fix(insights): honest discovery footer, commensurable better-days ranking, settings tap-target + prop cleanup

- Mood discovery footer no longer pairs a mood-only numerator with the
  full behaviour-by-outcome denominator. Reword it to surface the mood
  relationships out of all day-to-day pairs the engine tested across
  every metric, so the count it reports is honest about its scope.
  Update all six locales.
- Rank the better-days board on a comparable scale: standardize the tag
  mood-point delta into a Cohen's-d effect (delta over the pooled
  with/without mood SD, clamped to [0,1]) so it sits alongside a metric
  |r| instead of the old |delta|/2 heuristic. Fall back to |delta|/2 only
  when the pooled SD is undefined. The pooled SD is ranking-only and is
  stripped before the row enters the snapshot payload; the raw delta and
  r the cards show are unchanged.
- Normalize the Withings and Whoop resume buttons to the min-h-11 scale
  token, dropping the arbitrary min-h-[44px] spelling.
- Give the API token copy button a 44px tap target on small viewports,
  scaling back to the tighter icon size from sm up.
- Drop the dead cached prop from InsightStatusCard and every caller; the
  card has carried no cached affordance since the badge was removed.

messages/de.json

@@ -2564,7 +2564,7 @@
       "discovery": {
         "title": "Entdeckte Muster",
         "description": "Tag-zu-Tag-Zusammenhänge rund um deine Stimmung, die auch nach Korrektur für viele gleichzeitige Tests bestehen.",
-        "footer": "{discovered} von {tested} getesteten Zusammenhängen haben die Korrektur für Fehlentdeckungen überstanden.",
+        "footer": "Herausgefiltert aus {tested} Tag-zu-Tag-Paaren, die über all deine Werte getestet wurden, nach der Korrektur für Fehlentdeckungen.",
         "empty": "Bisher hat kein Stimmungszusammenhang die statistische Hürde genommen. Logge weiter.",
         "disclaimer": "Statistisch gefiltert, aber weiterhin beobachtend – ein Muster, das einen Blick wert ist, keine Ursache.",
         "moodUp": "Höheres {factor} geht in deinen Daten tendenziell mit besserer Stimmung am Folgetag einher.",

messages/en.json

@@ -2564,7 +2564,7 @@
       "discovery": {
         "title": "Discovered patterns",
         "description": "Day-to-day relationships involving your mood that hold up after correcting for testing many pairs at once.",
-        "footer": "{discovered} of {tested} relationships tested survived the false-discovery correction.",
+        "footer": "Surfaced from {tested} day-to-day pairs tested across all your metrics, after the false-discovery correction.",
         "empty": "No mood relationships cleared the statistical bar yet. Keep logging.",
         "disclaimer": "Statistically filtered, but still observational — a pattern worth watching, not a cause.",
         "moodUp": "Higher {factor} tends to go with higher next-day mood in your data.",

messages/es.json

@@ -2564,7 +2564,7 @@
       "discovery": {
         "title": "Patrones descubiertos",
         "description": "Relaciones de un día a otro con tu estado de ánimo que se mantienen tras corregir el haber probado muchos pares a la vez.",
-        "footer": "{discovered} de {tested} relaciones probadas superaron la corrección de falsos descubrimientos.",
+        "footer": "Extraídas de {tested} pares de un día a otro probados en todas tus métricas, tras la corrección de falsos descubrimientos.",
         "empty": "Todavía ninguna relación de ánimo superó el umbral estadístico. Sigue registrando.",
         "disclaimer": "Filtrado estadísticamente, pero aún observacional: un patrón que vale la pena observar, no una causa.",
         "moodUp": "Un mayor {factor} suele coincidir con mejor ánimo al día siguiente en tus datos.",

