Across our restaurant network, 83.5 percent of guests visit once and never come back. The 16.5 percent who return generate 51.4 percent of all visits. That ratio, computed this month from millions of guest profiles across 1,000+ locations, is why this guide to how to market a restaurant is built around bringing guests back, not just finding new ones. Five pillars, 33 strategies, five findings you will not read anywhere else.
Restaurant marketing in 2026 means capturing every guest’s data automatically, bringing first-time guests back before they lapse, answering every review, and judging campaigns by verified visits and revenue. Across the Bloom network, 16.5 percent of guests produce 51.4 percent of visits.
Most restaurant marketing advice is a list of tactics with no evidence attached. This page is different in one way: every claim about guest behavior below was computed from Bloom Intelligence platform data in September 2026, across guests, orders and reviews from 1,000+ restaurant locations. The five findings are marked. The 33 strategies are organized into the five pillars that the findings point to.
Be the source AI search, Google and voice assistants cite.
Strategies 1 to 6 02Capture every guest, then know who pays the bills.
Strategies 7 to 13 03Campaigns that branch on what the guest does.
Strategies 14 to 20 04Every review answered, every complaint turned into operations.
Strategies 21 to 27 05Visits and revenue verified, not estimated.
Strategies 28 to 33Guests ask ChatGPT, Google AI Mode, Perplexity and Siri where to eat. Those engines pull from the same ranking systems as classic search, and they cite pages with substance. Discoverability work happens on your website, your listings and your reviews, not in your ad budget.
Google says its AI Overviews and AI Mode are rooted in its core Search ranking systems. There is no special AI markup to add. Answer the question your page targets in 30 to 45 plain words within the first 300 words, on a page you own, with evidence nobody else has. Our restaurant SEO, AEO and voice guide walks through the page pattern.
Cited by Google AI Overviews, ChatGPT, Perplexity and GeminiVoice and AI queries are questions: is there parking, do you take walk-ins on Friday, is the patio dog friendly. Put those questions and one-sentence answers on your location pages in the words a guest would use. Assistants read the page that already contains the sentence they need.
Read aloud by Siri, Alexa and Google AssistantRestaurant, LocalBusiness, Menu and opening hours markup earn rich results. Review and AggregateRating markup is allowed only on pages that display the reviews being summarized, under Google’s review snippet policy. FAQPage and HowTo were removed from Google Search in 2026. Markup that does nothing is a maintenance cost.
Rich results with zero policy riskIdentical name, address and phone on every directory. One page per location with hours, menu, photos and the reviews your guests actually wrote. The local pack rewards completeness and consistency, and a location page with substance is the page AI engines quote when someone asks for a recommendation nearby.
Higher Maps placement and more direction requestsFill every field: attributes, services, menu, dish photos, Q and A. Post weekly. Answer the questions guests leave. Then push the same verified details to Apple Business Connect, Yelp, TripAdvisor and Bing so every listing agrees. Google Business Profile is free and it is the front door most guests use.
More profile views, calls and reservationsAI engines recommend restaurants with substance: a real menu, real hours, real guest reviews and consistent sentiment. Roka Akor earned top AI search placement in its category in three markets by making its site the authoritative source, backed by verified guest data from its restaurant customer data platform. Read the Roka Akor case study.
Authority a bigger ad budget cannot buyMost restaurants market to every guest the same way. The data says the guests are not the same. Two findings below show where the visits come from and how the profiles get captured in the first place.
The 16.5 percent who return generate 51.4 percent of all visits. The 1.5 percent with ten or more visits generate 24.7 percent.
This is the whole restaurant marketing problem in two numbers. Most marketing budgets chase the first visit. The visits, and the revenue, live in the second visit and beyond. A guest who reaches ten visits averages 28 visits in the window. A one-visit guest averages one.
Apply the network shares to a location with 10,000 captured guest profiles. About 8,350 visited once. About 1,500 came back at least once and produced roughly half of every visit the location has ever recorded. About 150 crossed ten visits and produced a quarter of them. If cadence breaks for 200 of those 1,500 returning guests, a win-back sequence recovering the network’s 38 percent brings back 76 regulars. Compare that to the cost of finding 76 new guests, then getting even 16.5 percent of those to return.
