From 03925ddece00b220087adc56313d4f1b54833429 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Marc-Andr=C3=A9=20Bombeck?= <mbombeck@gmail.com>
Date: Tue, 2 Jun 2026 16:43:41 +0200
Subject: [PATCH] feat(insights): time-range pills with period-over-period
 deltas on metric pages
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Add a 7d / 30d / 90d / 1y segmented control to every HealthKit metric page,
between the stat strip and the chart, with a sentiment-coloured "vs prior
period" delta caption beside it.

New route GET /api/analytics/range?type=<MeasurementType>&range=<7d|30d|90d|1y>
returns the current-window aggregate, the previous comparable window, and the
composed delta. Single metric type by construction, so the read is one
rollup-tier call covering the trailing 2N days sliced into the two halves —
no per-type fan-out, no Prisma-pool burst. Reuses the WMY readers
(readBestGranularityRollups + aggregateWmyBuckets) and the two-window pattern
the year-ago baseline already uses; count/min/max/mean/sum compose across
buckets, SD/slope/r2 stay out (not composable). The cached summaries slice is
untouched so the hottest dashboard cache key never fragments.

The delta caption reuses the dashboard's metric-aware up-good/up-bad colour
logic — extracted from trend-card into a shared trend-sentiment module — with
the metric's good direction sourced from the metric-status registry
(target-band metrics render neutral, matching deviation framing). When the
prior window has no data the caption shows "no prior-period data" rather than
a misleading 0%. The range choice persists across metrics via a localStorage
seam and defaults to 30d so nothing regresses. Cache keyed per (type, range)
through the query-key factory.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
(cherry picked from commit eac87a9f8b144f4d8e7a8b950528a19d5dcf8726)
---
 docs/api/openapi.yaml                         | 182 ++++++++++++++++++
 messages/de.json                              |  16 ++
 messages/en.json                              |  16 ++
 messages/es.json                              |  16 ++
 messages/fr.json                              |  16 ++
 messages/it.json                              |  16 ++
 messages/pl.json                              |  16 ++
 src/app/api/analytics/range/route.ts          |  72 +++++++
 src/components/charts/trend-card.tsx          |  75 ++------
 .../healthkit-metric-page-status.test.tsx     |   7 +
 .../insights/healthkit-metric-page.tsx        |  50 ++++-
 .../insights/metric-range-delta.tsx           |  86 +++++++++
 src/components/insights/time-range-pills.tsx  |  64 ++++++
 src/hooks/use-analytics-range.ts              |  56 ++++++
 src/hooks/use-insights-layout-prefs.ts        |  55 ++++++
 .../analytics/__tests__/range-delta.test.ts   | 120 ++++++++++++
 src/lib/analytics/range-delta.ts              | 180 +++++++++++++++++
 src/lib/insights/trend-sentiment.ts           |  66 +++++++
 src/lib/openapi/registry.ts                   |   1 +
 src/lib/openapi/routes.ts                     |  96 +++++++++
 src/lib/query-keys.ts                         |  10 +
 21 files changed, 1151 insertions(+), 65 deletions(-)
 create mode 100644 src/app/api/analytics/range/route.ts
 create mode 100644 src/components/insights/metric-range-delta.tsx
 create mode 100644 src/components/insights/time-range-pills.tsx
 create mode 100644 src/hooks/use-analytics-range.ts
 create mode 100644 src/lib/analytics/__tests__/range-delta.test.ts
 create mode 100644 src/lib/analytics/range-delta.ts
 create mode 100644 src/lib/insights/trend-sentiment.ts

diff --git a/docs/api/openapi.yaml b/docs/api/openapi.yaml
index 2e4cf533e..6395d7cab 100644
--- a/docs/api/openapi.yaml
+++ b/docs/api/openapi.yaml
@@ -30,6 +30,7 @@ tags:
   - name: Measurements
   - name: Medications
   - name: Mood
+  - name: Analytics
   - name: Insights
   - name: Dashboard
   - name: Notifications
@@ -1954,6 +1955,86 @@ paths:
         "401": *a1
         "422": *a2
         "429": *a3
+  /api/analytics/range:
+    get:
+      tags:
+        - Analytics
+      summary: Single-metric period-over-period range delta
+      description: v1.9.0 — returns the current-window aggregate, the previous comparable window, and the composed delta for
+        ONE metric type over a `7d` / `30d` / `90d` / `1y` range. Single-type by construction (the metric page is
+        single-metric), so the read is one rollup-tier call covering the trailing 2N days sliced into the two halves —
+        no per-type fan-out. Additive route; the `/api/analytics` envelope is unchanged. Auth via cookie or Bearer.
