The Sentiment page is the AI-powered analytical view of the firm’s feedback. Compass does sentiment classification and topic extraction on every feedback entry; this page turns those signals into a sentiment story.
What this page answers
What’s the firm’s overall feedback sentiment? Is it improving or declining over time? What topics are driving positive sentiment, and which drive negative? How does sentiment vary by segment? What’s the firm saying about its strengths and growth areas?
What’s on the page
Four headline cards
Net Sentiment Index — The headline NSI with period-over-period comparison
Positive Feedback — Count of positive entries
Neutral Feedback — Count of neutral entries
Negative Feedback — Count of negative entries
NSI ranges from −1 (uniformly negative) to +1 (uniformly positive): more positive feedback lifts the score, more negative feedback lowers it, and neutral feedback pulls it towards zero.
Sentiment Breakdown
A three-ring gauge chart with concentric rings for Positive (outer), Neutral (middle), Negative (inner). Each ring is sized by the share of feedback in that category.
Sentiment Trend
A line / area chart with three series (Positive / Neutral / Negative) showing daily counts. Useful for seeing whether the positive-to-negative ratio is shifting.
Net Sentiment Index Trend
A chart of NSI over time, with an area band showing the variability around the average. A rising line is improving sentiment; a band that widens is increasing variability (some segments much happier, others much less).
Trending Topics & Sentiment
A column chart with topics (AI-extracted aspects) as categories. For each topic, the columns show counts of positive / neutral / negative mentions. The pattern tells you which topics are sources of positive feedback (mostly green) vs which are problem topics (mostly red).
Net Sentiment by Segment
A tabbed table with Department / Role / Location tabs. Each tab lists segments with:
Feedback — Count of feedback given
Requests — Count of feedback requests
NSI — The segment’s Net Sentiment Index, colour-coded
The colour-coding is calibrated deliberately high: NSI below 0.3 is highlighted red (negative or low-positive — values within a whisker of zero are left unhighlighted), 0.3–0.8 gets no highlight, and 0.8 and above is highlighted green. A segment at 0.4 isn’t great even though it’s “positive” — the bar for green is meaningfully higher than zero.
Subjects Discussed
A tabbed table with Skill / Organization tabs. Each tab shows subjects mentioned in feedback with mention counts and a per-subject NSI. A skill or organisational topic with low NSI is one people are unhappy about.
Key Descriptor-Topic Pairs
Two word clouds in tabs (Strengths / Growth Areas). The Strengths cloud shows the descriptive terms most often paired with positive sentiment; the Growth Areas cloud shows those paired with negative sentiment. Word size scales with frequency.
For example, “clear” + “communication” might dominate Strengths while “lacking” + “delegation” might dominate Growth Areas.
Where the data comes from
Every chart and table on this page is built from Compass’s language analysis of your feedback:
Sentiment classification — The positive / neutral / negative read Compass produces for each analysed feedback entry
NSI — An aggregated score combining positive / neutral / negative counts
Aspect extraction — Topics (aspects) Compass found in the feedback text
Qualifiers — Descriptive terms Compass paired with aspects
Subjects — Skill and organisational subjects extracted
The analysis happens in the background as feedback entries are completed. Entries not yet analysed are excluded.
How to read it
NSI is the headline. Trend matters more than absolute level — is it going up or down over the period?
Trending topics is where insight lives. A topic with high positive mentions is something the firm is doing well at (worth celebrating and protecting); a topic with high negative mentions is what to address.
Strengths and growth-area word clouds are conversation starters. They condense thousands of feedback entries into the few terms people use most. Useful in town halls, leadership meetings, and culture work.
Subjects Discussed at low NSI is your action list. A topic that comes up a lot with negative sentiment is what to investigate first.
Why sentiment is reliable at scale but not individually
Sentiment classification is good but imperfect — for any single feedback entry, the AI might get the read wrong. At a cohort level, individual errors cancel out — the average sentiment of 1,000 entries is reliable. At an individual level, the sentiment of one entry might be wrong.
Use sentiment data at the segment, topic, or cohort level. Don’t use it to make decisions about an individual on the basis of one feedback entry.
Tips
Compare segments with the Net Sentiment by Segment table. Firm-wide NSI can hide a segment that’s markedly unhappy — negative NSI in a segment is rare but always meaningful.
Track the NSI trend, not the absolute value. What “good” looks like varies firm to firm — the trend tells you whether your firm is improving.
Topic mentions are a sample, not a census. Compass extracts topics from feedback text, so the topics you see reflect what people write about, not necessarily what they think about.
Empty page? If the analysis hasn’t run yet, the page shows a notice and returns early.