Your Google reviews are a free menu consultant. Here's how to mine them

Most restaurant owners treat Google reviews like a report card. You check the score, maybe reply to the bad ones, feel good or bad about the average, and move on. That's a reasonable approach if your goal is reputation management. It's a terrible approach if your goal is understanding your customers.

Here's the thing about a five-star review: someone sat down, finished their meal, went home, opened their phone, and decided to write something. They didn't have to. There was no prompt, no incentive, no follow-up email nudging them to rate you. They just wanted to tell someone about it. That's not data you should be averaging into a score and forgetting about.

Annica Wainwright has spent a lot of time thinking about this. She's the co-founder of Two Forks, a consultancy that works with restaurants like Pizza Pilgrims, Caravan, Waka, Roses Thai and The Shim on menu strategy. Before that, she spent 16 years as a restaurant critic. She knows what it feels like to eat somewhere exceptional, and she knows the gap between what operators think is exceptional and what guests actually walk away talking about.

At Flipdish's Dished Live '25 conference, she walked through the methodology her team uses before any menu project starts. Reading Google reviews isn't a nice-to-have step in that process. It's one of the first things they do.

Not a quick scan. A full year of reviews, every location, read one by one.

What you're actually looking for

Automated sentiment tools give you percentages. Positive: 76%. Neutral: 14%. Negative: 10%. That tells you roughly nothing about your menu.

What Wainwright's team is doing is looking for patterns specifically inside the five-star reviews. Not all reviews. Not the three-star ones where someone liked the food but found the service slow. The people who loved it, who felt compelled to say so. The question they're trying to answer is deceptively simple: what do your happiest customers have in common?

At Tonkatsu, the answer came through clearly once they read everything. Five-star reviewers kept mentioning the gyoza. They mentioned the wings. They talked about cocktails. And a lot of them brought up something specific about the food itself: the house-made noodles, the broth made from scratch, the fact that the kitchen makes the gua fresh when almost every competitor buys it frozen. The guests who raved weren't just people who had a bowl of ramen. They were people who discovered something, ordered more, and felt like they were in on a secret.

The problem was that the menu wasn't helping anyone get there. Customers were getting stuck on the very first decision, whether to get a small or large bowl, and a lot of them just ordered ramen and tap water and left. Perfectly pleasant visit. Not a five-star one.

That gap between what made people happiest and what people actually ordered was sitting in the review data the whole time.

The process you can run without a consultant

This doesn't require budget. It requires time and a spreadsheet.

Step 1: Copy or export all your reviews from the last 12 months.

Do every location if you have more than one. Twelve months captures seasonal shifts and any menu changes you've made, without going so far back that the feedback stops being relevant to how you're running things now.

Step 2: Read only the four and five-star reviews.

Right now you're not doing a complaints audit. You're building a picture of what a genuinely great visit at your restaurant looks like in practice. Save the negative reviews for a different session.

Step 3: Note every dish mentioned by name.

Simple spreadsheet: dish name, how it was mentioned, any context you notice (ordered with a drink, came as part of a group, celebrating something). Don't summarise or paraphrase. Write down what you actually see in the review. You'll notice names repeating faster than you'd expect.

Step 4: Look for combinations, not just counts.

A dish that shows up in 35% of five-star reviews is worth knowing. A dish that shows up in 35% of five-star reviews and is almost always mentioned alongside drinks is genuinely actionable. You're not just counting what people ordered. You're mapping what a great visit tends to involve as a whole.

Step 5: Put what you found next to your actual sales data.

This is the uncomfortable step for a lot of operators. The dishes your happiest customers rave about are often not your highest-volume sellers. When that's the case, the menu isn't doing its job. Something in the layout, the description, or the photography is creating friction between a customer arriving and that customer finding the dish that would have made their night.

A few patterns that tend to come up

Across different restaurants and different review sets, Wainwright's team keeps finding similar things. Worth knowing before you start your own analysis.

Guests who knew how the food was made were far more likely to leave five stars. At Tonkatsu it was the noodles and the broth. The specific detail doesn't matter as much as the principle. Customers who feel like they understand something about the craft behind their food experience the meal differently. It literally changes how it tastes to them. That context usually reaches them through the menu, through a staff member, or through something they read. If none of those channels are carrying the story, you're leaving five-star reviews on the table.

The dishes your chef is proudest of are rarely the ones with the best menu placement. Menus accumulate history. The dishes with the most prominent real estate were probably the bestsellers when the menu was first designed, maybe years ago. Kitchens evolve, palates shift, and the thing your team is genuinely excited about often ends up at the bottom of a section with no image and a description that's basically a list of ingredients.

Almost no five-star review mentions only tap water. Read enough reviews and this becomes obvious pretty fast. Guests who ordered a cocktail, or got genuinely interested in a soft drink they hadn't seen before, had better visits. They spent more time at the table, engaged more with the food, and felt more like they'd had a proper experience. Wainwright's team noticed this at Tonkatsu in particular. Food and drinks were on separate sides of a two-panel menu, which meant customers were making two disconnected decisions rather than experiencing them as one meal. That structural problem was showing up in the review data long before anyone had articulated it as a problem.

