Your competitors are running a free focus group. Almost nobody reads the transcript.

The owner of a pet boarding kennel outside Boise told me he reads every review his business gets, usually within an hour of it landing. Sixty one reviews over four years, all of them read twice, most of them answered. Then I asked how many reviews he’d read for the three boarding places within four miles of him.

Zero. Not one, in four years of competing with them.

Those three businesses had 412 reviews between them. Every one of those reviews was a customer in his category, in his zip code, describing in their own unpaid words what they wanted, what they got, and which gap between the two made them angry enough to type. He had a running, public, permanently archived focus group operating on his behalf, and the transcript was free, searchable, and sitting four miles away.

This post argues that the highest-value reading in local marketing is your three closest competitors’ review pages, that their one-star reviews are your positioning handed over for nothing, that their five-star reviews are actually the more useful half, and that the correct way to use all of it is to read everything and respond to nothing.

Why the usual version of this advice is useless

Search for competitor review analysis and you will be sold a benchmarking dashboard: your 4.3 against their 4.5, a sentiment gauge, a word cloud, a monthly PDF with an arrow on it. Some of these tools cost $300 a month.

A competitor’s star rating tells you almost nothing you can act on. It’s an average of self-selected writers across years of staffing changes, compressed into one decimal. Two businesses at 4.4 can be failing in opposite directions. And a sentiment score is worse, because it collapses the only thing that matters, which is the specific named expectation a customer had, into a number between zero and one. Nobody ever changed a service because sentiment was 0.62.

The word cloud is the purest version of the problem. It will tell you that customers of a boarding kennel talk about “dogs,” “staff,” and “price.” Thank you.

The value is in the sentences. Reading 412 reviews by hand takes a weekend, and a weekend is cheap for what comes out of it. There’s an academic version of this argument, if you want one: Netzer, Feldman, Goldenberg and Fresko’s 2012 paper in Marketing Science, “Mine Your Own Business,” showed that the raw text of consumer discussion could reconstruct a market’s competitive structure closely enough to match the maps built from sales and survey data. They needed a text-mining pipeline because they were reading tens of thousands of posts across a national category. You are reading four hundred about three businesses near you. You are the pipeline.

Pick three, and pick them honestly

Three is the number for a reason that isn’t arbitrary. BrightLocal’s 2026 consumer search behavior research, run on a panel of 1,227 US consumers, found that 72% of consumers look at three or fewer businesses before deciding, and that even the people still in the research stage mostly stop at five. Your prospect is not surveying the market. They are opening three tabs, and you are either one of them or you are not in the conversation.

So the three you study are not the three you resent. They are the three that appear next to you when a real buyer runs the query that actually brings you money. Run that search yourself, on a phone, logged out, from the part of town your customers live in. Whoever is sitting beside you in that map pack is your competitive set, even if they’re smaller than you, newer than you, or worse than you. Especially if they’re worse than you and winning anyway. That’s the most instructive case in the whole exercise.

Read all their platforms, not just Google. BrightLocal’s 2026 Local Consumer Review Survey has consumers consulting an average of six sources before choosing a local business, and Google’s share of that attention slipping year over year. The complaints that never make it to Google often land somewhere with less friction: Yelp, Facebook, Nextdoor, a category site like Rover or Angi, sometimes a subreddit. Different platforms surface different anger. Facebook reviews skew toward people who feel personally wronged. Category sites skew toward process complaints, because the buyer went in with a transaction in mind.

What to actually pull out

Open a spreadsheet. Seven columns, and the seventh is the one that does the work.

Date. Star rating. Platform. Which competitor. The literal quote, copied exactly, no paraphrasing. Then: what did this customer think they had bought? And finally: is this an execution failure or an expectation failure?

That last split is the entire method. An execution failure is a business doing its own stated job badly on a given day: the van was late, the groomer was rude, the order was wrong. It’s real, it matters to that customer, and it is almost worthless to you, because it’s a staffing problem they will probably fix and you will probably have too.

An expectation failure is different. It’s a customer who got exactly what the business intended to deliver and was still angry, because the business had let them believe they were buying something else. That gap is structural. It lives in their pricing page, their phone script, their intake form, the photo at the top of their site. They cannot fix it without changing what they sell, which means they usually don’t fix it, which means it is still there next year.

Expectation failures are the ones worth your weekend. They’re a map of the promises your category’s customers believe they’re making and nobody in your market is actually keeping.

Reading for the cluster, not the review

A single one-star is noise. Somebody was having a bad Tuesday. What you’re hunting is the same complaint, in different words, from unrelated people, across more than one competitor. Cross-competitor repetition is the signal that you’ve found a category expectation rather than one business’s bad hire.

My rough threshold: five independent mentions across at least two competitors, inside eighteen months. Under that, hold it and keep reading. Over that, it’s real, and it’s probably worth changing something you do.

