playground / A/B test
Build a synthetic experiment in 60 seconds.
We'll run it through the same statistics that ship on every Winnow account.
Configure n, effect size, distribution shape, outliers, and missing-data pattern on the left. Watch the distribution and descriptive stats update live in the middle. Click "See how our statistics would conclude" on the right to pipe the dataset through Winnow's actual statistical/ Python module — the same mSPRT, Bayesian, and CUPED implementations that run on every customer experiment.
Stats run server-side on the real module· No login required· Synthetic data only — never uploaded
1. Configure
live
Sample size
n per arm5,000
allocation50 / 50
Outcome
metric typecontinuous
baseline mean280.0
effect size-7.0
distributionlognormal
Realism
5% of values replaced with 10× the mean — one giant order, one bot session
missing-data patternMAR
missing at random — explainable from observed data (e.g., fast runs drop)
missing rate12%
2. Synthetic data
8,800 rows · MAR 12% masked
control · n=4,400 · μ=539.6treatment · n=4,400 · μ=528.1
| stat | control | treatment | delta |
|---|---|---|---|
| n (after missing) | 4,400 | 4,400 | — |
| mean | 539.6 | 528.1 | -11.53 |
| median | 323.73 | 310.01 | -13.72 |
| std | 882.37 | 886.58 | — |
| IQR (25–75) | 228.5–483.5 | 220.2–476.2 | — |
| skew | 6.13 | 6.72 | heavy right tail |
| kurtosis (excess) | 50.94 | 62.42 | — |
| % missing | 12.0% | 12.0% | MAR — explainable |
| outliers (|z|>3) | 112 | 108 | — |
3. Test it
via /api/v1/playground/analyze
Click the button above to pipe your synthetic data through the real statistical module. We run mSPRT with auto-cure CUPED, the SRM check, skewness-aware normality floor, and surface what we'd decide.