+      parameters:
+        - in: query
+          name: type
+          schema:
+            type: string
+            enum:
+              - WEIGHT
+              - BLOOD_PRESSURE_SYS
+              - BLOOD_PRESSURE_DIA
+              - PULSE
+              - BODY_FAT
+              - SLEEP_DURATION
+              - ACTIVITY_STEPS
+              - BLOOD_GLUCOSE
+              - TOTAL_BODY_WATER
+              - BONE_MASS
+              - OXYGEN_SATURATION
+              - HEART_RATE_VARIABILITY
+              - RESTING_HEART_RATE
+              - ACTIVE_ENERGY_BURNED
+              - FLIGHTS_CLIMBED
+              - WALKING_RUNNING_DISTANCE
+              - VO2_MAX
+              - BODY_TEMPERATURE
+              - FAT_FREE_MASS
+              - FAT_MASS
+              - MUSCLE_MASS
+              - SKIN_TEMPERATURE
+              - PULSE_WAVE_VELOCITY
+              - VASCULAR_AGE
+              - VISCERAL_FAT
+              - AUDIO_EXPOSURE_ENV
+              - AUDIO_EXPOSURE_HEADPHONE
+              - TIME_IN_DAYLIGHT
+              - WALKING_STEADINESS
+              - AUDIO_EXPOSURE_EVENT
+              - RESPIRATORY_RATE
+              - BODY_MASS_INDEX
+              - LEAN_BODY_MASS
+              - WALKING_HEART_RATE_AVERAGE
+              - WALKING_ASYMMETRY
+              - WALKING_DOUBLE_SUPPORT
+              - WALKING_STEP_LENGTH
+              - WALKING_SPEED
+            description: "The measurement type to read (single metric — no fan-out). Closed enum: an unknown type 422s."
+          required: true
+          description: "The measurement type to read (single metric — no fan-out). Closed enum: an unknown type 422s."
+        - in: query
+          name: range
+          schema:
+            type: string
+            enum:
+              - 7d
+              - 30d
+              - 90d
+              - 1y
+            description: "Trailing window: `7d` / `30d` / `90d` / `1y`. The previous comparable window is the equally-sized span
+              immediately before it."
+          required: true
+          description: "Trailing window: `7d` / `30d` / `90d` / `1y`. The previous comparable window is the equally-sized span
+            immediately before it."
+      responses:
+        "200":
+          description: Current + previous window aggregates and the delta.
+          content:
+            application/json:
+              schema:
+                $ref: "#/components/schemas/AnalyticsRangeResponseEnvelope"
+        "401": *a1
+        "422": *a2
+        "429": *a3
   /api/insights/metric-status:
     get:
       tags:
@@ -6334,6 +6415,107 @@ components:
       additionalProperties: false
       description: Acknowledgement that a full assessment warm was enqueued for the calling user. The generation runs out of
         band; the text lands in the read-only status routes.
+    AnalyticsRangeResponseEnvelope:
+      type: object
+      properties:
+        data:
+          $ref: "#/components/schemas/AnalyticsRangeResponse"
+        error:
+          type: "null"
+        meta:
+          type: object
+          properties:
+            requestId:
+              type: string
+          additionalProperties: false
+      required:
+        - data
+        - error
+      additionalProperties: false
+    AnalyticsRangeResponse:
+      type: object
+      properties:
+        range:
+          type: string
+          enum:
+            - 7d
+            - 30d
+            - 90d
+            - 1y
+          description: The range that was read (echoes the request).
+        windowDays:
+          type: integer
+          minimum: -9007199254740991
+          maximum: 9007199254740991
+          description: Trailing-window length in days for the chosen range.
