Social media for local restaurants: what moves covers

Review platforms, on the evidence available. Two economics papers built on Yelp data are the closest thing this topic has to causal proof, and nothing of comparable rigour exists for Instagram, TikTok or Facebook driving restaurant visits. Almost every guide to restaurant social media is written as if the opposite were true.

Michael Anderson and Jeremy Magruder's "Learning from the Crowd" (The Economic Journal, 2012) used a regression discontinuity design on Yelp's half-star rounding. A restaurant averaging 3.24 stars and one averaging 3.26 are near-identical in quality but display differently, so comparing either side of that arbitrary line isolates the displayed number rather than the food. Across Yelp restaurants with 20 or more reviews between January 2009 and January 2011, an extra displayed half-star was associated with San Francisco restaurants selling out their 7pm tables 19 percentage points more often, a 49% relative increase.

The second bar is the one an independent operator should read twice. Restaurants with no external accreditation — no Michelin star, no San Francisco Chronicle Top 100 listing — sold out 27 percentage points more often after gaining a half-star, against 19 for the sample overall. If nobody has written about your 40-seat place, your rating is not one signal among many. It is the signal.

Michael Luca's Harvard Business School working paper (2011) matched Yelp ratings to Washington State Department of Revenue records and found a one-star increase associated with 5-9% higher revenue — an effect that came entirely from independent restaurants, with none measurable for chain-affiliated ones. Both papers land on the same segment: reputation platforms move demand for businesses whose reputation is not established elsewhere.

None of this makes feeds irrelevant. The National Restaurant Association found that sentiment holding across a majority of adults in every age group. Discovery is real. What the record lacks is any study connecting a restaurant's own posting to a booked table.

Below is every channel a restaurant is normally told to work, matched against the strongest evidence for it and the claim that evidence cannot support. The fourth column is the point.

Read the TikTok row honestly. It is thin because the evidence is thin, not because TikTok does nothing. A video can find a large audience while the dining room serves the same covers it did last Tuesday, and nothing published tells you which outcome to expect.

On the published evidence, photos carry a longer signal than reviews do. Mengxia Zhang and Lan Luo's Management Science paper analysed 755,758 photos and 1,121,069 reviews across 17,719 U.S. restaurants on Yelp between October 2004 and December 2015. Consumer-posted photos predicted whether a restaurant would still be open up to three years ahead. Reviews were predictive only one year ahead.

The detail inside that finding matters more than the headline. What related most strongly to survival was the informativeness of the photos — specifically the proportion that were food photos — rather than attributes like composition and brightness. Food photos carried the largest positive association, followed by exteriors, then interiors. A stylised brand shoot and a clear picture of the actual plate are not doing the same job, and the second is what tracked with staying open.

The effect concentrated where the Yelp rating studies concentrated: photos carried more predictive power for independent, young or mid-aged, and medium-priced restaurants than for chains or the price extremes. Three research teams, on different questions with different methods, keep landing on the same kind of business: the independent restaurant with no reputation established elsewhere.

What this implies in practice

• Prioritise clear photographs of plated dishes over brand imagery. Informativeness beat aesthetics.
• Keep a current exterior shot so people recognise the building from the street. Exteriors ranked second behind food.
• Make it frictionless for guests to add their own, because the study measured consumer-posted photos rather than the restaurant's uploads.
• Report photo volume and food-photo share monthly, alongside review count and rating.

Nobody knows, and you should be sceptical of anyone who hands you a number. The advice to post three to five times a week appears on dozens of near-identical pages, frequently prefaced with "studies show". No such study is ever named, linked or described. There is no published experiment testing restaurant posting frequency against covers, revenue or foot traffic.

It is not the only orphan claim here. The most-quoted statistic in this category — that roughly three quarters of people use social media to decide where to eat — traces to a hospitality report whose own source line is a bundle of four references, one of them a statistics-aggregator blog. Another perennial, that around a third of millennials avoid restaurants with a weak Instagram presence, comes from a 2017 UK restaurant-chain PR survey and has been recycled for nine years without anyone re-fielding it.

