Service · Reviews + reputation
Know how buyers read you before they book a call.
Review strategy, directory presence and third-party validation, measured by one score out of a hundred: 412 sources across search, AI answers and review sites, recomputed every six hours, with alerts that fire when it moves.
Perception index
Composite score
Polling
Index, today
72/100
30-day change
+6pts
Search results
78 · healthy
AI answers
71 · healthy
Review sites
58 · neutral
Composite of three sub-indices, recomputed every six hours. Weighted by how much traffic each surface actually receives.
Sources polled
412
Refresh interval
6hrs
Mentions classified
86.4k
What it is
A smoke alarm, not a dashboard.
Reputation problems are cheap to fix early and expensive to fix late, and the gap between the two is usually measured in weeks of nobody noticing. Most SaaS companies find out from a lost deal, which is already too late.
So this is built to be ignored. You do not log in daily. The index sits in the background, and the only time it asks for your attention is when something has actually moved.
412 sources across search results, AI answers, G2, Capterra and directories, Reddit and communities, press, analysts and podcasts.
Every mention classified for polarity, then weighted by how much traffic it gets.
Four standing thresholds; crossing one sends an alert with the mentions attached.
Trend
Index over time
Composite perception index, monthly close. A point is roughly one percent of classified mentions moving a band.
Table view
| Month | Index |
|---|---|
| Jan | 54 |
| Feb | 56 |
| Mar | 55 |
| Apr | 59 |
| May | 58 |
| Jun | 61 |
| Jul | 64 |
| Aug | 62 |
| Sep | 66 |
| Oct | 69 |
| Nov | 70 |
| Dec | 72 |
The read-out
Polarity, not volume.
Polarity, not volume. Bars grow out from the centre line, so a quarter that got louder in both directions does not read as progress.
Strongly negative
Negative
Neutral
Positive
Strongly positive
Share of classified mentions per quarter, percent. Net is positive share minus negative share — neutral is deliberately excluded from it.
Table view
| Quarter | Str. neg | Negative | Neutral | Positive | Str. pos | Net |
|---|---|---|---|---|---|---|
| Q1 25 | 11 | 19 | 40 | 22 | 8 | +0 |
| Q2 25 | 9 | 18 | 41 | 24 | 8 | +5 |
| Q3 25 | 8 | 16 | 40 | 26 | 10 | +12 |
| Q4 25 | 7 | 15 | 39 | 27 | 12 | +17 |
| Q1 26 | 6 | 13 | 38 | 28 | 15 | +24 |
| Q2 26 | 5 | 12 | 36 | 30 | 17 | +30 |
How it works
Collect, classify, index, alert.
01
Collect
We poll 412 sources on a six-hour cycle — search results, AI answers, review sites, Reddit and communities, press, analysts and podcasts.
02
Classify
Each mention is scored for polarity and checked against a sample we grade by hand, so the classifier stays honest as language drifts.
03
Index
Scores are weighted by the traffic each surface actually receives and rolled into one number out of a hundred, plus three sub-indices: search, AI answers and reviews.
04
Alert
Four thresholds are armed by default. Crossing one sends the alert with the mentions that caused it, not just the number.
Signal & alerts
What feeds the score, and what wakes you up.
Contribution
Where the signal comes from
Share of index weight by source, 100 = largest. Weights are set once at setup and held, so a change in the index is a change in reality.
Table view
| Source | Weight index |
|---|---|
| Search results | 96 |
| AI answers | 88 |
| Review sites | 81 |
| Reddit & communities | 67 |
| Press & analysts | 59 |
| Podcasts & video | 44 |
Thresholds
What trips an alert
4 armed
Index drops > 5 pts in 24h
Critical
Negative share > 30%
Warning
Competitor named in an AI answer, you are not
Warning
Review rating falls below 4.0
Critical
Severity is paired with a label, never carried by colour alone. Sparklines show the last ten readings against each threshold.
Results
The gain is mostly in what you catch early.
Monitored and unmonitored companies get roughly the same number of problems. The difference is how long each one sits there unanswered.
Monitored
Unmonitored
Perception index by quarter, both starting at 54. The unmonitored cohort is the same company type with no alerting in place.
Table view
| Quarter | Monitored | Unmonitored |
|---|---|---|
| Q1 | 54 | 54 |
| Q2 | 57 | 55 |
| Q3 | 59 | 53 |
| Q4 | 63 | 56 |
| Q5 | 62 | 54 |
| Q6 | 67 | 57 |
| Q7 | 70 | 56 |
| Q8 | 74 | 58 |
FAQ
Questions we get.
Mostly about how the number is built, which is the right thing to be sceptical about.
What counts as a good score?
Context decides it, which is why the meter shows bands rather than a target. Above 60 is healthy for most SaaS companies and consultancies. What matters more than the level is the direction and the size of any single move — a six-point drop in a week is worth more of your attention than sitting at 64 for a year.
How is the index actually calculated?
Every mention is classified for polarity, weighted by the traffic its surface receives, and rolled into three sub-indices — search, AI answers and review sites — which combine into the composite. Weights are fixed at setup and held, so a change in the number is a change in reality rather than a change in method.
Is the classification automatic, and is it reliable?
Automatic, with a hand-graded sample every cycle so we can see when it drifts. Sarcasm and industry jargon are where it struggles, which is why alerts ship with the underlying mentions attached — you are meant to read them, not trust a number blindly.
Will this create more work for us?
It should create less. The default is silence: four thresholds, and nothing arrives unless one is crossed. Most clients hear from it a handful of times a quarter, which is roughly the number of times something genuinely needed a response.
Can we add our own sources or thresholds?
Yes to both. Clients usually add the two or three communities or directories that matter in their category, and tighten the review-rating threshold if G2 or Capterra drive their pipeline.
How do you measure AI answers?
We run a fixed set of buyer-style prompts (the questions your buyers actually ask) against the major AI assistants on every cycle, and record whether you are named, how you are described, and which sources the answer leans on. The prompt set is held constant, so a change in the score is a change in the answers, not in the questions.
Does it work for an early-stage company?
Yes, though the source list is smaller and the index moves in bigger steps, because fewer mentions means each one carries more weight. We adjust the alert thresholds accordingly so you are not woken up by normal variation.
Start here
Get your current index, free.
Start with a 45-minute trust audit: where buyers look before they book a call, what they find there, and where your competitors show up and you do not. You keep the map whether or not we work together.