The fixed reports answer common questions. The Custom tab answers yours. It's a small chart builder: choose what to group candidates by, what to measure, and which candidates to include, and Chosen draws the result. This page covers building a query and how literally to take the short "insight" lines that appear on some reports.
Building a query
The builder has three choices and an optional filter list.
- Group by a dimension — the categories along the bottom of your chart. Built-in options include source, stage, department, location, job type, experience level, and created or applied month. Any custom field you've added shows up here too, under a separate group.
- Pick a metric — the number being measured. Choices are count, average match score, average feedback, average experience in months, and hire rate. Numeric custom fields add their own average, sum, min, and max.
- Add filters, if you need them — up to five. Each is a field, an operator, and a value, and they narrow which candidates the chart counts.
- Choose a chart type — bar, line, pie, or donut — and select Generate.
A worked example: group by source, measure hire rate, filter to one department. You get hire rate per source for that team — a question no fixed report answers exactly. Swap the metric to count and you have applications per source instead.
Reading the results
The generated chart is a view, not a saved report — it lives until you change a choice or leave the tab, so a query you'll want again is worth rebuilding or noting down. Pick the chart type to match the question: bar or line for many categories, pie or donut for a handful of parts that sum to a whole.
The same caution from every other report applies here, and the builder makes it easier to trip over. Filter hard enough and you can group three candidates into five buckets — a chart that looks precise and means nothing. Watch the volume behind each bar before you trust its height.
About the "insight" lines
A few reports show a short sentence near the chart — Score vs Outcome, for instance, may say something like "candidates in later stages score 12% higher." It's fair to wonder how much weight those carry.
Be plain about it: these lines are computed, not authored. Chosen compares two figures in the report and states the difference as a sentence. There's no language model writing them and no analyst behind them — it's arithmetic, formatted to read like prose. That makes them reliable as far as they go, and shallow. The line tells you that later-stage candidates score higher; it doesn't know whether your match score is predictive, whether the sample is large enough to mean anything, or what you should do about it. Read an insight line as a pointer to look at the chart, never as the conclusion.
If you want an actual interpretation — "why is this happening, what should I change" — that's a question for HQ, the AI assistant, which can reason over your data and explain itself. The insight lines are signposts; HQ is the analyst.
Limits
The builder covers a fixed set of dimensions and metrics — broad, but not everything, and it won't join data the way a spreadsheet or BI tool would. A generated chart isn't saved or scheduled; it's there until you navigate away. Custom-field options only appear once the field exists and has data behind it. And the builder will happily chart a sample too small to trust — the honesty check is still yours. To get a chart out of Chosen, use the per-chart CSV or PNG export; for raw candidate data, see data and exports.