messages/fr.json

@@ -2564,7 +2564,7 @@
       "discovery": {
         "title": "Tendances détectées",
         "description": "Des relations d'un jour à l'autre autour de votre humeur qui tiennent après correction des tests multiples.",
-        "footer": "{discovered} relations sur {tested} testées ont passé la correction des fausses découvertes.",
+        "footer": "Issues de {tested} paires d'un jour à l'autre testées sur l'ensemble de vos mesures, après correction des fausses découvertes.",
         "empty": "Aucune relation d'humeur n'a encore franchi le seuil statistique. Continuez à enregistrer.",
         "disclaimer": "Filtré statistiquement, mais toujours observationnel : une tendance à surveiller, pas une cause.",
         "moodUp": "Un {factor} plus élevé tend à coïncider avec une meilleure humeur le lendemain dans vos données.",

messages/it.json

@@ -2564,7 +2564,7 @@
       "discovery": {
         "title": "Schemi rilevati",
         "description": "Relazioni giorno per giorno legate al tuo umore che reggono dopo aver corretto per i test multipli.",
-        "footer": "{discovered} relazioni su {tested} testate hanno superato la correzione per le false scoperte.",
+        "footer": "Estratte da {tested} coppie giorno per giorno testate su tutte le tue misure, dopo la correzione per le false scoperte.",
         "empty": "Nessuna relazione sull'umore ha ancora superato la soglia statistica. Continua a registrare.",
         "disclaimer": "Filtrato statisticamente, ma sempre osservazionale: uno schema da tenere d'occhio, non una causa.",
         "moodUp": "Un {factor} più alto tende a coincidere con un umore migliore il giorno dopo nei tuoi dati.",

messages/pl.json

@@ -2564,7 +2564,7 @@
       "discovery": {
         "title": "Wykryte wzorce",
         "description": "Relacje dzień po dniu dotyczące Twojego nastroju, które utrzymują się po korekcie wielokrotnych testów.",
-        "footer": "{discovered} z {tested} testowanych relacji przeszło korektę fałszywych odkryć.",
+        "footer": "Wyłonione z {tested} par dzień po dniu testowanych na wszystkich Twoich wskaźnikach, po korekcie fałszywych odkryć.",
         "empty": "Żadna relacja nastroju nie przekroczyła jeszcze progu statystycznego. Notuj dalej.",
         "disclaimer": "Przefiltrowane statystycznie, ale wciąż obserwacyjne — wzorzec warty uwagi, nie przyczyna.",
         "moodUp": "Wyższy {factor} zwykle wiąże się z lepszym nastrojem następnego dnia w Twoich danych.",

src/app/insights/blood-pressure/page.tsx

@@ -138,7 +138,6 @@ export default function InsightsBlutdruckPage() {
         icon={<HeartPulse className="h-5 w-5" />}
         text={status?.text ?? null}
         hasProvider={status?.hasProvider ?? false}
-        cached={status?.cached ?? false}
         updatedAt={status?.updatedAt ?? null}
         loading={isStatusLoading}
         preparing={status?.preparing ?? false}

src/app/insights/bmi/page.tsx

@@ -141,7 +141,6 @@ export default function InsightsBmiPage() {
         icon={<Ruler className="h-5 w-5" />}
         text={status?.text ?? null}
         hasProvider={status?.hasProvider ?? false}
-        cached={status?.cached ?? false}
         updatedAt={status?.updatedAt ?? null}
         loading={isStatusLoading}
         preparing={status?.preparing ?? false}

src/app/insights/medications/page.tsx

@@ -288,7 +288,6 @@ export default function InsightsMedikamentePage() {
         icon={<Pill className="h-5 w-5" />}
         text={status?.summary ?? null}
         hasProvider={status?.hasProvider ?? false}
-        cached={status?.cached ?? false}
         updatedAt={status?.updatedAt ?? null}
         loading={isStatusLoading}
         preparing={status?.preparing ?? false}

src/app/insights/mood/page.tsx

@@ -118,7 +118,6 @@ export default function InsightsStimmungPage() {
         icon={<Smile className="h-5 w-5" />}
         text={status?.text ?? null}
         hasProvider={status?.hasProvider ?? false}
-        cached={status?.cached ?? false}
         updatedAt={status?.updatedAt ?? null}
         loading={isStatusLoading}
         preparing={status?.preparing ?? false}

