feat/ui-impact

portfolioStats.anomalias ignores date range — KPI misleads

apps/platform/src/app/[locale]/(dashboard)/energia/comparativo/page.tsx:212

anomalias calls analyzeStore() with no range argument and counts stores where deviationPct > 15 unconditionally. The result is identical regardless of what period the user selects. The memo depends on [inRange] but anomalias never uses it. The KPI sub-label implies range context ('>15% sobre su base'). Either scope the count to the selected range (filter anomalies whose breakoutYm falls within the range), or reword the label to be honest ('en el dataset completo'). As-is the number is a constant displayed as if it responds to the date picker.

Recharts Tooltip formatter returns [value, name] — store count won't render as intended

apps/platform/src/app/[locale]/(dashboard)/energia/comparativo/page.tsx:407

The portfolioChart Tooltip formatter returns an array: [fmtKwh(v), `${n} tiendas`]. Recharts interprets this as [displayValue, displayName] for the single 'total' series entry — the store count appears as the series name label, not as a separate row. The intent (two rows: kWh + store count) requires a custom <Tooltip content={...}> component. Currently the kWh value displays correctly but '${n} tiendas' shows in the wrong position/style.

Custom date range from > to silently empties all charts and tables

apps/platform/src/app/[locale]/(dashboard)/energia/comparativo/page.tsx:274

The two <input type='month'> fields have min/max HTML attributes but no cross-field validation. When customFrom > customTo, inRange becomes (ym) => ym >= from && ym <= to which is always false. Every chart and table renders empty and the empty-state message 'Sin datos en el periodo seleccionado' gives no hint about the inverted range. Minimal fix: inside resolveRange, return { from: min(from,to), to: max(from,to) } for the custom case, or show a visible warning.

Table rows keyed by array index — stale reconciliation on range filter change

apps/platform/src/app/[locale]/(dashboard)/energia/comparativo/page.tsx:456

Rows use key={i} (array index). When the date range changes and rows are filtered, React reuses DOM nodes by position, which can cause brief visual glitches. Use a stable composite key: `${r.Tienda}-${r.Periodo}` is unique per store+period in both scope modes.

fmtKwh duplicates fmtCfe — one should be removed

apps/platform/src/app/[locale]/(dashboard)/energia/comparativo/page.tsx:133

fmtCfe (line 44) and fmtKwh (line 133) are byte-for-byte identical: both produce Math.round(n).toLocaleString('es-MX') + ' kWh'. fmtKwh was added in this diff without noticing the existing formatter. Keep fmtKwh (clearer, domain-generic), remove fmtCfe, update the two call sites. Two identical formatters with different names will silently diverge if one is updated.

Single-letter memo variable `a` for anomaly result

apps/platform/src/app/[locale]/(dashboard)/energia/comparativo/page.tsx:155

const a = useMemo(() => analyzeStore(storeId), [storeId]) — used in 9 places. A descriptive name (storeAnalysis, analysis, or anomaly) communicates intent at each call site without needing to trace back to the declaration.

subMonths and resolveRange have no unit tests; year-boundary edge uncovered

apps/platform/src/app/[locale]/(dashboard)/energia/comparativo/page.tsx

subMonths is pure with a non-obvious arithmetic path: t = y*12 + (m-1) - n. At year boundaries (e.g. subMonths('2023-01', 1) → '2022-12') the calculation is correct, but if t ever goes negative (n > total months from year 0) JS's % operator returns a negative remainder, yielding a malformed month string. The codebase has a clear pattern of co-located unit tests for _lib functions (findingView.test.ts). A __tests__/comparativo.test.ts covering subMonths boundary cases and resolveRange branches would pin the behavior.

anomaly.ts pure functions (DiD engine) have no test file

apps/platform/src/app/[locale]/(dashboard)/energia/comparativo/_lib/anomaly.ts

analyzeStore contains the core DiD logic: cohort scoping, thin-cohort guard, baseline anchoring, breakoutYm detection (≥2 consecutive months). The anomaly score is the primary KPI surfaced to users. No __tests__/ counterpart exists. Given the codebase's testing pattern, a test file here is expected.

portfolioRows called redundantly in export button onClick

apps/platform/src/app/[locale]/(dashboard)/energia/comparativo/page.tsx:435

The 'Exportar portafolio' button calls portfolioRows(metric).filter(inRange) inline. When scope === 'portafolio', rows (already memoized) is already the range-filtered portfolio rows — reusing it avoids the redundant call. Only needed as a fallback when scope === 'sucursal' (to export the full portfolio regardless of the selected store).

anomalias constant inside a range-dependent memo — extract to module level

apps/platform/src/app/[locale]/(dashboard)/energia/comparativo/page.tsx:212

STORES.map(s => analyzeStore(s.id)).filter(r => r?.deviationPct > 15).length is a module-level constant (analyzeStore has no side effects, STORES is static). Moving it out of the portfolioStats useMemo into module scope makes the intent explicit, eliminates unnecessary recomputation on every range change, and decouples the anomaly count from the date-range dependency it doesn't actually have.