Compare Reviews is the cross-cycle tab. Where the Reviews tab analyses one cycle in depth and Employee Trajectory follows the long arc per person, Compare Reviews answers a sharper question: did things move between these specific cycles — and which way? Every chart on the tab is a delta.
No new AI work runs here. The tab reuses the per-cycle extraction Compass already produced for each cycle — performance level, sentiment, quality, development areas, themes, risk indicators — and computes the differences between the cycles you pick. That means the quality of the comparison directly inherits the quality of the underlying extraction.
What this tab answers
Did the cohort’s performance mix improve or slip between cycles? Is review sentiment trending up or down? Are managers writing better reviews over time? Are risk signals growing or receding?
Picking cycles to compare
Compare Reviews replaces the single cycle picker with a Cycles multi-select in the sidebar:
You start with two empty selectors. Pick a cycle in each — the list is grouped by cycle folder, and only cycles with completed reviews appear.
Add cycle adds another selector, so you can compare three, four, or more cycles.
A remove (×) button appears next to each selector once you have more than two, so you can never drop below two slots.
A cycle already chosen in one selector is greyed out in the others — no duplicates.
Until you’ve selected at least two cycles, the pages show a “Select cycles to compare” prompt and the sidebar reminds you to Select at least 2 cycles to compare. Every change re-runs the view immediately.
Selection order doesn’t matter: cycles are always ordered chronologically by start date, so charts read left-to-right as time-ordered transitions.
The workspace Department, Role, and Location filters apply here too — narrow the comparison to a segment before reading the deltas. There’s no date-range filter on this tab; the cycles themselves define the time window.
The views
The sidebar offers five views: Introduction (an in-app explainer of how cycle comparison works), Overview (this article), Movement, Development, and Managers — each covered in its own article.
If a selected cycle hasn’t been through AI extraction yet, you’ll see a “No AI Extraction Data” notice. Extraction runs automatically in the background, so the page fills in once those cycles have been analysed.
What’s on the Overview page
Two headline KPIs (top row)
Cycles Compared — How many cycles are in the current comparison.
Strong+ Change — The percentage-point change in the share of the cohort rated Strong or Exceptional between the earliest and latest selected cycle. Positive (with an upward trend icon) means the top of the distribution grew; negative means it shrank.
Avg sentiment over time
A bar chart of the average sentiment score per cycle (−1 to +1). This is the fastest read on whether the overall tone of reviews is becoming more supportive or more critical across cycles.
Avg review quality over time
A bar chart of the average overall quality score per cycle (0–1). Rising quality usually means managers are writing richer, more specific reviews — which also makes every other signal on this tab more trustworthy.
Performance composition over time
A 100% stacked bar chart showing, for each cycle, the proportion of the cohort at each performance level: Exceptional, Strong, Solid, Concerning, Critical, and Insufficient Data. Because each bar totals 100%, you’re comparing shape, not headcount — useful when cycle populations differ.
Risk signals over time
A column chart with two series per cycle: Total Signals (all risk indicators Compass extracted — retention, disengagement, burnout and similar) and High Severity (the subset flagged high). The card only appears if at least one risk signal exists across the selected cycles.
Where the data comes from
Everything on this page is aggregated live from the per-cycle review extraction results of the cycles you selected. Nothing is recomputed by AI at comparison time — the page lines the stored per-cycle numbers up next to each other.
How to read it
Start with Strong+ Change and the composition chart together. A positive Strong+ Change with a thinning Concerning/Critical band is the healthy pattern; a positive change driven only by Solid people vanishing needs a population check.
Read sentiment and quality as a pair. If sentiment fell but quality rose, reviewers may simply be writing more honestly — that’s a calibration story, not a performance one.
Rising risk totals with flat performance is an early-warning pattern worth drilling into before it shows up in ratings.
Tips
Check the population before trusting a delta. People who appear in only one of the cycles shift the distributions without representing real movement — joiners and leavers can masquerade as change.
Watch for scale or reviewer changes between cycles. If the rating scale changed, or different managers wrote the reviews, some movement is drift rather than performance.
Filter to a department to make deltas actionable. A firm-wide shift is interesting; a department-level shift tells you where to act.
Use the Movement view next. The Overview tells you the distributions changed; Movement shows you exactly who moved where.