src/app/insights/pulse/page.tsx

@@ -151,7 +151,6 @@ export default function InsightsPulsPage() {
         icon={<Heart className="h-5 w-5" />}
         text={status?.text ?? null}
         hasProvider={status?.hasProvider ?? false}
-        cached={status?.cached ?? false}
         updatedAt={status?.updatedAt ?? null}
         loading={isStatusLoading}
         preparing={status?.preparing ?? false}

src/app/insights/weight/page.tsx

@@ -116,7 +116,6 @@ export default function InsightsGewichtPage() {
         icon={<Scale className="h-5 w-5" />}
         text={status?.text ?? null}
         hasProvider={status?.hasProvider ?? false}
-        cached={status?.cached ?? false}
         updatedAt={status?.updatedAt ?? null}
         loading={isStatusLoading}
         preparing={status?.preparing ?? false}

src/components/insights/__tests__/insight-status-card-flag-gate.test.tsx

@@ -24,7 +24,6 @@ const baseProps = {
   icon: null,
   text: "Your pulse is stable.",
   hasProvider: true,
-  cached: false,
   updatedAt: null,
 };
 

src/components/insights/__tests__/insight-status-card.test.tsx

@@ -24,7 +24,6 @@ const baseProps = {
   title: "Pulse",
   icon: null,
   hasProvider: true,
-  cached: false,
   updatedAt: null,
 };
 
@@ -90,14 +89,14 @@ describe("<InsightStatusCard>", () => {
     expect(html).toContain("No analysis yet.");
   });
 
-  it("never surfaces a cached badge, even when the text came from the warm cache", () => {
-    // v1.11.5 — the top-right "cached" label was removed. The caching
-    // behaviour is unchanged (the prop is still threaded), but the card
-    // no longer announces it, so the assessment reads as authoritative.
+  it("never surfaces a cached badge — the card has no cached affordance", () => {
+    // v1.11.5 — the top-right "cached" label was removed and the dead
+    // `cached` prop was dropped from the contract. The caching behaviour
+    // upstream is unchanged; the card simply never announced it, so the
+    // assessment reads as authoritative.
     const html = render(
       <InsightStatusCard
         {...baseProps}
-        cached
         text="Pulse stayed inside the band."
       />,
     );

src/components/insights/insight-status-card.tsx

@@ -15,14 +15,6 @@ interface InsightStatusCardProps {
   icon: React.ReactNode;
   text: string | null;
   hasProvider: boolean;
-  /**
-   * Whether the rendered text came from the warm cache rather than a
-   * fresh generation. Retained on the contract so every caller can keep
-   * threading it, but the card no longer surfaces a "cached" badge — the
-   * label devalued the assessment and the caching is an implementation
-   * detail the user does not need to see.
-   */
-  cached: boolean;
   updatedAt: string | null;
   loading?: boolean;
   /**

src/components/insights/metric-status-card.tsx

@@ -43,7 +43,6 @@ export function MetricStatusCard({
       icon={icon}
       text={status?.text ?? null}
       hasProvider={status?.hasProvider ?? false}
-      cached={status?.cached ?? false}
       updatedAt={status?.updatedAt ?? null}
       loading={isLoading}
       preparing={status?.preparing ?? false}

src/components/insights/mood/__tests__/mood-discovered-relations.test.tsx

@@ -111,7 +111,7 @@ describe("<MoodDiscoveredRelations>", () => {
     expect(html).toContain("lower next-day");
   });
 