+        granularity:
+          type: string
+          description: Rollup granularity the read resolved against (`DAY` / `WEEK` / `MONTH` / `YEAR`, or `none` on a coverage
+            miss).
+        current:
+          description: Aggregate over the current window `[now-N, now)`.
+          $ref: "#/components/schemas/AnalyticsWindowAggregate"
+        previous:
+          description: Aggregate over the previous comparable window `[now-2N, now-N)`.
+          $ref: "#/components/schemas/AnalyticsWindowAggregate"
+        delta:
+          anyOf:
+            - type: number
+            - type: "null"
+          description: "`current.mean - previous.mean`; null when either window has no data (never a misleading 0)."
+        deltaPct:
+          anyOf:
+            - type: number
+            - type: "null"
+          description: "`delta / previous.mean` as a fraction (0.03 = +3 %); null when the prior window has no / zero mean (no
+            divide-by-zero). The client shows 'no prior-period data' in that case."
+      required:
+        - range
+        - windowDays
+        - granularity
+        - current
+        - previous
+        - delta
+        - deltaPct
+      additionalProperties: false
+      description: Single-metric period-over-period aggregate. Reads the current and previous comparable windows from the WMY
+        rollup tier and composes a count-weighted-mean delta. `count/min/max/mean/sum` are linearly composable across
+        buckets; SD/slope/r² are intentionally excluded (not composable).
+    AnalyticsWindowAggregate:
+      type: object
+      properties:
+        count:
+          type: integer
+          minimum: -9007199254740991
+          maximum: 9007199254740991
+          description: Reading count composed across buckets.
+        min:
+          anyOf:
+            - type: number
+            - type: "null"
+          description: Window minimum; null when empty.
+        max:
+          anyOf:
+            - type: number
+            - type: "null"
+          description: Window maximum; null when empty.
+        mean:
+          anyOf:
+            - type: number
+            - type: "null"
+          description: Count-weighted mean across buckets; null when empty.
+        sum:
+          anyOf:
+            - type: number
+            - type: "null"
+          description: Cumulative total for cumulative metrics (steps, energy, distance); null when no bucket carries a sum.
+      required:
+        - count
+        - min
+        - max
+        - mean
+        - sum
+      additionalProperties: false
     MetricStatusResponseEnvelope:
       type: object
       properties:
diff --git a/messages/de.json b/messages/de.json
index b8f1a6c9f..0b4fce63a 100644
--- a/messages/de.json
+++ b/messages/de.json
@@ -1601,6 +1601,22 @@
       "confidenceLow": "Geringe Konfidenz",
       "pendingLabel": "Trend-Kommentar wird generiert"
     },
+    "range": {
+      "label": "Zeitraum",
+      "noPriorData": "keine Daten im Vorzeitraum",
+      "option": {
+        "7d": "7T",
+        "30d": "30T",
+        "90d": "90T",
+        "1y": "1J"
+      },
+      "priorLabel": {
+        "7d": "ggü. vorherigen 7T",
+        "30d": "ggü. vorherigen 30T",
+        "90d": "ggü. vorherigen 90T",
+        "1y": "ggü. Vorjahr"
+      }
+    },
     "correlationRow": {
       "title": "Zusammenhänge",
       "subtitle": "Vorab definierte Hypothesen, gegated auf n ≥ 14 und p < 0,05",
diff --git a/messages/en.json b/messages/en.json
index ca23e5853..2d8dd0c2f 100644
--- a/messages/en.json
+++ b/messages/en.json
@@ -1601,6 +1601,22 @@
       "confidenceLow": "Low confidence",
       "pendingLabel": "Generating trend commentary"
     },
+    "range": {
+      "label": "Time range",
+      "noPriorData": "no prior-period data",
+      "option": {
+        "7d": "7d",
