You can't manage what you refuse to measure, and brand cohesion is one of the things people refuse to measure. The reason is honest enough. The obvious measurement, an aggregate reputation score, is worse than useless. It hides decline for years. A brand built over decades absorbs eighteen bad months without shifting a decimal point, in the same way a restaurant with a 4.5 rating built over twelve years absorbs a bad chef without the number moving.
We use a different measurement in another part of our work, and it applies here directly.
The city-guide method
Helping people feel at home in new places has been a favourite pastime for decades. There are new ways to look for patterns.
When we build a shortlist of the interesting bars, restaurants and shops in a city, the failure mode is easy to describe. You list a place that has coasted for two years and everyone in the trade knows it, but the aggregate rating is still 4.4, and the tourist reviews are still cheerful, and the guide gets forwarded and reread and the writer is exposed.
The rule we developed is: for every pick, read the newest reviews rather than the top ones. Google surfaces the oldest and most enthusiastic. Decline lives in the language of the recent ones.
Specific signals, in rough order of reliability:
- Language in recent reviews. Used to be. Not what it was. Still good but. We have been coming for years. These phrases are the tell, and they appear long before the rating moves.
- Velocity up, rating down. A place has been discovered. Sixteen thousand reviews on a cafe means a queue, not a cafe.
- Award trajectory, not award presence. Ranked thirty last year and absent this year says more than any absolute position.
- Trade press silence. Constant coverage for five years and nothing for two usually means the interesting people left. A bar's reputation is a head bartender, and when that name moves without a successor being written about, the room is coasting on its fit-out.
- Menu fossilisation. The same signature dish or drink quoted in write-ups a decade apart.
The read works in any language now. Kyoto reviews in Japanese, Bangkok reviews in Thai. The barrier that used to gate this eval to English-speaking cities came down with the rest of them.
None of these needs a big data project. Each of them is checkable in an afternoon. For a single restaurant.
For a brand it used to be a season. Years of posts across dozens of authors, decks revised quarter by quarter, employee comms, analyst notes, trade coverage, results calls. Reading all of that at language level was a team of analysts and a lot of time, and nobody was going to fund it. Language models have just made this a well-designed prompt. Multivariate pattern-matching across a large text corpus is exactly what they are good at, and it is exactly what a brand team couldn't do by hand. The reason nobody has been running this eval at brand scale isn't that nobody knew about it. It's that until this year, it cost more than it was worth.
What this looks like for a brand
The city-guide eval reads for decay in the language around a venue. A brand eval reads for decay in the language around the brand. Same shape, different surfaces.
- Language on your own recent surfaces. Your most recent LinkedIn posts, employee comms, sales collateral shipped in the last quarter. Same phrases: used to be, we've evolved, at heart we are still, in essence, moving forward, more relevant than ever. All of these say the same thing. Something has drifted and the drafter is smoothing over it.
- Volume up, engagement down. More posts, fewer meaningful reads. This is the brand equivalent of the queued cafe with the falling rating. Attention was bought, not earned, and the room is running on presence.
- Trajectory in industry press, not presence. Analyst mentions, sector coverage. Named in three roundups last year and absent from this year's is a signal. Which specific things you were named for, and whether the naming is fresher or staler, matters more than the count.
- Trade press silence. Constant coverage for a stretch, then nothing. Your brand is a small number of specific voices; when they move on and their successors aren't written about with the same weight, the room is coasting on its earlier hires.
- Message fossilisation. The same three claims quoted in your own materials for a decade. The trap is that you can still say them and mean them, but the buyer stopped noticing them the day the fourth competitor started saying the same thing.
Every one of these is checkable in an afternoon. Very few brand teams do.
Why the aggregate scores keep hiding it
The reason brand teams keep buying aggregate reputation trackers is that the number is comforting. It only moves once every eighteen months, and when it moves, everyone in the room has a plausible external explanation. The decline was there before the score noticed. It was in the language on your surfaces. It was measurable at the sentence level six or nine months earlier.
The point isn't that the trackers are wrong. It's that they detect things too late. By the time the number moves, the cohesion you were trying to defend has already been spent.
The trackers were the only affordable option. Reading the surfaces at language level required a team of analysts and a lot of time. That constraint just quietly disappeared. The eval that was always more accurate is now also cheaper.
The city guide has the same problem in miniature and solves it the same way. Read the language on the surfaces the audience actually sees. Sample recent, not top. Look for the specific tells. Grade the fade.
What to do about it
The eval isn't the fix. It's the diagnostic. If a brand is fading in the specific ways above, the fix is what it's always been: recover the position, restate it in your own words, put it on the surfaces machines and buyers hit first, hold it there long enough to compound. If it isn't fading, the eval is a cheap way to know you can keep spending on what you're doing.
Either way, an afternoon with the right prompt, reading your own recent surfaces the way we read a Kyoto restaurant's newest reviews, will tell you more about the state of your brand than most of what your dashboard is showing you.
Plan B runs this diagnostic on request. If you want to know whether your brand is coasting, get in touch.