-  it("filters to mood pairs and shows the honest tested footer", () => {
+  it("filters to mood pairs and shows the honest full-family footer", () => {
     const html = renderWith(
       [
         pair({ behaviour: "TIME_IN_DAYLIGHT", outcome: "MOOD" }),
@@ -122,7 +122,8 @@ describe("<MoodDiscoveredRelations>", () => {
     expect((html.match(/data-slot="mood-discovered-pair"/g) ?? []).length).toBe(
       1,
     );
-    // footer reports 1 mood pair of 12 tested
-    expect(html).toContain("1 of 12");
+    // Footer's denominator is the full behaviour×outcome family the engine
+    // tested, not the mood-only subset — so it must report 12, not 1 of 1.
+    expect(html).toContain("from 12 day-to-day pairs");
   });
 });

src/components/insights/mood/mood-discovered-relations.tsx

@@ -124,7 +124,6 @@ export function MoodDiscoveredRelations() {
       </ul>
       <p className="text-muted-foreground text-[11px]">
         {t("insights.mood.discovery.footer", {
-          discovered: moodPairs.length,
           tested: data.pairsTested,
         })}
       </p>

src/components/settings/api-section.tsx

@@ -303,7 +303,7 @@ function ApiTokensCard() {
                 type="button"
                 variant="outline"
                 size="icon"
-                className="h-9 w-9 shrink-0"
+                className="min-h-11 min-w-11 shrink-0 sm:h-9 sm:w-9"
                 onClick={() => void handleCopyToken()}
                 aria-label={t("settings.tokenCopy")}
                 data-slot="settings-api-token-copy"

src/components/settings/integrations-section.tsx

@@ -464,7 +464,7 @@ function WithingsCard({
               onClick={() => resume.mutate()}
               disabled={resume.isPending}
               data-testid="withings-resume-button"
-              className="min-h-[44px] sm:min-h-0"
+              className="min-h-11"
             >
               {resume.isPending ? (
                 <Loader2 className="mr-1 h-3.5 w-3.5 animate-spin motion-reduce:animate-none" />
@@ -897,7 +897,7 @@ function WhoopCard({
               onClick={() => resume.mutate()}
               disabled={resume.isPending}
               data-testid="whoop-resume-button"
-              className="min-h-[44px] sm:min-h-0"
+              className="min-h-11"
             >
               {resume.isPending ? (
                 <Loader2 className="mr-1 h-3.5 w-3.5 animate-spin motion-reduce:animate-none" />

src/lib/insights/__tests__/mood-aggregates.test.ts

@@ -638,7 +638,8 @@ describe("computeBetterDays", () => {
           withoutDays: 10,
           withAvg: 4.4,
           withoutAvg: 2.4,
-          delta: 2, // effect 1.0 (capped)
+          delta: 2,
+          pooledSd: 1, // |d| = 2 → standardized effect 1.0 (capped)
           pValue: 0.001,
           confidence: "high",
         },
@@ -680,6 +681,7 @@ describe("computeBetterDays", () => {
           withAvg: 2.2,
           withoutAvg: 3.8,
           delta: -1.6,
+          pooledSd: 0.9,
           pValue: 0.01,
           confidence: "medium",
         },
@@ -698,6 +700,88 @@ describe("computeBetterDays", () => {
     expect(board[0].delta).toBe(-1.6);
   });
 