+        "30d": "30d",
+        "90d": "90d",
+        "1y": "1y"
+      },
+      "priorLabel": {
+        "7d": "vs prior 7d",
+        "30d": "vs prior 30d",
+        "90d": "vs prior 90d",
+        "1y": "vs prior year"
+      }
+    },
     "correlationRow": {
       "title": "Relationships",
       "subtitle": "Pre-defined hypotheses, gated on n ≥ 14 and p < 0.05",
diff --git a/messages/es.json b/messages/es.json
index b34fa799b..81e7c7f83 100644
--- a/messages/es.json
+++ b/messages/es.json
@@ -1601,6 +1601,22 @@
       "confidenceLow": "Confianza baja",
       "pendingLabel": "Generando comentario de tendencia"
     },
+    "range": {
+      "label": "Período",
+      "noPriorData": "sin datos del período anterior",
+      "option": {
+        "7d": "7d",
+        "30d": "30d",
+        "90d": "90d",
+        "1y": "1a"
+      },
+      "priorLabel": {
+        "7d": "vs 7d anteriores",
+        "30d": "vs 30d anteriores",
+        "90d": "vs 90d anteriores",
+        "1y": "vs año anterior"
+      }
+    },
     "correlationRow": {
       "title": "Correlaciones",
       "subtitle": "Hipótesis predefinidas, con umbral n ≥ 14 y p < 0,05",
diff --git a/messages/fr.json b/messages/fr.json
index cf4ceb691..69ba52e45 100644
--- a/messages/fr.json
+++ b/messages/fr.json
@@ -1601,6 +1601,22 @@
       "confidenceLow": "Confiance faible",
       "pendingLabel": "Commentaire de tendance en cours de génération"
     },
+    "range": {
+      "label": "Période",
+      "noPriorData": "aucune donnée sur la période précédente",
+      "option": {
+        "7d": "7j",
+        "30d": "30j",
+        "90d": "90j",
+        "1y": "1an"
+      },
+      "priorLabel": {
+        "7d": "vs 7j précédents",
+        "30d": "vs 30j précédents",
+        "90d": "vs 90j précédents",
+        "1y": "vs année précédente"
+      }
+    },
     "correlationRow": {
       "title": "Corrélations",
       "subtitle": "Hypothèses prédéfinies, validées si n ≥ 14 et p < 0,05",
diff --git a/messages/it.json b/messages/it.json
index 175434875..360007928 100644
--- a/messages/it.json
+++ b/messages/it.json
@@ -1601,6 +1601,22 @@
       "confidenceLow": "Confidenza bassa",
       "pendingLabel": "Commento di trend in generazione"
     },
+    "range": {
+      "label": "Periodo",
+      "noPriorData": "nessun dato nel periodo precedente",
+      "option": {
+        "7d": "7g",
+        "30d": "30g",
+        "90d": "90g",
+        "1y": "1a"
+      },
+      "priorLabel": {
+        "7d": "vs 7g precedenti",
+        "30d": "vs 30g precedenti",
+        "90d": "vs 90g precedenti",
+        "1y": "vs anno precedente"
+      }
+    },
     "correlationRow": {
       "title": "Correlazioni",
       "subtitle": "Ipotesi predefinite, vincolate a n ≥ 14 e p < 0,05",
diff --git a/messages/pl.json b/messages/pl.json
index 22b4b7bc5..2e37e134e 100644
--- a/messages/pl.json
+++ b/messages/pl.json
@@ -1601,6 +1601,22 @@
       "confidenceLow": "Niska pewność",
       "pendingLabel": "Komentarz trendu jest generowany"
     },
+    "range": {
+      "label": "Okres",
+      "noPriorData": "brak danych z poprzedniego okresu",
+      "option": {
+        "7d": "7d",
+        "30d": "30d",
+        "90d": "90d",
+        "1y": "1r"
+      },
+      "priorLabel": {
+        "7d": "vs poprzednie 7d",
+        "30d": "vs poprzednie 30d",
+        "90d": "vs poprzednie 90d",
+        "1y": "vs poprzedni rok"
+      }
+    },
     "correlationRow": {
       "title": "Korelacje",
       "subtitle": "Hipotezy zdefiniowane wcześniej, z progiem n ≥ 14 i p < 0,05",
diff --git a/src/app/api/analytics/range/route.ts b/src/app/api/analytics/range/route.ts
new file mode 100644
index 000000000..bebb0a292
--- /dev/null
+++ b/src/app/api/analytics/range/route.ts
@@ -0,0 +1,72 @@
+/**
+ * v1.9.0 — single-metric period-over-period range read.
