separate
Audience	Unique eligible people; unique people suppressed before send
Sending	Unique people attempted; delivered; failed; still pending
Reply	Unique eligible delivered recipients whose full fixed response window elapsed and whose reply outcome was reconciled and known; the subset of that exact population who replied inside the fixed window; open, immature, ineligible, and unresolved reply records
Primary action	Unique eligible delivered recipients whose full fixed action window elapsed and whose primary-action outcome was reconciled and known; the subset of that exact population with one valid primary action inside the window; open, immature, ineligible, and unresolved action records
Opt-out	Unique eligible delivered recipients whose full fixed response window elapsed and whose opt-out outcome was reconciled and known; the subset of that exact population who opted out inside the fixed window; open, immature, ineligible, and unresolved opt-out records
Visit	Unique eligible delivered recipients whose full fixed visit window elapsed and whose visit outcome was reconciled and known; the subset of that exact population who completed an identified in-person campaign visit with a matched paid, non-voided settled check inside the fixed window; open, immature, ineligible, and unresolved visit records
Sales	Matched settled-check dollars; discounts; refunds; voids; applicable campaign costs
Check matching	Unique eligible identified in-person campaign arrivals whose full fixed check-reconciliation window elapsed and whose match outcome was reconciled and known; the subset of that exact population whose paid, non-voided settled check was linked inside the fixed reconciliation window; unmatched arrivals plus open, immature, ineligible, and unresolved match records
Return	Unique eligible identified campaign guests who completed the initial campaign visit with a matched paid, non-voided settled check, whose full fixed return window elapsed, and whose return outcome was reconciled and known; the subset of that exact population who completed another identified in-person visit with a matched paid, non-voided settled check inside the fixed return window; open, immature, ineligible, and unresolved return records

Close the message window only when every attempted record has a known delivery result or is explicitly left pending. Close the visit window only after the promised visit period has elapsed and the known outcomes are reconciled. A guest whose visit window remains open does not belong in a failed-visit denominator.

For recipient rates, use the first promotional message in the campaign. Keep confirmations and follow-ups as separate message counts and costs rather than adding the same person to the audience again.

Use these exact formulas:

Delivery rate = unique delivered recipients ÷ unique recipients with a known delivery result (delivered + failed). Exclude people suppressed before send and records still pending.
Reply rate = the subset of the denominator's unique eligible delivered recipients who replied inside the fixed response window ÷ unique eligible delivered recipients whose full response window elapsed and whose reply outcome was reconciled and known. Hold early replies until the cohort matures; report open, immature, ineligible, and unresolved reply records separately.
Primary-action rate = the subset of the denominator's unique eligible delivered recipients with one valid primary action inside the fixed action window ÷ unique eligible delivered recipients whose full action window elapsed and whose primary-action outcome was reconciled and known. Hold early actions until the cohort matures; report open, immature, ineligible, and unresolved action records separately. When the primary action is a reply, the same event may appear in the separately labeled Reply and Primary action records, but count it once inside each measure.
Opt-out rate = the subset of the denominator's unique eligible delivered recipients who opted out inside the fixed response window ÷ unique eligible delivered recipients whose full response window elapsed and whose opt-out outcome was reconciled and known. Hold early opt-outs until the cohort matures; report open, immature, ineligible, and unresolved opt-out records separately, along with pre-send suppressions and failed sends.
Verified paid-visit rate = the subset of the denominator's unique eligible delivered recipients who completed an identified in-person campaign visit with a matched paid, non-voided settled check inside the fixed visit window ÷ unique eligible delivered recipients whose full visit window elapsed and whose visit outcome was reconciled and known. This measures visits the system can verify, not every anonymous arrival. Hold early visits until the cohort matures; report open, immature, ineligible, and unresolved visit records separately.
Check-match coverage = the subset of the denominator's unique eligible identified in-person campaign arrivals whose paid, non-voided settled check was linked inside the fixed check-reconciliation window ÷ unique eligible identified in-person campaign arrivals whose full fixed check-reconciliation window elapsed and whose match outcome was reconciled and known. Hold early matches until the cohort matures; report unmatched arrivals plus open, immature, ineligible, and unresolved check-match records separately.
Verified repeat-visit rate: First define the denominator as all unique eligible identified campaign guests who completed the initial campaign visit with a matched paid, non-voided settled check, whose full fixed return window elapsed, and whose return outcome was reconciled and known. The numerator is the subset of that exact group who completed another identified in-person visit with a matched paid, non-voided settled check in the window. Hold early returns out until the cohort matures; report open, immature, ineligible, and unresolved records separately.