+  it("ranks tags by the standardized (Cohen's-d) effect, not the raw delta", () => {
+    // "tight" has the smaller raw delta but a far smaller pooled SD, so its
+    // standardized effect (0.8 / 0.4 = 2.0 → capped 1.0) beats "wide"
+    // (1.2 / 2.0 = 0.6). The legacy |delta|/2 heuristic would have ordered
+    // them the other way (0.6 vs 0.4), so this pins the new behaviour.
+    const tagInfluence: TagInfluence = {
+      flat: [
+        {
+          tag: "wide",
+          labelKey: null,
+          categoryKey: null,
+          icon: null,
+          withDays: 10,
+          withoutDays: 10,
+          withAvg: 4.2,
+          withoutAvg: 3.0,
+          delta: 1.2,
+          pooledSd: 2.0,
+          pValue: 0.02,
+          confidence: "medium",
+        },
+        {
+          tag: "tight",
+          labelKey: null,
+          categoryKey: null,
+          icon: null,
+          withDays: 10,
+          withoutDays: 10,
+          withAvg: 4.0,
+          withoutAvg: 3.2,
+          delta: 0.8,
+          pooledSd: 0.4,
+          pValue: 0.001,
+          confidence: "high",
+        },
+      ],
+      structured: [],
+    };
+    const board = computeBetterDays(tagInfluence, {
+      sleep: emptyCorr,
+      steps: emptyCorr,
+      pulse: emptyCorr,
+      weight: emptyCorr,
+      bloodPressureSystolic: emptyCorr,
+    });
+    expect(board.map((f) => f.key)).toEqual(["tight", "wide"]);
+    // Raw deltas shown in the UI are untouched by the ranking change.
+    expect(board[0].delta).toBe(0.8);
+    expect(board[1].delta).toBe(1.2);
+  });
+
+  it("falls back to the |delta|/2 heuristic when a tag has no pooled SD", () => {
+    const tagInfluence: TagInfluence = {
+      flat: [
+        {
+          tag: "constant",
+          labelKey: null,
+          categoryKey: null,
+          icon: null,
+          withDays: 10,
+          withoutDays: 10,
+          withAvg: 4.0,
+          withoutAvg: 2.0,
+          delta: 2,
+          pooledSd: null, // both groups perfectly constant
+          pValue: 1,
+          confidence: "low",
+        },
+      ],
+      structured: [],
+    };
+    const board = computeBetterDays(tagInfluence, {
+      sleep: emptyCorr,
+      steps: emptyCorr,
+      pulse: emptyCorr,
+      weight: emptyCorr,
+      bloodPressureSystolic: emptyCorr,
+    });
+    expect(board).toHaveLength(1);
+    expect(board[0].effectSize).toBe(1); // min(1, |2|/2)
+  });
+
   it("returns an empty board when nothing clears the gates", () => {
     const board = computeBetterDays(
       { flat: [], structured: [] },
@@ -723,6 +807,7 @@ describe("computeBetterDays", () => {
       withAvg: 4,
       withoutAvg: 3,
       delta: 1 - i * 0.01,
+      pooledSd: 1,
       pValue: 0.02,
       confidence: "low" as const,
     }));

src/lib/insights/mood-aggregates.ts

@@ -313,6 +313,13 @@ export interface TagInfluenceRow {
   withoutAvg: number;
   /** withAvg − withoutAvg, rounded. Positive = higher mood with the tag. */
   delta: number;
+  /**
+   * Pooled mood SD across the with/without groups (Cohen's-d denominator).
+   * `null` when neither group has testable spread (both constant). Lets the
+   * board standardize the raw delta into a unitless effect comparable to a
+   * Pearson |r|. Not surfaced in the UI — ranking only.
+   */
+  pooledSd: number | null;
   /** Welch two-sided p-value for the difference of means. */
   pValue: number;
   /** Discrete confidence band derived from p-value + per-group sample size. */
@@ -431,6 +438,22 @@ function influenceForTag(
   const delta = withAvg - withoutAvg;
   const minGroupDays = Math.min(withVals.length, withoutVals.length);
 
