Brand Spectrometer · quick answers
A fast, AI-driven way to see how different audiences actually perceive your brand — and where they disagree. Short, plain answers below: the first few get you using it in under a minute, the rest explain what you're seeing and what it costs.
Just open the tool. A real example brand is already loaded, so you can click around straight away — turn cohorts on and off, hover the heatmap, flip the noise switch. No setup, no waiting.
No. No account, no download, no install. It's a web page. Open it and use it.
Three ways — pick whichever is easiest:
All three live in the Load an atlas section of the tool.
Yes, and it's the simplest path. You don't need a spreadsheet or any prep.
Your own survey or panel data is often the best input. If you already run brand tracking or audience research, hand those results to your AI with the prompt — the Spectrometer will read the eight-dimension shape of each audience straight from your own data.
The page draws the full picture from that. (Heads up: a single-AI reading is a shape — good for exploring, but not yet a proven measurement. More on that under "can I trust the numbers?".)
One small text file. You don't have to write it or even understand it — the AI step above produces it for you, and the prompt tells the AI exactly what to make. If you're curious, you can see a worked example or read the step-by-step guide.
Almost always one of these:
Re-running the AI step with the copied prompt usually fixes a bad file in one go.
How different audiences see the same brand. Each audience (a "cohort") gets a score from 0 to 10 on eight things:
Semiotic, Narrative, Ideological, Experiential, Social, Economic, Cultural, Temporal. In plain terms: its symbols, its story, what it stands for, how it feels to use, who's seen with it, money and value, where it sits in culture, and whether it feels fresh or fading.
Two audiences can like a brand equally yet see a totally different shape. Overlay them and you can see that gap. The heatmap then shows which audiences are far apart and which are close.
The read is drawn as a spectrum strip — eight wavelength lanes, one per dimension, each on its own 0–10 scale (no radar, whose silhouette is an artifact of axis order, not the brand). A detail dial lets you start simple and add cohorts, sentiment and uncertainty one step at a time. How to read it →
Each reading has two parts: how strongly a dimension shows up (the line's height) and whether it shows up positively or negatively — its valence. On the spectrum strip the dot on each line slides from negative (left) to positive (right). A line sitting high with its dot on the left is the dangerous one: an audience that cares a lot, and not in your favour. See how to read it → (Valence needs sentiment data in the atlas; a scores-only atlas shows the lines without the dot.)
By default the distance map shades each gap by size — the bigger the gap between two audiences, the deeper the blue. Flip the noise switch and it grades every gap against the instrument's own margin of error instead:
| Color | Means |
|---|---|
| Green — real | The two audiences clearly differ. You can act on it. |
| Yellow — close call | Probably a difference, but small. Treat it gently. |
| Greyed out — too close to tell | Any gap is within the margin of error. The tool greys it out rather than dressing up noise as a finding. |
So "too close to tell" gets no colour at all — it greys out, leaving only the differences worth trusting. Most tools hide this; this one shows it on purpose. (The grade needs a measurement-grade reading taken by several AI operators; a single-AI atlas shows the size-of-gap map only.)
A normal tracker rolls everyone into one average — one score, one funnel, one number that drifts up or down. That average hides the most useful thing: that your investors, your power users, and your critics may be seeing almost different brands.
The Brand Spectrometer keeps those audiences apart and shows the eight-part shape of each, so you can see exactly where they diverge — and it flags which of those gaps are real versus noise. It's the unusual idea behind Spectral Brand Theory: a brand isn't one fixed thing, it's a spread of perceptions, and the spread is where the decisions are. And because it reads from public material with AI, you get a first look in minutes, not a fielded survey in weeks.
Yes. Nothing you load leaves your browser. There's no server and no database on our end. Your file is read and drawn right on your own device. We never see it, so there's nothing for us to store or leak.
No — on purpose. There isn't one single "true" version of a brand. Different people honestly see it differently, and that spread is the thing we measure. So the tool shows you "here's how these audiences read it," never "here's the final verdict."
Here's the idea most people find surprising — and it's the whole reason the tool works the way it does. A brand isn't in your logo, your product, or your ad, any more than color is in an object. Newton's point about light was that the rays themselves carry no color; color is completed in the eye that sees them. A brand is the same: the company sends signals, and each audience completes them into a perception through its own filter.
That's why the same brand is honestly a different brand to your investors, your power users, and your critics — and why chasing one "true" version is chasing something that was never in the object to begin with. The Brand Spectrometer measures a brand where it actually lives: in the audiences, one cohort at a time. It isn't reading your brand off your assets — it's reading the perception your audiences complete.
Yes. The tool is open and runs in your browser at no cost — no account, no API key, nothing to install. The method behind it is open too: the research is published and the code is open-source. There's nothing to buy to get the full read.
One honest limit: a reading you build yourself in a couple of minutes is a useful shape, not yet a proven measurement. Turning a shape into a measurement takes rigor — multiple independent runs and checked data — which is the method's job, not the screen's. Both use the same open tool. See "can I trust the numbers?".
A rigorous reading is built so it can be checked: feed it the same input and it gives the same answer; anyone can re-create the result from public data; and it only reports differences big enough to clear its own margin of error (the colors above). A quick build-it-yourself reading is great for exploring, but it skips that checking — so treat it as a sketch, not a verdict.
We measured exactly that. A frozen panel of brand artifacts is re-read across model versions, so the movement caused by versions alone becomes its own margin of error — the version floor. At the July 2026 measurement, no version pair — not even models released 18 months apart reading the same texts — moved brand readings beyond the ordinary margin between contemporaneous models (the PRISM-T measurement).
And the answer is checked forward, not assumed: readings taken under the current protocol are epoch-stamped against the measured version floor, and the frozen panel is re-read whenever a vendor ships a new version. If an update ever does move readings, the stamp will say so — a reading never silently changes under you.
No. Brands are named only as examples, for analysis. No logos, no endorsement, no sponsorship. A reading is our best estimate from public material — not a statement from the brand.
The Brand Spectrometer is the measuring tool behind Spectral Brand Theory. The eight dimensions, the audience/shape idea, and the margin-of-error method are written up in published research, and the tool's testing is open and repeatable. Read it all at spectralbranding.com.
Open the tool and try it →