+ *
+ * `GET /api/analytics/range?type=<MeasurementType>&range=7d|30d|90d|1y`
+ * returns the current-window aggregate, the previous comparable window, and
+ * the composed delta. Single metric type by construction — the Insights
+ * metric page is single-metric — so the read is one
+ * `readBestGranularityRollups` covering the trailing `2N` days, sliced into
+ * the two halves. No per-type fan-out, no Prisma-pool burst (the opposite of
+ * the deleted 15-way live walk).
+ *
+ * The cached `/api/analytics?slice=summaries` slice is deliberately left
+ * untouched: adding a range dimension to its single shared cache key would
+ * fragment the hottest cache on the dashboard. This dedicated route carries
+ * its own per-(type, range) cache slot on the client (`queryKeys.analyticsRange`).
+ *
+ * iOS-safe: additive new route, no change to the `/api/analytics` envelope.
+ */
+import { NextRequest } from "next/server";
+import { z } from "zod/v4";
+
+import { apiHandler, requireAuth } from "@/lib/api-handler";
+import { apiSuccess, returnAllZodIssues } from "@/lib/api-response";
+import { annotate } from "@/lib/logging/context";
+import { measurementTypeEnum } from "@/lib/validations/measurement";
+import {
+  ANALYTICS_RANGES,
+  computeRangeDelta,
+  type AnalyticsRange,
+} from "@/lib/analytics/range-delta";
+import type { MeasurementType } from "@/generated/prisma/client";
+
+export const dynamic = "force-dynamic";
+
+const rangeQuerySchema = z.object({
+  type: measurementTypeEnum,
+  range: z.enum(ANALYTICS_RANGES),
+});
+
+export const GET = apiHandler(async (request: NextRequest) => {
+  const { user } = await requireAuth();
+
+  const parsed = rangeQuerySchema.safeParse({
+    type: request.nextUrl.searchParams.get("type"),
+    range: request.nextUrl.searchParams.get("range"),
+  });
+  if (!parsed.success) {
+    annotate({
+      action: { name: "analytics.range.invalid-params" },
+      meta: { issue_count: parsed.error.issues.length },
+    });
+    return returnAllZodIssues(parsed.error, 422);
+  }
+
+  const type = parsed.data.type as MeasurementType;
+  const range = parsed.data.range as AnalyticsRange;
+
+  const result = await computeRangeDelta(user.id, type, range);
+
+  annotate({
+    action: { name: "analytics.range.read" },
+    meta: {
+      type,
+      range,
+      granularity: result.granularity,
+      current_count: result.current.count,
+      previous_count: result.previous.count,
+    },
+  });
+
+  return apiSuccess(result);
+});
diff --git a/src/components/charts/trend-card.tsx b/src/components/charts/trend-card.tsx
index cfc73b02a..8addfbfb6 100644
--- a/src/components/charts/trend-card.tsx
+++ b/src/components/charts/trend-card.tsx
@@ -11,6 +11,17 @@ import {
 } from "@/components/ui/tooltip";
 import { useTranslations, useFormatters } from "@/lib/i18n/context";
 import type { ComparisonBaseline } from "@/lib/dashboard-layout";
+import {
+  getTrendSentiment,
+  sentimentColorClass,
+  type TrendDirectionSentiment,
+  type TrendSentimentDirection,
+} from "@/lib/insights/trend-sentiment";
+
+// v1.9.0 — the metric-aware sentiment mapping moved to a shared module so the
+// range-delta caption paints the same colour for the same signal. Re-exported
+// here so existing `@/components/charts/trend-card` importers keep the type.
+export type { TrendDirectionSentiment, TrendSentimentDirection };
 
 interface SecondaryMetric {
   /** Sub-value latest reading (e.g. diastolic when latest is systolic). */
@@ -19,70 +30,6 @@ interface SecondaryMetric {
   avg30: number | null;
 }
 
-/**
- * Maps a metric's "up means" direction to colour sentiment for the small
- * trend arrow on each tile. v1.4.6 P4 stripped the colour entirely after
- * the original up=red / down=green mapping was wrong for half the metrics
- * (mood up = good, BP up = bad, pulse up = neutral). v1.5 phase-5 restores
- * the colour but per-metric:
- *
- *   - `up-good`   — higher value is better (mood, sleep hours, steps).