WiFi login creates 68.9 percent of new guest profiles. Online ordering adds 15.1 percent, reservations 7.4 percent, in-store POS 1.0 percent. Website sign-up forms, the only channel that asks the guest to volunteer, produce 7.6 percent.
If your guest database grows only when someone fills in a form, you are working with a twelfth of the guests who actually walked in. The channels that capture automatically do the heavy lifting, and they run without a marketer touching them.
One profile per guest across WiFi, POS, reservations, online ordering, reviews and surveys. A CRM stores contacts. A restaurant CDP knows who visited, what they ordered, how often they come and what they said. Every other strategy on this page runs better on unified data, and several are impossible without it.
One view of every guestFinding 02: 92.4 percent of new guest profiles on our network arrive through channels the guest never fills in, led by guest WiFi at 68.9 percent. Online ordering, reservations and POS add the rest. A sign-up form on the counter reaches the one guest in twelve who volunteers.
Automatic capture at scaleRecency, frequency and spend, layered with what the guest said in reviews and surveys. Behavioral segments (new, regular, super guest, cooling off, at risk) update themselves from actual visits, not from a list someone built in March.
Segments that never go staleFinding 01: the guests who return are 16.5 percent of profiles and 51.4 percent of visits. Treat them differently. Recognize them at the door, message them less often and more personally, and protect them first when frequency slips. The other 83.5 percent get one job: the second visit.
Marketing spent where the visits areA guest who used to come every two weeks and has not been in for five is telling you something. Score every guest on visit cadence and act when the cadence breaks. Automated win-back campaigns recover 38 percent of at-risk guests across our network.
38 percent at-risk recoveryLifetime value is not average ticket times visits. It is projected revenue across the relationship, adjusted for the risk that the relationship ends. Know the number per segment and the budget conversation changes: a super guest is worth a phone call, a one-visit guest is worth one well-timed email. Retention math here.
Spend matched to guest valueEvery guest in the 83.5 percent already cost you an acquisition dollar. A welcome message within 48 hours, a second-visit nudge inside three weeks, and a reason to come back on a slow night. Our first-time guest return data shows how narrow that window is.
The highest-return dollar in restaurant marketingA calendar blast sends the same message on the same day whatever the guest did. A conditional workflow asks a question at every step (did they open, did they come back, did they spend) and routes each guest down a different path. It is the difference between a mailing and a relationship, and it is the mechanism behind the 38 percent recovery figure on this page.
A conditional workflow is an automation where every step branches on guest behavior. Opened or not. Clicked or not. Returned or not. The same workflow produces a different journey for every guest, and a goal step at the end counts only the guests who actually came back and spent.
That goal step is what makes a recovery rate a verified number rather than an estimate.
Stop exporting lists. Define a trigger (guest enters the at-risk segment) and the branches (opened, did not open, returned, did not return). The workflow decides who qualifies in real time and the list maintains itself.
Hours back every weekWelcome on the first WiFi login. Birthday on the day. Anniversary of the first visit. Win-back when cadence breaks. Our restaurant email benchmarks show triggered messages return far more recipients than calendar blasts, because the timing comes from the guest, not from the marketer.
Relevant every time, automatic every timeText messages get read fast, and unsubscribed fast when overused. Keep restaurant SMS for a table opening tonight, an offer that expires today, or a recovery message to a guest who left unhappy. Let the workflow control frequency so no guest gets two texts in a week.
A channel that stays open“Cooling off” sits between regular and at risk: frequency has slowed, the relationship is intact. A light touch here, a note from the manager or a favorite dish back on special, costs a fraction of a full win-back sequence later.
Recovered while it is still cheapBirthday messages are the most reliable single-message performers in restaurant email. The anniversary of a guest’s first visit is the one most restaurants never send. Recognition for a guest crossing into your top tier should arrive the week it happens. Set them once and they run.
The messages guests actually openBuild Meta and Google audiences from your own guest segments: lapsed guests to win back, lookalikes of super guests, active regulars suppressed. Then judge the campaign on completed orders and reservations, not clicks. That is how the Agentic Ad Manager works, and our restaurant advertising playbook covers the setup.
