Segment Analysis

Segment Analysis compares scores across groups of candidates rather than across tests or questions — the report for "did our New York cohort outperform Chicago" or "how did the 'Referral' label compare to 'Direct Apply'." It buckets your candidates' results by group, label, or tag and shows the score statistics for each bucket side by side.

Choosing What to Compare

• Test (required) — pick one or more tests before anything else. This isn't optional: averaging percentage scores across different tests would blend unrelated scales into a meaningless number, so the report always needs a specific test (or tests) to compare within.
• Task — optional, narrows further.
• Date range — optional, filters by session date.

Segmenting: Group By / Then By

Pick a primary dimension to bucket by — Group, Label, Tag (choose which tag key), or Task — and optionally a second dimension to sub-divide each bucket further, producing a two-level comparison table instead of one flat list. Choosing the same dimension twice, or the same tag key twice, isn't allowed since it wouldn't produce a meaningful split.

Group and Task give each candidate exactly one bucket. Labels are different: a candidate can carry several labels at once, so a single result can count toward more than one label bucket — the totals across label buckets won't necessarily add up to your overall submission count, and the report notes this on screen.

Reading the Table

Each bucket (or each primary × secondary pair, if you set a second dimension) gets its own row: Submitted (count), Average Score, Minimum, Maximum, and Standard Deviation — letting you spot not just which segment scored higher on average, but which one was more consistent.

A segment with very few submissions can show an extreme average purely by chance — check the Submitted count alongside the average before treating a segment comparison as meaningful.

Good to know