- *                   ↑ green, ↓ orange.
- *   - `up-bad`    — higher value is worse (BP, weight, body fat).
- *                   ↑ orange, ↓ green.
- *   - `neutral`   — direction doesn't carry a value judgement (pulse,
- *                   "BP in target %" — those have their own range
- *                   colouring on the avg7/avg30 numbers already).
- *
- * Strictly affects the ↑/↓/→ arrow next to the latest reading. Chart
- * lines, axes, and the avg7/avg30 colour classes are untouched.
- */
-export type TrendDirectionSentiment = "up-good" | "up-bad" | "neutral";
-
-/**
- * v1.4.33 F5 — single sentiment helper so the headline arrow, the 7-day
- * delta value, and the comparison-overlay caption all paint the same
- * colour for the same signal. Before this helper each consumer ran its
- * own branch — the arrow read `slope30.direction`, the delta value read
- * `trend7Delta > 0`, and the comparison caption read `compareDelta > 0`.
- * When the 30-day regression and the 7-day delta disagreed (weight down
- * over a month, up over a week) the tile painted a green arrow next to
- * an orange value, which read as "two metrics in one tile".
- *
- * `change` is a signed magnitude of the movement under inspection (the
- * 7-day delta in units, or the slope-projected-over-7-days for tiles
- * with only `slope30`). `sentiment` is the metric's improvement
- * direction.
- *
- *  - `'positive'` — the change moves the metric toward its goal
- *    (e.g. weight on `up-bad` going down, mood on `up-good` going up).
- *    Renders green.
- *  - `'negative'` — the change moves away from the goal. Renders orange.
- *  - `'neutral'` — either the sentiment is neutral (pulse, BP-in-target),
- *    or the change is below the noise floor (|change| < 0.05). Renders
- *    muted.
- */
-export type TrendSentimentDirection = "positive" | "negative" | "neutral";
-
-function getTrendSentiment(
-  change: number | null | undefined,
-  sentiment: TrendDirectionSentiment,
-): TrendSentimentDirection {
-  if (change == null || Math.abs(change) < 0.05) return "neutral";
-  if (sentiment === "neutral") return "neutral";
-  const isUp = change > 0;
-  const isGood =
-    (sentiment === "up-good" && isUp) ||
-    (sentiment === "up-bad" && !isUp);
-  return isGood ? "positive" : "negative";
-}
-
-function sentimentColorClass(direction: TrendSentimentDirection): string {
-  if (direction === "positive") return "text-dracula-green";
-  if (direction === "negative") return "text-dracula-orange";
-  return "text-muted-foreground";
-}
-
 interface TrendCardProps {
   label: string;
   latest: number | null;
diff --git a/src/components/insights/__tests__/healthkit-metric-page-status.test.tsx b/src/components/insights/__tests__/healthkit-metric-page-status.test.tsx
index d03212622..5a398339f 100644
--- a/src/components/insights/__tests__/healthkit-metric-page-status.test.tsx
+++ b/src/components/insights/__tests__/healthkit-metric-page-status.test.tsx
@@ -33,6 +33,13 @@ vi.mock("@/hooks/use-auth", () => ({
 
 vi.mock("@/hooks/use-insights-layout-prefs", () => ({
   useInsightsLayoutPrefs: () => ({ layout: null, compareBaseline: false }),
+  useInsightsRangePref: () => ({ range: "30d", setRange: () => {} }),
+}));
+
+// v1.9.0 — the page reads the period-over-period range delta; mock the hook
+// so the test stays deterministic and never touches `fetch`.