If another person pays, the guest uses cash, or the identity never reaches the check, keep the confirmed arrival in an unmatched-visit column. Do not add an estimated check to directly matched sales. Fix the match for the next campaign, and use like-for-like covers or sales only as a separately labeled observed change.

Measure what the checks can support, not one inflated “ROI” number

The check tells you matched gross sales; your cost ledger tells you what the campaign consumed. Neither one alone proves causal profit.

Track messaging charges, the campaign's share of software or agency fees, copy or creative cost, staff time assigned to the send and replies, payment or fulfillment costs, and the actual incremental cost of any incentive. If a vendor bundles several services, allocate the shared fixed charge on one written basis and keep that basis the same across campaigns. Do not count the same food or reward cost again under a second label.

For a cost-only operating measure:

Full campaign cost per unique identified campaign guest with a verified completed visit and matched paid, non-voided settled check = all applicable campaign costs ÷ unique identified guests who completed an in-person campaign visit with a matched paid, non-voided settled check.

For a cautious sales view, apply your restaurant's stated contribution-margin rate before campaign costs to directly matched settled sales, then subtract the campaign costs not already included in that margin. Call the result a matched-sales contribution estimate, not incremental profit. Some matched guests would have visited anyway, and anonymous or different-payer checks can be missed.

Compare that record with the same daypart in like-for-like periods. If covers, first-time identified guests, repeat guests, average order value, or contribution moved, label it observed lift unless your test design supports an incremental estimate. Platform-attributed sales remain platform-attributed. A forecast remains modeled opportunity.

Message price still matters, but a low per-text price is expensive when nobody follows the reply or checks the visit. Keep carrier segments, platform fees, and overages in the cost ledger. The deeper SMS marketing cost breakdown belongs beside the campaign scorecard, not in place of it.

If you also text food influencers, keep those conversations out of this guest-campaign report. Use the restaurant influencer marketing benchmarks to measure that separate coordination job.

Let the weak step choose what you fix next

Do not send a bigger batch until you know where the first one broke. Each weak number points to a different repair.

Many suppressed records: stop and improve how new permission records are collected. Do not “re-permission” an unverified list by sending it a marketing text.
Many failed sends: clean invalid records and work with the provider on delivery. Do not rewrite the offer first.
Delivery is healthy but replies are weak: test the reason to visit, clarity, sender identity, and one requested action.
Replies are healthy but reservations or arrivals are weak: inspect response speed, available times, booking handoff, reminders, and whether staff can honor what the text promised.
Arrivals are recorded but check matches are weak: fix identification at the host stand or checkout and retest the POS path.
Matched sales appear but the contribution estimate is weak: repair the offer or costs. More volume will magnify weak economics.
Initial visits are healthy but matured repeat visits are weak: improve the next-visit reason and timing. Do not count guests whose full follow-up window has not elapsed.
Opt-outs or complaints rise: pause the send, confirm suppression, and review audience, expectations, frequency, and copy with counsel and the provider.

Continue only when the message reaches eligible people, the restaurant can fulfill the response, matched checks leave enough contribution after costs, and guest complaints stay inside the limit you set before launch. Repair the specific handoff when one step fails. Stop when permission is uncertain, the restaurant cannot honor the response, the technical path cannot prove a paid check, or the economics remain below your written threshold after a clean retest.

The best SMS system earns the next decision from a settled check

Restaurant SMS marketing becomes useful when every message has a permission record before it, a person responsible for the reply, and a visit-and-check record after it. That is what turns a phone list into a controlled way to bring guests back without pretending that every delivery, click, or reservation became revenue.

One operator captured the danger of adding another disconnected tool:

“Just because tools exist does not mean they should all be implemented.”

— Restaurant operator discussing marketing systems, January 2026

Feast becomes relevant when the hard part is keeping restaurant messages, identified guest activity, and the available sales records connected. It is the full restaurant-marketing product for retention and loyalty work; confirm your exact POS, messaging setup, and match coverage before buying because no system identifies every anonymous viewer, cash check, or different payer.

Book a demo of Feast to see whether your restaurant can connect the text, the visit, and the settled check.