Reading results

Results compare each variant’s conversions against its exposures and report which one is winning. Read them in the dashboard or over the API.

Fetch results

GET /v1/experiments/:id returns the analysis. The :id accepts the experiment id or its key.

curl https://statskit.ai/api/v1/experiments/hero-headline-test-a1b2 \
  -H "Authorization: Bearer vsk_live_…"

{
  "experiment": "hero-headline-test-a1b2",
  "goal": "signup",
  "status": "running",
  "variants": [
    {
      "key": "control",
      "exposed": 4120,
      "conversions": 214,
      "uplift": 0,
      "significance": null,
      "winner": false
    },
    {
      "key": "b",
      "exposed": 4090,
      "conversions": 267,
      "uplift": 25.8,
      "significance": 0.97,
      "winner": true
    }
  ],
  "metrics": []
}

What the numbers mean

exposed: visitors put into that variant (unique exposures).
conversions: visitors who fired the goal event after being exposed.
uplift %: the variant’s conversion rate relative to control. Control’s uplift is 0 by definition; a positive number means the variant converts better.
significance: how confident the result is. Under the bayesian engine this reads as the probability the variant beats control (higher is stronger). Under the frequentist engine you get a p-value instead (lower is stronger).
winner: flagged on the leading variant once the result clears the engine’s threshold. Until then no variant is marked, which usually means you need more data.

Bayesian vs frequentist

The engine was chosen at creation (statsEngine, default bayesian) and decides how these fields are computed:

bayesian: reports the probability a variant beats control. Intuitive to read mid-flight (“b has a 97% chance to win”).
frequentist: reports a p-value and significance against control, the classic hypothesis test.

Attach custom metrics

Beyond the primary goal, you can attach custom metrics to an experiment as a primary metric and up to ten guardrails (metrics you want to make sure a variant does not hurt). Each is analysed direction-aware and returned in the metrics array of the results.

PUT /v1/experiments/:id/metrics replaces the attached set. Reference metrics by key or id.

curl -X PUT https://statskit.ai/api/v1/experiments/hero-headline-test-a1b2/metrics \
  -H "Authorization: Bearer vsk_live_…" \
  -H "Content-Type: application/json" \
  -d '{
    "primary": "activation_rate",
    "guardrails": ["refund_rate", "support_tickets"]
  }'

In the dashboard

The dashboard shows the same analysis with per-variant rates, uplift, the significance bar, and the winner call, plus any attached custom metrics.