What does exist is signal from the companies running the ranking systems, absent from the pages asserting weekly quotas. In its Q4 2025 earnings call, Meta reported Instagram Reels watch time in the U.S. up more than 30% year over year, 75% of Instagram recommendations now coming from original posts after raising original-content prevalence by 10 percentage points in a single quarter, and Facebook surfacing over 25% more same-day Reels than the prior quarter. Those describe what the systems favour: freshness and originality.

Instagram is also unusually explicit about what suppresses distribution. Its official ranking explainer names the reels it deprioritises in recommendations:

• Low-resolution or watermarked reels, which includes anything exported from another app with its badge burned in.
• Reels that are muted, or that carry borders.
• Reels that are majority text.
• Reels already posted on Instagram.

The company names reshares, complete watch-through rate, likes and audio-page visits as its ranking predictions. Two consequences follow: recycling a watermarked TikTok onto Reels works against a documented suppression rule, and a fifteen-second clip watched through is worth more than a ninety-second one abandoned at ten.

A defensible policy: post when there is something real to show, favour original footage shot on the premises, and stop treating a weekly number as a strategy. If an outside team runs this, ask which of the claims above they can source. That question sorts social media management vendors faster than any portfolio review.

Probably not as a first move. Pew Research Center's "Americans' Social Media Use 2025", an address-based survey of 5,022 U.S. adults fielded between February and June 2025, found YouTube reaching 84% of U.S. adults, Facebook 71%, Instagram 50% and TikTok 37%. Facebook and YouTube are each visited daily by around half of all adults; TikTok by 24%.

The age split decides platform choice. 80% of adults aged 18-29 reported using Instagram against 19% of those 65 and older, and roughly half of 18-29s open TikTok daily against 5% of over-65s. YouTube and Facebook are the only platforms Pew asked about that a majority in every age group uses. If your dining room skews older, TikTok is a small and expensive audience.

Reach is not relevance, though. A platform reaching 71% of adults is not reaching 71% of people within a fifteen-minute drive deciding where to eat tonight. Pair Pew with what you observe: the age mix in your dining room on a Friday, whether your covers skew resident or visitor, and which platform your regulars already tag you on.

That judgement looks different two towns apart. A waterfront restaurant in Monterey serving visitors competes for attention at trip planning and again when someone is standing on the street with a phone, which puts Google Maps, review recency and food photos ahead of feed cadence. A family diner in Hollister serving the same households every month is running a retention business, where Facebook reach and a loyalty mechanic do more than short-form video will.

Not if the reward depends on the review being positive. The Federal Trade Commission's Rule on the Use of Consumer Reviews and Testimonials (16 CFR Part 465) took effect on October 21, 2024. This is a federal rule with civil penalty exposure, not a platform policy you can appeal.

• §465.4 bans compensating reviewers when the payment is conditioned on the review expressing a particular sentiment.
• §465.5 bans insider reviews — from owners, managers and their relatives — without clear disclosure of the relationship.
• §465.7 bans review suppression.
• §465.8 bans buying or selling fake indicators of social media influence, including bot followers, likes and views.

Asking every guest for an honest review is fine. Asking only the visibly happy ones, or rewarding five stars specifically, is now a rule violation, and §465.8 puts bought followers and likes in the same category. The compliant version of a review programme is boring by design: ask everyone, ask the same way, incentivise nothing that depends on the outcome.

Responding is where the leverage sits instead. BrightLocal's 2026 Local Consumer Review Survey of 1,002 U.S. consumers found 80% saying they are likely to use a business that responds to all its reviews, while 50% view generic responses negatively. That is stated intent rather than observed behaviour, so read it as directional — but the cost of a specific reply over a template one is a minute.