A worked example, complaint cluster to offer

Back to Boise. Fernwood Pet Lodge, thirty runs, boarding and daycare, midweek occupancy sitting around 62% and the owner convinced his problem was that his facility was older than the place that opened in 2024 with the glass play yard and the webcams.

He read 412 reviews over two evenings and a Saturday. Sixty four of them were one or two stars. Nineteen of those sixty four, across all three competitors, said some version of the same thing:

“I texted twice on Thursday asking how she was doing and never heard back. Found out at pickup she hadn’t eaten since Tuesday.”

“No news the whole week. I know no news is supposed to be good news but I was in Costa Rica picturing the worst.”

“They were great with him, honestly. But I had to call three times to get anyone to tell me anything and that’s not okay when it’s your dog.”

Notice the third one. The dog was fine. The care was fine. She gave two stars anyway, and she said why: being uninformed about her own animal for a week was the product defect. That’s an expectation failure, and it appeared at all three competitors, which is what made it a category finding rather than gossip about one understaffed front desk.

None of the three websites promised updates. None of them promised no updates either. They all led with facilities: square footage, play yards, the webcam thing. The customers had filled in the silence with an assumption, and the assumption lost every time.

Here is what Fernwood did with it.

The offer change. Every boarding stay now includes what they call the 6 o’clock: one photo and two specific sentences about the dog’s day, sent by text before 6pm, every day, including the day of drop-off. Specific means “Maple ate her whole dinner and spent forty minutes with the collie group,” not “Maple had a great day!” If the 6 o’clock doesn’t arrive, that night is free. The guarantee is on the booking confirmation, so nobody has to remember to claim it.

The landing page line. The old hero said “Boise’s home away from home for your best friend.” Which is every kennel in America. The new one says: “You’ll hear from us at 6pm every day. Not just when something goes wrong.” That second sentence is doing the real work, because it names the fear the nineteen reviews were actually describing.

The staffing consequence. Thirty dogs, two sentences and a photo each, is about forty minutes of somebody’s afternoon. That is a real cost and it had to come from somewhere. They moved the kennel tech’s shift by an hour and cut the Saturday deep-clean rotation to every other week. Every guarantee you make out of a competitor’s complaint cluster costs something. If it costs nothing, you probably haven’t promised anything.

Eleven weeks later midweek occupancy was 79%. The owner’s read on why: the guarantee gave people a concrete reason to switch away from a place they were only mildly unhappy with. Mild unhappiness doesn’t move anybody. A specific unmet expectation, named out loud by somebody else, does.

The five-stars are the more useful half

Everyone who does this exercise starts with the one-stars, because conflict is interesting and because schadenfreude is a hell of a motivator. Then they stop, having collected a list of things not to do.

The five-stars are where the money is, and it took me an embarrassingly long time to work out why. A one-star tells you what a business failed to deliver. A five-star tells you what a customer thought was worth stopping their day to write about, unpaid, for strangers. That’s a much rarer signal. Praise is the category telling you what it actually buys on, as opposed to what everyone in the category advertises.

Those two things are usually not the same thing, and the gap is often absurd.

Fernwood’s competitors’ five-stars, 61 of them read closely: the glass play yard was mentioned four times. The webcams, seven, and three of those seven were complaints about the webcam being down inside an otherwise glowing review. Square footage, zero. What did get mentioned, in 38 of 61, was a human being by name. “Marisol remembered he can’t have chicken.” “Dev texted me a picture unprompted.” “Whoever was working Sunday night stayed late so I could get him after my flight.”

Three businesses were spending their entire marketing budget on facilities. Their happiest customers were writing about staff continuity. The market was buying a named person who knows my dog, and all three websites were selling real estate.

That finding changed more at Fernwood than the complaint cluster did. It put four staff bios with photos on the site, with how long each person had been there. It put a line in the booking flow asking which handler your dog had last time, and an actual attempt to honor it. It changed what the owner said in his own review replies: instead of “thanks for the kind words,” he started naming the person the customer had dealt with, which turns a generic reply into evidence for the exact thing his market rewards.

Run this on your own category and you’ll usually find the same shape. Restaurants advertise the menu; the five-stars are about being recognized on a second visit. Law firms advertise verdicts; the five-stars are about somebody returning a call on a Saturday. Dentists advertise technology; the five-stars are about not being made to feel stupid for waiting three years. The category advertises the capital expense and buys on the emotional one. Your competitors’ five-stars are the cheapest way to find out which is which in your town, specifically, rather than in general.

The line, and it is not a blurry one

Everything above is reading. The moment it becomes writing, you are in a different category of activity, and the platforms have written the rules down.

