Sentiment Analysis: Quickly Read Your Customers’ Minds Using Google’s AI

You probably read every review the moment it comes in. A nice one makes your whole morning, and a bad one sticks with you for the rest of the day (maybe you even replay it on the drive home).

But reading reviews one by one means you only ever hear from one customer at a time.

And one review on its own doesn’t tell you much. It’s only when you line up fifty of them that you notice four different people mentioned the parking, or that every complaint about slow service came in on a weekend.

Fact is, Google already reads your reviews this way, all together, and it shows the results to your customers. So it makes sense for you to see them too!

The short answer

Google Maps groups what customers write into topics, like service, price or wait time, and can show whether people generally feel good or bad about each one. That’s sentiment analysis. You can do a simple version yourself, on your reviews and your competitors’, to find what to fix first.

What is sentiment analysis, in plain words?

It sounds a lot more technical than it is. Sentiment analysis is just reading a review for two things: what the customer is talking about, and how they feel about it.

Take a review like this one: “The khao soi was amazing, but we waited almost forty minutes for it.”

That’s one review talking about two separate things. There’s the food, which they loved, and the wait, which they didn’t. Whatever star rating they gave you (a three? a four?) squashes both into a single number, and you lose the part that’s useful to you.

Sentiment analysis pulls them apart again: food is positive, wait time is negative. Do that across all of your reviews and you get something far more helpful than an average, which is a list of what customers love about you and what keeps annoying them.

How does Google’s AI actually read a review?

Google doesn’t publish exactly how Google Maps does it. But Google also offers this kind of analysis to developers through its Cloud Natural Language tool, and its documentation explains the basics pretty clearly.

First, the text gets broken into words, and each word is tagged with what kind of word it is. The nouns are usually what the review is about (the khao soi, the wait), and the adjectives are usually how the customer felt about it (amazing, or not so amazing).

Then the text gets a score from -1.0, which is very negative, to 1.0, which is very positive. There’s also a second number called magnitude, which measures how strongly someone feels, good or bad.

Why two numbers? Because, as Google’s documentation points out, a score near zero can mean a review is neutral, or that it’s full of strong feelings pulling in opposite directions. Our khao soi review is the second kind, and magnitude is what tells the two apart.

The same tools can also score each thing a review mentions separately, so the khao soi gets its own positive score and the wait gets its own negative one. Do that over a few hundred reviews and the same handful of things keep coming up. Those are your themes!

Nobody outside Google can tell you how closely Google Maps follows this exact method, but the idea is the same, and you can see the results on your own profile.

Where can I see what Google thinks my reviews are saying?

Right above your reviews. Open Google Maps on your phone, search for your business the way a customer would, and scroll down to the reviews.

Depending on how many reviews you have, you might see a row of short topics there, like “wait time”, “price” or the name of a dish. Google calls these Place Topics, and says they’re based on the main themes in your reviews. Customers can tap one to see only the reviews about it, and Google says a topic may also show a prevalent sentiment, meaning whether people mostly felt good or bad about that thing.

Just a heads up: topics only appear once a business has enough reviews, and there’s no way to edit them yourself, because they come from your customers’ words rather than yours. (If a topic is abusive, Google’s advice is to flag the reviews behind it.)

So a customer might already know what people think of your wait times before they’ve read a single review! And in some countries, Google goes a step further with AI-written review summaries that sum a place up in a few lines.

Can review themes really warn me about problems early?

Yes, and honestly, that’s the best reason to do any of this!

Your star rating is slow to move. It’s an average of every rating you’ve received, so once you’ve got a decent number of reviews, a problem has to annoy a lot of people before the number drops by much (we did the math on this in why reviews are your most powerful growth tool). Themes show up much earlier.

Picture a noodle shop off Sukhumvit with a steady 4.6. Over the last two months, five reviews have mentioned waiting, all from weekend lunchtimes. Most are still four stars because the food is still great, so the rating barely moves.

Read those five together, though, and a pattern shows up. Weekend lunch is also when delivery orders spike, and the counter fills up with riders waiting for pickups, so the people sitting down to eat get served after the delivery bags go out.

Nobody wrote “please sort out your delivery pickups”, but the reviews pointed straight at it. A pickup shelf by the door, or one person handling bags on weekends, and it’s fixed long before it shows up in the stars.

It works for the good stuff too. If two reviewers mention the same staff member by name, that person is part of why customers come back. Good to know the next time you’re planning the roster!

What can my competitors’ reviews tell me?

Quite a lot, and they’re completely public! Nothing stops you from reading the reviews of the places your customers are most likely to choose instead of you, and running the same exercise on them.

One rule, though: use what you learn to get better, NOT to go after them. Don’t comment on their reviews, and never post (or pay for) reviews about them, because Google’s content policy bans exactly that.

Their negative themes are the gaps you might be able to fill. Say you run a dental clinic, and the clinic two sois over keeps getting complaints that nobody answers the phone. If your team picks up quickly, that’s something you’ve got over them, and right now nobody comparing the two of you can see it.

So make it visible in your business description and your posts! Reviews from happy customers who bring it up on their own will back you up even better.

Their positive themes matter too. If every competitor nearby gets praised for the same thing, that’s probably what customers in your area see as standard, and you don’t want to be the one place falling short.

And if you’re watching more than a couple of competitors, or comparing several branches, local market intelligence does this for you: it benchmarks you against the businesses near you and tracks how customers feel about every location.

How do I find my top three strengths and weaknesses?

You don’t need any special tools for this. A notebook works, or a simple spreadsheet. Give yourself an hour, grab a coffee, and work through these steps.

1. Pick your reviews. The last three months is a good start, or your last thirty reviews if you don’t get many.

2. For each one, write down what the customer talked about. Look for the nouns: the dish, the staff, the price, the wait, the parking. Ignore the star rating for now!

3. Mark each as positive or negative. A mixed review gets one of each, like our khao soi customer.

4. Group words that mean the same thing. “Slow”, “queue” and “took forever” all go under wait time.

5. Count them up. Your three most-praised themes are your strengths, and your three most-complained-about are your weaknesses. Don’t be surprised if something lands on both lists (staff often does).

6. Compare your list with the Place Topics I mentioned earlier. If Google is showing customers a topic you didn’t expect, find out why.

7. Do it again next month. One month is a snapshot, but the change from month to month tells you whether a fix is working.

Then decide on one thing to do for each. Make sure customers can see your strengths before they visit, in your description and your photos. For your weaknesses, pick one practical fix each, starting with whichever comes up most.

And when a fix is in place, say so the next time a review mentions it. A reply like “thanks for letting us know, we’ve added a second person at the counter on weekends” tells every future reader it’s been dealt with, and the people reading your replies usually aren’t the reviewer at all. Not sure how to word it? Our formula for replying to any review walks you through it.

What if I have too many reviews to do this by hand?

Then the spreadsheet stops being fun pretty quickly! Doing this for one shop with a few dozen reviews a month is very manageable. Doing it for hundreds of reviews, or across several branches, is basically a part-time job.

That’s where a review management tool with sentiment tracking built in earns its place. It reads every review as it comes in, picks out the recurring themes and shows you how sentiment is trending, so you get the same top-three list without the tally marks.

Either way, start with this month’s reviews. Your customers have already done the hard part by writing it all down for you. All you need to do is read them together!