Recency compounds with that threshold. A strong historic average with nothing recent behind it reads as a restaurant that used to be good. The same survey found 97% of consumers read reviews for local businesses, 71% on Google and 45% now through ChatGPT or other AI tools, ahead of Apple Maps at 27%. Review text is no longer only read by people.

Google's own documentation closes the loop: local results are based primarily on relevance, distance and prominence, "more reviews and positive ratings can help your business's local ranking", and "there's no way to request or pay for a better local ranking on Google". Note what it does not claim — nothing there says photos affect ranking. Photos earn their place on Zhang and Luo's evidence, not a ranking promise. Getting the listing right is ordinary local SEO work.

Measure the things closest to a booked table, and be honest that the top of the funnel is directional. The chain from a social impression to a POS ticket has a break in it that no analytics package closes.

Direction requests deserve their own line in a monthly report. Google Business Profile counts how many people tapped for driving directions to your restaurant, which is about as close as a free metric gets to someone intending to walk in. Neither it nor call volume tells you which post caused it.

The bottom of that chain is where the answer lives. Match-back attribution means reconciling activity against point-of-sale records: a promo code used in one campaign only, a reservation source field staff actually fill in, an honest "how did you hear about us" at the host stand. For an independent restaurant it is usually the only defensible way to connect a campaign to revenue.

A defensible monthly scorecard

1. Review count, average rating and review recency. 74% of consumers said they prioritise reviews from the last three months.
2. Photo volume and the share that show food. Food photos carried the strongest survival association across 17,719 restaurants.
3. Direction requests and calls from your Google Business Profile. The closest free proxy for intent to visit.
4. Match-back to the POS. Promo code redemptions, reservation source, and what guests say at the host stand.
5. Saves and shares on short-form video, not views. Instagram names reshares and complete watch-through rate among its ranking predictions.

Follower counts belong nowhere on that list. They correlate with nothing anyone has published about restaurant revenue, and under §465.8 buying them is a federal violation rather than merely a bad idea.

One filter for pitch decks. If someone quotes a headline revenue lift from social media, check whose revenue. Deloitte Digital's 2025 State of Social surveyed 390 director-level-and-above respondents at U.S. B2C companies with 500 or more employees and $500 million or more in annual revenue, and states plainly that the performance figures are executive self-reports rather than audited financials. It describes what enterprise marketers believe happened somewhere orders of magnitude larger than a Salinas taqueria.

The market is larger and more fragmented than it feels from inside one dining room. Bureau of Labor Statistics data for 2024 counts 3,476 private food-service and drinking-place establishments across the five Central Coast counties, employing 59,346 people. Monterey County alone paid $503.7 million in annual restaurant wages.

The visitor economy changes the arithmetic in the coastal counties. See Monterey reported Monterey County visitor spending reaching $3.1 billion in 2024, up 5.7%, with dining spend growing more than 7% — outpacing overall growth. Tourism supported 27,596 jobs and $310.1 million in state and local tax receipts.

A restaurant taking a meaningful share of covers from that spending is marketing to people who will never see a locally targeted feed post, because they were not local when they decided. Their sequence runs through trip planning, a map search on arrival, then a rating and a set of photos on a phone at a street corner. Every step of that runs through the review-and-listing layer — and 45% of consumers in BrightLocal's 2026 survey said they now read local reviews through ChatGPT or another AI tool, which reads review text and listing data rather than a feed.

The inland counties invert that. San Benito's 102 establishments serve a resident base that returns, where repeat visits carry the business and a loyalty mechanic has something real to work on. The question is not which platform is best. It is what share of your covers will come back, because that decides whether you are running a discovery problem or a retention problem. We go deeper on that split on our hospitality and tourism page.

One closing bias check. Every source here carrying real statistical weight studied review platforms and photographs; every source that studied feeds measured reach rather than revenue. That is not an argument that feeds do nothing. It is an argument that the burden of proof falls where the evidence is thinnest, and the thinnest evidence sits under the most expensive line in the average restaurant marketing budget.