You cannot respond on their listing. Not in the sense that it’s frowned upon. The owner reply feature is bound to the verified profile, so there is no mechanism for you to post an owner response under somebody else’s review. What’s available to you is posting a review, answering a question in their Q&A, or voting on their content, and all three of those are exactly what Google’s Maps content policy covers. The fake engagement policy states that merchants cannot post content on a competitor’s place or business to undermine that business’s reputation, and the prohibited and restricted content policy treats content written by someone with a conflict of interest, which it spells out as including industry competitors, as removable, and separately bans posting content about a competitor to manipulate their ratings. Yelp’s content guidelines say the same in plainer language: business owners don’t write reviews of competitors, and don’t pressure people into changing or removing reviews.

Removal of the review is the mild outcome. Enforcement against manipulation can land on the profile doing the manipulating, which means trading your own listing for a one-star you posted under a pseudonym from the shop wifi. People do get caught. The IP and the account history are not as anonymous as the burner Gmail feels.

Don’t contact their unhappy reviewers. The temptation is real and the mechanics are easy: a woman posts a furious two-star about a competitor’s response time, her name and often her Facebook are right there, and a friendly message offering a better experience seems like the most efficient sales call available. Do not. Here I want to be precise about what kind of rule this is, because the honest answer is less quotable than the one above. Google’s policies are explicit about what you may not post on a competitor’s listing; they are not written as a single clause forbidding you to message a stranger off-platform. Yelp’s guidelines are closer, covering pressure applied to reviewers. What actually stops this is narrower and broader at once: the platforms’ general spam and harassment terms, the fact that scraping reviewer identities for outreach runs into consent problems in several states and in the EU, and the specific, entirely foreseeable outcome where she screenshots your message and posts it. That screenshot is a worse review than any you’ve ever received, because it’s evidence of method rather than evidence of a bad day.

Never name a competitor in your own replies. Not to compare, not to defend yourself, not even favorably. A reply that says “unlike some places in town, we always call back” tells every reader that you spend your time thinking about other businesses, and it hands your competitor a screenshot. Write the reply as if they don’t exist. The positioning should be visible in what you promise, never in who you mention.

The one time competitors legitimately enter your review workflow is when you believe one of them is posting fake reviews on your listing, which is a documentation and flagging problem with its own procedure. We wrote that up separately in what to do about a competitor’s fake review. The rule of thumb that keeps you out of trouble: their reviews are an input to your product decisions and never an input to your publishing decisions.

The part nobody tells you: half of what you find has a shelf life

The complaint clusters expire. Not all of them, but the ones you’d most like to build on, because a competitor who loses enough business to a problem eventually fixes it. If your entire differentiation is a guarantee built on their weak spot, you have bought a lease, not a position, and you won’t be told when it ends.

Which is why the five-star findings are worth more. What a category rewards moves slowly. Fernwood’s competitors can hire a front desk person and answer the phone by spring. What they can’t do in a quarter is build four years of staff continuity, and continuity is what their own happiest customers keep writing about. Build on the thing that takes time to copy.

The other half of the shelf-life problem is dosage. Do this twice a year, in a block, with a deadline. I’ve watched an owner turn it into a daily habit and it went badly in a specific way: he started reading his competitors’ reviews the way you read an ex’s social media, which is to say for his own emotional weather rather than for information. His replies got defensive. He started counting their review velocity against his. Two sittings a year, spring and fall, spreadsheet closed in between, is the version that stays useful.

One more failure mode, the sneaky one. Mining a competitor’s complaints makes it very easy to build your whole identity out of not-being-them, and negative positioning reads as insecurity even when the customer can’t say why. The 6 o’clock guarantee works because it’s a promise about Fernwood, stated without reference to anyone. The version that fails is the same guarantee phrased as “unlike other kennels, we actually update you.” Same fact, half the credibility.

Where this fits in the rest of the work

None of this replaces answering your own reviews, which remains the highest-return hour in the whole discipline and is covered in our guide to reputation management for small businesses. Competitor reading is a strategy input. It tells you what to promise. Your own review responses are where you demonstrate, in public and repeatedly, that you keep the promise.

The two connect more directly than most owners expect. Once the five-star analysis told Fernwood the category buys on named staff, the reply template changed, and every reply after that became a small ad for the thing the market had already said it wanted. If you want a starting structure for those, there’s a set of positive-review reply templates worth editing into your own voice, and if the volume has grown past what one person can sustain by hand, the tooling comparison lives in our roundup of review response software.

Go read the transcript

Block out a Saturday. Open three tabs. Copy the sentences, not the stars, and sort them into things those businesses did badly and things their customers believed they were buying. Read the praise more slowly than the anger. Then close the tabs and don’t type a single word into any of them.

Your competitors have been paying for this research for years, in churned customers and bad weeks and a thousand small disappointments typed out at eleven at night. The findings are published. Nobody sends you an invoice for reading them, and nobody is going to tell you they exist.