- winner if posterior probability > 50%, unsuccessful otherwise
- Removes inconclusive — binary outcome based on Bayesian signal
- Label "Loser" renamed to "Unsuccessful"
- Updated all stored experiment verdicts in DB
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Change getWeekStartSunday to getISOWeekMonday for timestamp calculation
- Ensures week alignment between data aggregation and ISO week range calculation
- Fixes incorrect treatment_after average calculation
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Add ISO week calculation functions (getISOWeek, getISOWeekMonday, calculateISOWeekRange)
- Expand experiment dates to full ISO weeks (Monday-Sunday) instead of exact date ranges
- Use actual data.length for averaging metrics instead of theoretical durationWeeks
- Baseline now takes same number of complete weeks immediately before experiment
Fixes incorrect units sold average calculation (was 6.5, now correctly shows 4.33 for 13 units over 3 weeks)
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
The generic regex /sale/i in ads column detection was matching "Advertising Cost
of Sales" (the ACOS % column) before reaching "7 day total sales" (the actual
attributed sales amount). This caused attributedSales30d to receive the ACOS
percentage instead of the real sales value, making experiment ACOS = 0 or 100%.
- Replace /sale/i with specific regexes: /\d+\s*day.*sale/i, /total\s+sale/i, etc.
- Replace /cost/i with regexes that exclude "cost of sales" columns
- Replace /unit/i with specific regexes for unit columns
- Add diagnostic logging for ads Excel column headers and first-row values
- Revert ACOS 100% fallback (root cause now fixed)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Change ?? to || for salesAds/units/revenue in weekly aggregation to guard against NaN
- Add || 0 fallback to ads-only records in mergeSalesAndAdsData
- Fix parseCurrency to check isNaN on all EU-format return paths
- Handle ACOS edge case: cost > 0 with adRevenue = 0 now returns 100% instead of 0%
- Add diagnostic console logging to computeDiD for data flow tracing
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace the basic CRUD experiment tracker with a scientifically rigorous A/B testing system:
- Add DiD analysis engine (services/experimentAnalysis.ts) that computes treatment vs control
group comparisons across 7 metrics (units, sessions, CVR, CTR, ROAS, revenue, ACOS)
- Implement Bayesian verdict system (Winner/Loser/Inconclusive) using posterior probability
with normal CDF approximation (Abramowitz & Stegun erf, no external deps)
- Build counterfactual time series for trend charts (actual vs estimated without change)
- Rewrite ExperimentsView as single component with 3 inline sub-views (list, detail, create)
replacing the previous modal-based ExperimentDetail and ExperimentForm
- Add control group support, change annotations (before→after diffs), and SEO experiment type
- New types: ExperimentChangeAnnotation, DiDMetricResult, DifferenceInDifferencesResult,
ExperimentVerdict
- Simplify App.tsx by removing experiment modal state (5 useState hooks eliminated)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