Ad spend judged on closed checksFinding 02 again: guest WiFi creates 68.9 percent of new profiles. It also recognizes a returning device automatically, so rewards can trigger by email or text without anyone downloading anything. A captive portal is the loyalty program most of your guests will actually join.
Loyalty for guests who never install appsYour review feed is the highest-resolution record of what happens in your dining room, written by the people paying for it. Three findings below on where the words are, how much the platform you watch distorts the picture, and how rarely anyone answers.
66.6 percent of 5-star reviews include a comment longer than a few words. For 1-star reviews it is 89.5 percent, and 84.4 percent for 2-star.
The operational detail (which dish, which shift, which server, how long the wait) is concentrated in the reviews most operators want to look away from. Read them as data, not as insults. 1 and 2 star reviews are 6.8 percent of the total, and they name the problem more often than the other 93.2 percent name the praise.
Finding 03 says the detail is in the low reviews. On its own that is a reading problem: nobody has time to read every comment across six platforms. The view below groups every comment into food, service, environment and staff, scores each theme, tracks it week over week, and tags it with an impact level. The theme card at the left shows inconsistent food quality falling from four mentions to three. That is the sentiment signal.
On its own it is still only one signal. The same guest record carries the check from the POS and the visit history from WiFi and reservations, so a rising complaint theme can be read against order volume for the item and against the value of the guests complaining. A theme raised by three first-time guests is a note. The same theme raised by regulars whose frequency has already slipped is a revenue problem with a number attached. Strategies 25 and 26 are how you work it.
Tock 4.80, OpenTable 4.63, Google 4.56, TripAdvisor 4.31, Yelp 4.25. A 0.55 star spread describing the same kitchens.
1 and 2 star reviews are 13.4 percent of Yelp reviews, 7.4 percent of Google reviews and 3.4 percent of OpenTable reviews. Benchmark your reputation on one platform and you inherit that platform’s bias. Only a unified view across every platform shows your real sentiment baseline.
38.1 percent of all reviews received a reply from the restaurant. For 1 and 2 star reviews it was 39.3 percent. 78.1 percent of all reviews were 5 stars.
The reply rate barely moves with the rating. Restaurants are not triaging their worst reviews, they are answering about four in ten of everything. A public reply is the one marketing message every future guest reads before they choose you. Check your own reply share against 39.3 percent this week.
Finding 04: the same restaurants average 4.80 on Tock and 4.25 on Yelp. Watching one platform gives you one platform’s bias. Unified reputation management puts Google, Yelp, OpenTable, TripAdvisor, Facebook and Tock in one inbox and shows the real baseline.
One inbox, one honest numberFinding 05: 38.1 percent of reviews get a restaurant reply. Guests reading reviews see the silence too. A reply that names the dish and the server, in the language your guests use, is a marketing message aimed at your next guest. AI drafts it in your voice; you approve it or let it publish.
Reply share you can measure weeklyA short survey within 24 hours of the visit reaches the guest before any public review exists. A low score routes to a manager and a private recovery message. A high score routes to a one-tap review link. Every response is tagged with the guest’s segment, so you know whether the unhappy guest was a regular.
Bad experiences handled in privateAcknowledge the specific complaint, do not argue, move the conversation to a private channel, then log the issue against the location and the shift. How to respond to negative restaurant reviews is a repeatable process, and the log is what turns Finding 03 into operations.
Recover the guest and fix the causeFinding 03: 89.5 percent of 1-star reviews carry a written comment. Cluster them by topic and location. Twelve mentions of “wait” at one location this month, at lunch, while lunch checks are down, is an alert for the general manager, not a task for the marketer. Sentiment analysis does the clustering.
Problems seen before they reach the P and LA single NPS hides the story. Split every survey metric by guest tier and you learn whether your regulars or your first-timers are the unhappy ones, which changes the fix entirely. Guest identity plus sentiment is where the useful decisions live.
