What is a Heat Score, and when should I act on it?
A Heat Score is a 0–100 reading of how much your category is genuinely talking about one subject right now. It combines how much is posted, how far it reaches, how hard people engage, how many separate accounts are involved and how many platforms it appears on. At 60 or above, write into it this week.
- Independent on 3 platforms14 accounts
Three things you get
- 0–100, with five named bands
- One loud account cannot fake it
- Says when to write, not only what
What comes back
The shape of it, drawn at a size the figures can be read. Placeholders, not a customer's numbers.
How is a Heat Score calculated?
Three measures are averaged, then two multipliers are applied to them. The measures are frequency at 35%, reach at 30% and engagement rate at 35% — how much is being published, how far each post travelled against what is normal for its platform, and what share of the people who saw it did something about it.
The detail+–
The two multipliers are what make the number worth reading. Source diversity scales the result between 0.35 and 1 depending on how many separate accounts are involved. Cross-platform corroboration multiplies by 1.15 when a subject appears on two of the five platforms read and 1.3 when it appears on three or more.
That ordering is deliberate. The measures say how loud something is; the multipliers say whether the loudness is evidence. A subject can be published often, travel far and still be discounted by two thirds because it all came from one account.
Why does one account posting ten times score so low?
Because one account posting ten times is one opinion, not a trend, and a score that cannot tell those apart is a score anybody can game by posting more. Diversity is computed as one minus the Herfindahl index over the accounts involved: 1 when every post comes from a different account, falling towards 0 as they concentrate in one.
The detail+–
A single account does not score zero, though. It scores 0.35 — weak evidence rather than no evidence. Zeroing it would hide the subject nobody else has noticed yet, which is occasionally the most valuable kind there is.
Why is the reach baseline different on every platform?
Because a view is a different act on each one. An autoplaying short is not a chosen watch, and treating ten thousand of each as the same event ranks whichever platform inflates hardest. Each post is measured against what counts as a lot on its own platform: 100,000 views on TikTok, 40,000 on Instagram, 30,000 on X, 20,000 on YouTube and 15,000 on Reddit.
The detail+–
Those five numbers are the ones most worth arguing with, which is why they are here rather than buried. If your category behaves differently from ours, you now know exactly which figure to disagree with.
What do the five bands mean, and what do I do in each?
Cold is under 20: attention is spread thin and nothing is moving. Cool is 20 to 39, warm is 40 to 59, hot is 60 to 79 and viral is 80 or above.
The detail+–
Hot and viral mean write into the named subject this week, because a surging subject has a window measured in days and the same piece published 30 days later is an archive entry. Warm means the strongest subject is gaining and has not peaked. Cool and cold are the calm windows for the evergreen pages that pay off slowly — reported as such rather than padded with something to do.
Why does the score depend on accounts I pick myself?
Because heat is only meaningful inside a category, and a category is defined by who it listens to. According to the Princeton and Georgia Tech researchers who measured generative search, the efficacy of visibility tactics "varies across domains, underscoring the need for domain-specific optimization methods" — the same holds for what counts as loud.
The detail+–
So the accounts are yours to choose, up to 25 read on each refresh, and they are filed into catalogs you name. Heat is read per catalog, so a subject running hot in one theme is not diluted by a quiet one next to it.
Is the score different before I have chosen any accounts?
Yes, and it says so on the screen. Before there are accounts to read, subjects are found by searching the platforms for your category's own phrases. That is a weaker claim than the one above — it says the platforms have something to show, not that the people you care about are talking about it — so it is scored differently rather than presented as the same number.
The detail+–
Source diversity is dropped entirely. A search returns different accounts by construction, so counting them would make the strongest signal in the account version meaningless here and every subject anybody looked up would come back hot. In its place: how much came back against how much was asked for, how far it travelled, how hard people engaged, and how recent it is — with volume gating the rest, because four posts are not a subject however well they did.
Recency counts for more in this version, at 45 per cent of the reading. When the sample is whatever a platform decided to return, a phrase whose results are all a month old has stopped moving — and that is the most useful thing a search can tell you, while the account version never has to ask it.
Also asked
- How recent does a post have to be to count? +
- Anything from the last 30 days. The first 7 days count in full, then a post decays in a straight line to nothing at the edge of the window.
- Can I compare Heat Scores between two of my catalogs? +
- Yes, and that is the comparison it is built for. Both are read against the same platform baselines, so a hot catalog and a cool one differ because the subjects differ, not because the scales do.
- Does a low Heat Score mean there is nothing to do? +
- No. It means the reactive work is not urgent, which makes it the right week for evergreen pages. A quiet category is reported as quiet rather than dressed up as an emergency.
- Is the score shown as a single number? +
- It is shown as a number, a band and the evidence behind it: how many posts, from how many separate accounts, on how many platforms. The number is the summary, not the argument.
Last updated 3 September 2026
Become the brand AI recommends.
Start with the free check. Your web address and nothing else.