What actually changed at Tonkatsu

The Google review analysis was the starting point, not the whole picture. The team also did observational research inside the restaurants, ran customer surveys, held Zoom sessions where people roleplayed ordering from the menu, and ran workshops with kitchen and front-of-house staff. All of it pointed to the same underlying issue: the menu wasn't guiding people toward the experience that made them happiest.

A few specific things changed in the redesign.

The upfront choice between a small and large bowl of ramen was removed as the first thing a customer saw. Previously it was the opening decision, which meant people spent mental energy on portion size before they'd thought about anything else. The small bowl still exists, but it's now in a small separate box with a message about food waste. That reframes it as an option for people who genuinely need a smaller portion rather than a primary choice for everyone walking in.

The "sides" label was dropped entirely. Instead of a section called "on the side," those dishes were arranged around the ramen section in an upside-down L-shape, so a customer's eye naturally moved through them before landing on the main bowl. The header that had been making them feel like an afterthought was simply gone.

Food and drink were brought onto the same central spread for the first time, so both decisions got made together rather than separately as if they were unrelated.

Every dish that received a photograph saw a sales lift. The housemade ginger and sansho soda went up 61%. Pickles went up 137%, partly because customers could finally picture what they were actually ordering. Overall spend per head lifted 12.4% year-on-year.

All of that was already visible in the five-star reviews. The happiest customers had been ordering sides and drinks and leaving raving. The menu just wasn't helping the average customer have that same experience.

If you want to track whether changes like these are actually shifting order behaviour rather than just improving sentiment, Flipdish's online ordering platform gives you that visibility at the item level, so you can see the impact in real sales rather than waiting for the next wave of reviews to roll in.

Three questions worth asking after you've read the reviews

Reviews give you hypotheses. Customer conversations confirm them or tell you you've read it wrong.

Wainwright calls these her three magic questions. They work because they're open-ended and specific enough to produce useful answers rather than polite ones.

"What do you like most about our menu?"

Run the responses through a free word cloud tool like Worditout.com. The words that come up biggest are the ones customers naturally reach for when they describe your food. Those words should be in your marketing copy. Usually they're not, because your marketing was written by someone in a meeting who hasn't eaten there recently.

"What would you change, if you were in charge?"

The "if you were in charge" part is doing real work. It shifts the customer from guest to collaborator, which makes them noticeably more honest. At Caravan, customers kept saying they wanted chips available at brunch. This wasn't a kitchen development project. It wasn't a new supplier conversation. It was a scheduling decision. They made chips available at brunch. Sales at brunch lifted significantly. The information had been there to find.

"How would you describe us to a friend who's never been?"

This one pulls double duty. It shows you how customers actually position your restaurant in their own minds, which is often quite different from how you think you're positioned. And it gives you language. Real language, from real people, describing a meal they actually had. Wainwright calls it "magic dust" for content teams, and she's right. A genuine quote from a customer explaining why they dragged friends back on a random Thursday is more persuasive than anything written in a brief.

One additional tip she shared: end any survey with a question that invites people into something bigger. Something like "We're planning some tastings and interviews over the coming months, can I count on you to get involved?" The people who say yes become your ongoing focus group. They care, they're articulate about their experience, and they're almost always the same people leaving those five-star reviews.

On running this across multiple sites

For a single-site independent, this is probably a Saturday morning. One year of reviews at one location is usually between 50 and 300 entries. You'll start spotting patterns within the first 40 or so.

For operators running several sites, it gets genuinely revealing. If one location gets consistent review mentions of a particular dish that never comes up at a neighbouring site, that's worth investigating. It might be a training gap, a quality inconsistency, or just a menu that looks slightly different between locations because it was last updated at different times. All of those are fixable once you can see them clearly.

Flipdish's restaurant management system centralises menu management across locations, so when review analysis points you toward a change, you're not having to implement it manually at each site.

The actual point

Your Google reviews are not a score. They're a record of what a great visit at your restaurant looks like, written by people who had one, in their own words, naming the dishes they ate and the drinks they ordered and the details that made them feel something.

Wainwright's team reads all of it before they change a single thing on a menu. At Tonkatsu, everything they found in the reviews turned out to be true. The happiest customers had the sides, had the drinks, and knew the story behind the food. The menu wasn't helping most customers get anywhere near that experience. Changing the structure, not the recipes, not the pricing, not the concept, lifted spend per head by 12.4%.

Your reviews probably contain a version of the same story. Most owners just haven't sat down to read them properly yet.

This post draws on insights from Annica Wainwright's talk at Flipdish's Dished Live '25 event. Annica is co-founder of Two Forks, a hospitality consultancy specialising in menu strategy and customer insight.