Topic Detection clusters every post, review and chat into themes, sized by volume, scored by confidence, tracked month over month. No tagging, no taxonomy meetings. Just the picture.
Every post is read, embedded and assigned a place in a two-tier hierarchy. Top-level Topics for the picture; Subtopics underneath for the why. Each classification carries a confidence score.
Big enough to cover any product. Small enough that you can read the whole map in 10 seconds.
Every classification carries a 0–1 score. Below 0.7 stays in a review queue. Above 0.85 publishes automatically.
New categories emerge, dead ones fold in. The map reflects what your customers are saying this month, not last quarter.
Every topic and subtopic has a year of history attached. Spot the long curve climbing weeks before it spikes in support. Spot a complaint quietly dying because you actually shipped the fix.
A 12-month area showing post count per week. Hover to see which release, ad campaign or competitor move kicked it.
Three lines on one chart: raw posts, net votes (up minus down), and aggregate sentiment. Same scale, same week.
Pin one topic, overlay another. See whether your shiny new feature stole all the oxygen from another conversation.
Each theme shows representative quotes, vote weight, sentiment polarity and the confidence score the model assigns it. Use this to tell whether a cluster's a coherent ask or a fuzzy edge.
Three real (anonymised) lines per theme, picked from the median of each cluster. The voice, not the abstraction.
Each theme carries an aggregate sentiment from –100% to +100%, weighted by recency. The vote bar shows it.
A topic-level confidence score tells you which clusters are tight (high-confidence) versus fuzzy edges still worth a glance.
Detect, automate, analyse, agent, MCP: the full Sleek Intelligence stack. Topic Detection is the analytics layer. Wire on the rest for an end-to-end portal that runs itself.
When a subtopic jumps 40% week-over-week, that's tomorrow's ticket flood. Sleekplan surfaces it now, before your CS team is buried.
Stop arguing about which feature 'feels' important. Compare topic volumes and growth side-by-side; ship what the numbers want.
A theme is growing. Open the topic, scan the subtopics, read the typical requests. Now you know what to scope.
Themes are already there, charts are already drawn. Pull a screenshot, drop it in the board update. No more 'someone needs to categorise the last 200 posts'.
Topic Detection is on every Sleek Intelligence plan, with a year of history out of the box. Flip it on, scroll the map.