+  // Pooled SD (Cohen's-d denominator) so the board can standardize this raw
+  // mood-point delta into a unitless effect comparable to a Pearson |r|.
+  // Unbiased (n−1) per-group variances pooled on (nWith + nWithout − 2) df;
+  // null when there is no testable spread (both groups perfectly constant).
+  const nWith = withVals.length;
+  const nWithout = withoutVals.length;
+  const varWith =
+    withVals.reduce((s, v) => s + (v - withAvg) ** 2, 0) / (nWith - 1);
+  const varWithout =
+    withoutVals.reduce((s, v) => s + (v - withoutAvg) ** 2, 0) /
+    (nWithout - 1);
+  const pooledVar =
+    ((nWith - 1) * varWith + (nWithout - 1) * varWithout) /
+    (nWith + nWithout - 2);
+  const pooledSd = pooledVar > 0 ? Math.sqrt(pooledVar) : null;
+
   const pValue = welch.status === "ok" ? welch.pValue : 1;
   // A zero delta with no spread is not an "influence" — drop it so the
   // board never shows a 0.0 row.
@@ -446,6 +469,7 @@ function influenceForTag(
     withAvg: round(withAvg, 2),
     withoutAvg: round(withoutAvg, 2),
     delta: round(delta, 2),
+    pooledSd: pooledSd === null ? null : round(pooledSd, 3),
     pValue,
     confidence: influenceConfidence(pValue, minGroupDays),
   };
@@ -586,8 +610,16 @@ function metricConfidence(
  *    non-"keine" strength (|r| ≥ 0.2) are folded in — a near-zero r is not
  *    an association.
  *
- * Ranking is by `effectSize` desc (|r| for metrics, |delta|/2 capped for
- * tags), then by confidence, then key for stability. Observational only.
+ * Ranking is by `effectSize` desc, then confidence, then key for stability.
+ * The two sources are put on one comparable scale: metric rows use |r|
+ * (already unitless in [0,1]); tag rows use a standardized effect — the raw
+ * mood-point `delta` divided by the pooled with/without mood SD (Cohen's d),
+ * clamped to [0,1]. A |d| ≥ 1 (a mean shift of one full SD) saturates the
+ * scale, mirroring an |r| near its ceiling, so neither source dominates the
+ * board for scale reasons alone. When a tag has no pooled SD (both groups
+ * perfectly constant), it falls back to the legacy |delta|/2 heuristic — a
+ * rare degenerate case that can't be standardized. The raw delta / r the UI
+ * shows is unchanged; this only governs the sort order. Observational only.
  */
 export function computeBetterDays(
   tagInfluence: TagInfluence,
@@ -606,7 +638,13 @@ export function computeBetterDays(
       direction: row.delta >= 0 ? "up" : "down",
       n: Math.min(row.withDays, row.withoutDays),
       confidence: row.confidence,
-      effectSize: Math.min(1, Math.abs(row.delta) / 2),
+      // Cohen's-d standardization so the tag effect is commensurable with a
+      // metric |r|; fall back to the legacy |delta|/2 only when the pooled
+      // SD is unavailable (both groups perfectly constant).
+      effectSize:
+        row.pooledSd && row.pooledSd > 0
+          ? Math.min(1, Math.abs(row.delta) / row.pooledSd)
+          : Math.min(1, Math.abs(row.delta) / 2),
       delta: row.delta,
       r: null,
     });

src/lib/insights/mood-status.ts

@@ -8,6 +8,7 @@ import {
 import { pearsonCorrelation } from "@/lib/analytics/correlations";
 import { applyPayloadBudget } from "@/lib/insights/bucket-series";
 import {
+  type TagInfluenceRow,
   MOOD_GREEN_MAX,
   MOOD_GREEN_MIN,
   MOOD_ORANGE_MAX,
@@ -47,6 +48,17 @@ import { returnTimeoutFallback } from "@/lib/insights/timeout-fallback";
 import { annotate } from "@/lib/logging/context";
 import { toBerlinDayKey } from "@/lib/tz/resolver";
 
+/**
+ * Drop the ranking-only `pooledSd` before a tag-influence row enters the
+ * snapshot payload — it exists solely to standardize the better-days sort
+ * and has no place in the model prompt or the UI.
+ */
+function stripRankingOnly(row: TagInfluenceRow): Omit<TagInfluenceRow, "pooledSd"> {
+  const { pooledSd: _pooledSd, ...rest } = row;
+  void _pooledSd;
+  return rest;
+}
+
 export async function generateMoodStatusForUser(
   userId: string,
   options?: {
@@ -352,10 +364,13 @@ export async function generateMoodStatusForUser(
       tagInfluence:
         tagInfluence.flat.length > 0 || tagInfluence.structured.length > 0
           ? {
-              flat: tagInfluence.flat.length > 0 ? tagInfluence.flat : null,
+              flat:
+                tagInfluence.flat.length > 0
+                  ? tagInfluence.flat.map(stripRankingOnly)
+                  : null,
               structured:
                 tagInfluence.structured.length > 0
-                  ? tagInfluence.structured
+                  ? tagInfluence.structured.map(stripRankingOnly)
                   : null,
             }
           : null,