A score that tells you what to doAcross our network, 78.1 percent of reviews are 5 stars and 6.8 percent are 1 or 2 stars (Finding 05). Track your own two shares monthly. A 5-star share below 78.1 percent with an unchanged kitchen is a review-capture problem. A rising low-star share is an operations problem. They need different fixes.
Two numbers that govern reputationOpens and clicks are vanity. The metric that matters is guests who came back and what they spent, traced from the message to the return visit to the check. Every strategy in this pillar exists to make that trace possible and then act on it.
Closed-loop attribution follows a campaign from the send, to the guest’s return (confirmed by a WiFi login or reservation), to the POS transaction, to the items on the check. Every dollar attributed is observed, not modeled.
It is the difference between a marketing tool that reports opens and a platform that proves revenue. It is also why our client retention is 99.3 percent.
Attribution bolted on at quarter end is guesswork. Connect WiFi, POS and messaging to the same guest profile before the first campaign sends, so every return visit and every check can be tied to the message that caused it. Restaurant marketing attribution, explained.
Verified ROI from the first sendEvery campaign has four numbers: sends, opens, guests who returned and revenue those guests spent. The first two are interesting. The second two are decision-grade. Optimize against returns and revenue and most of your email debates end.
Decision-grade campaign dataUse the numbers on this page: 83.5 percent one-visit share, 78.1 percent 5-star share, 6.8 percent low-star share, 38.1 percent review reply share, 38 percent at-risk recovery, and a network average ticket of $38.73 (orders March to August 2026). More restaurant benchmarks here.
Know where you stand before you spendA group average hides the location that is printing money and the one that needs help. Report campaign returns and revenue per location, monthly. Intervene where the number is wrong and leave the rest alone. Full-stack attribution covers the per-location build.
Interventions where they matterA repeat visit driven by a discount is rented. Real loyalty is a frequency lift that holds after the offer stops. Compare visit cadence for guests who redeemed against guests who returned without an offer, and optimize for the second group.
Frequency that outlasts the couponReview attribution weekly. Tune workflows monthly. Audit sentiment and rating shares quarterly. Re-run your own version of Findings 01 to 05 twice a year. The restaurants that compound are the ones with a calendar. Our blog carries the next round of plays as the data moves.
Advantage that compoundsDo four things and let them run: automatic guest capture at the door, a welcome sequence, a win-back sequence, and a reply to every review by Friday. Then complete your Google Business Profile. That is strategies 8, 13, 17 and 22, and none of them costs media dollars. The coffee shop marketing guide is the small-operator version of this page.
Your opening month is the largest cohort of first-time guests you will ever have, and Finding 01 says 83.5 percent of them will not come back on their own. Capture every one of them from the first night, send the welcome within 48 hours, and schedule the second-visit nudge before you spend a dollar on awareness.
Report every metric on this page per location, not as a group average. Finding 04 applies inside your own group too: two locations with the same kitchen standards can carry very different review profiles depending on which platform their guests use. The fine dining marketing and fast casual marketing guides cover segment-specific plays, and the brewery and taproom guide covers the tasting-room model.
AI engines choose the restaurant. Google’s AI Overviews and AI Mode draw on the same ranking systems as Search, so the page with the most substance and the clearest answer wins the citation. Strategies 1 to 6.
Automatic capture replaced the sign-up form. 92.4 percent of new guest profiles on our network arrive through WiFi, ordering, reservations and POS. Strategy 8.
Campaigns branch on behavior. The recovery rates that matter come from workflows with a goal step, not from calendar blasts. Strategies 14 to 20.
Reputation is managed across platforms at once. A 0.55 star gap between platforms describing the same kitchens makes single-platform monitoring a blind spot. Strategies 21 to 27.
Paid advertising is verified against closed checks. Agentic ad management plans, runs and judges Google and Meta campaigns on completed orders and reservations rather than clicks. Strategy 19 and the Agentic Ad Manager.
Click a name to read the case study
Ten questions from search data and sales calls, each answered in a few sentences with the network figure where one exists.
In a 30-minute session we will pull your one-visit share, your capture mix, your review reply share and your platform rating gap, then show which of the 33 strategies moves them first. No pitch deck.