How to Read and Interpret Benchmarks Correctly
📅 2026-06-10
StatFacts insights follow one pattern: if you change X, Y moves by about Z—with a range, a confidence label, and linked sources. This guide explains the fields you see on every card.
The core fields
| Field | What it tells you |
|---|---|
intervention |
What changed (e.g. remove one signup step) |
outcome |
What was measured (e.g. signup completion) |
effect_label |
The headline range (e.g. +12–18%) |
sample_context |
Who, where, and when the number applied |
confidence |
How strong the evidence is |
sources |
Primary references to verify |
Effect labels
- +12–18% usually means a relative percent change (e.g. 20% → 23%).
- +8% with unit
percent_pointmeans percentage points (e.g. 30% → 38%).
Always check the insight detail page for effect_unit—mixing these up is the most common citation mistake.
Confidence at a glance
| Label | Meaning |
|---|---|
| Meta-analysis | Multiple studies aggregated |
| A/B test | Controlled experiment on real traffic |
| Study | Published research; causality not guaranteed |
| Estimate | Directional industry rule-of-thumb |
Sample context
sample_context is the guardrail. A benchmark from mobile B2B SaaS may not apply to your desktop marketplace. When context differs, treat the insight as a hypothesis, not a forecast.
Sources
We link primary references where possible. Open them before citing in investor decks, client proposals, or published content.
Next guides
Was this page helpful?
Your feedback helps us improve StatFacts