Surveys produce answers; decisions need numbers. The KPI system sits between the two. It scores incoming responses into metrics, colours each one against a benchmark you set, and puts them on one dashboard that updates as answers arrive. You can use the standard metrics as-is, design your own scoring, and combine existing metrics into new ones — without a formula editor or exporting anything.
How it fits together
There are three parts, and they stay separate on purpose.
Metrics are defined once per workspace in Settings → Metrics. A metric knows how answers should be scored and what counts as healthy. It doesn’t know or care which survey those answers come from.
Questions feed metrics. In the survey editor you open a question’s settings and tick the metrics it should count toward; you can also create a new metric right there without leaving the editor.
One survey can feed several metrics, and one metric can collect answers from many surveys. Because metrics live at the workspace level, a company-wide CSAT can pull from your post-support survey, your onboarding survey, and your yearly relationship survey at the same time — one definition, one trend line, no double bookkeeping.
The dashboard shows the result. Every metric gets a tile with its current score and a trend chart of the last 30 days, next to the responses coming in.
The standard metrics
The metrics most survey programs run on come as presets. Pick one, and it arrives configured: the right formula, the right answer scale, and a sensible benchmark. Everything about it stays editable.
Create New Metric Dialog
- NPS (Net Promoter Score) asks a 0-10 recommendation question and scores the percentage of promoters (9-10) minus the percentage of detractors (0-6). The result runs from -100 to +100, healthy at 50 or above.
- CSAT (customer satisfaction) is the share of satisfied answers — the top two boxes on a 1-5 satisfaction scale, shown as a percentage. Healthy at 80% or above.
- CES (customer effort) is the average of a 1-7 effort rating. Healthy at 5.5 or above.
- Completion rate — the percentage of people who open a survey and finish it — is measured automatically for every survey. There is nothing to set up.
If you want to size a survey before running it, the NPS calculator and CSAT calculator work through the maths, and the NPS template is a ready-made questionnaire.
Health bands instead of raw scores
A score on its own doesn’t tell you whether to act. Is an NPS of 31 fine? Every metric therefore carries three bands — healthy, warning, and needs attention — and everything that displays the score takes its colour from the band it sits in: green, amber, or red. You set the two thresholds by dragging two handles, so “good” means what it means for your business, not an industry average.
New Metric: NPS
The trend chart follows the same rule, and the line changes colour mid-chart the day your score crosses a threshold. When last quarter’s chart is green on the left and amber on the right, you can point to the day things changed.
Direction is configurable too. For most metrics higher is better; for effort, complaints, or churn it’s the opposite, and one switch flips which end of the scale is green.
Score answers your way
The presets are configurations, not special cases — you can build the same kind of metric from scratch. A custom metric scores either a rating scale or a plain number.
For a rating metric you choose the formula: net score (the NPS arithmetic on your own scale), percent positive or percent negative (top-box and bottom-box), average, or median. Then you mark which answers count as good and which as bad by clicking them:
Which answers are good or bad
A metric can accept several scale lengths at once, so a 1-5 question in one survey and a 0-10 question in another can feed the same score. Marking answers has a second effect: each response gets a sentiment. The response feed shows a face next to every answer and lets you filter to just the negative ones — useful when the score dips and you want to read why.
Number metrics track quantities instead of opinions — order value, ticket count, hours lost — using an average, a sum, or a simple count. Health bands are optional here; a number without a target just shows neutral.
Combined metrics
Any two metrics imply a third. Completions and opens imply a completion rate; expected and experienced quality imply a gap. Combined metrics let you define these in the same dialog by picking what you want to measure rather than writing a formula:
Create New Metric Dialog
- Rate / conversion — one metric divided by another, shown as a percentage. Opt-in rate, response rate, any part-of-a-total.
- Blended score — several metrics averaged into one index, each with its own importance weight. This is how composite indexes are built — see how to build a CX index for a worked example with published weights and benchmarks.
- Difference — one metric minus another. Expectation gaps, before-and-after comparisons.
- Worst / Best — the lowest or highest of several metrics. A floor that flags the moment any one input slips, which an average would hide.
Inputs are the metrics already in your workspace, and the output format follows the recipe — a rate becomes a percentage, and an index becomes a number. A combined metric behaves exactly like any other: its own tile, its own trend chart, its own health bands, recalculated the moment any input moves. The inputs keep their own tiles beside it, so when the number moves you can see which input moved it.
Live, from the first answer
KPIs read the same responses you see in the feed — there is no separate analytics pipeline and nothing to refresh. When a new response arrives, the tiles update. A metric that has no responses yet shows a dash and points you at the surveys it should be connected to, so a half-configured dashboard tells you what’s missing instead of showing zeros.
Metrics are included in every workspace. Start from a template, link a question, and the first response puts a number on the board.