+vi.mock("@/hooks/use-analytics-range", () => ({
+  useAnalyticsRange: () => ({ data: undefined, isLoading: false }),
 }));
 
 const analyticsMock = vi.fn();
diff --git a/src/components/insights/healthkit-metric-page.tsx b/src/components/insights/healthkit-metric-page.tsx
index dcbe22f2f..2fa7b6ff8 100644
--- a/src/components/insights/healthkit-metric-page.tsx
+++ b/src/components/insights/healthkit-metric-page.tsx
@@ -7,11 +7,21 @@ import { Sparkles } from "lucide-react";
 
 import { useAuth } from "@/hooks/use-auth";
 import { useInsightsAnalytics } from "@/hooks/use-insights-analytics";
-import { useInsightsLayoutPrefs } from "@/hooks/use-insights-layout-prefs";
+import {
+  useInsightsLayoutPrefs,
+  useInsightsRangePref,
+} from "@/hooks/use-insights-layout-prefs";
+import { useAnalyticsRange } from "@/hooks/use-analytics-range";
 import { useTranslations } from "@/lib/i18n/context";
 import type { InsightMetric } from "@/lib/insights/metric-availability";
+import {
+  getMetricStatusMeta,
+  metricIdForMeasurementType,
+} from "@/lib/insights/metric-status-registry";
 import type { MetricStatusMetricId } from "@/lib/insights/metric-status-registry";
+import { sentimentFromMetricDirection } from "@/lib/insights/trend-sentiment";
 import type { ChartOverlayKey } from "@/lib/dashboard-layout";
+import type { MeasurementType } from "@/generated/prisma/client";
 import { Button } from "@/components/ui/button";
 import { HealthChartDynamic } from "@/components/charts/health-chart-dynamic";
 import { MetricStatusCard } from "@/components/insights/metric-status-card";
@@ -19,6 +29,8 @@ import { MetricEmptyState } from "@/components/insights/metric-empty-state";
 import { MetricStatStrip } from "@/components/insights/metric-stat-strip";
 import { MeasurementDiversityNudge } from "@/components/insights/measurement-diversity-nudge";
 import { MetricTargetSummary } from "@/components/insights/metric-target-summary";
+import { TimeRangePills } from "@/components/insights/time-range-pills";
+import { MetricRangeDelta } from "@/components/insights/metric-range-delta";
 import { SubPageShell } from "@/components/insights/sub-page-shell";
 
 /**
@@ -139,9 +151,30 @@ export function HealthKitMetricPage({
   const { user, isAuthenticated } = useAuth();
   const { t } = useTranslations();
   const { compareBaseline } = useInsightsLayoutPrefs(isAuthenticated);
+  const { range, setRange } = useInsightsRangePref();
 
   const { data: analytics, isEmpty } = useInsightsAnalytics(insightMetric);
 
+  // v1.9.0 — period-over-period range read. Single-metric, additive route;
+  // gated off the empty-data branch so a brand-new metric never fires it.
+  const { data: rangeData, isLoading: isRangeLoading } = useAnalyticsRange(
+    measurementType,
+    range,
+    !isEmpty,
+  );
+
+  // The delta's sentiment colour follows the metric's "good direction" from
+  // the metric-status registry (higher-better → up-good, lower-better →
+  // up-bad, target-band → neutral). Every HealthKit metric page is backed by
+  // a registry entry; a metric absent from the registry falls back to neutral.
+  const registryMeta = (() => {
+    const id = metricIdForMeasurementType(measurementType as MeasurementType);
+    return id ? getMetricStatusMeta(id) : null;
+  })();
+  const directionSentiment = registryMeta
+    ? sentimentFromMetricDirection(registryMeta.direction)
+    : "neutral";
+
   const rawSummary = analytics?.summaries?.[measurementType] ?? null;
   // The stat strip renders display-unit values. The summary holds stored
   // values, so when the page renders a scaled unit (e.g. WALKING_SPEED
@@ -207,6 +240,21 @@ export function HealthKitMetricPage({
       coachLaunch
       showAllValuesType={measurementType}
     >
+      {/* v1.9.0 — time-range pills + period-over-period delta, between the
+          stat strip and the chart. The selected range persists across metrics
+          via `useInsightsRangePref`. */}
+      <div
+        data-slot="metric-range-controls"
+        className="flex flex-wrap items-center justify-between gap-2"
+      >
+        <TimeRangePills value={range} onChange={setRange} />
+        <MetricRangeDelta
+          data={rangeData}
+          range={range}
+          directionSentiment={directionSentiment}
+          isLoading={isRangeLoading}
+        />
+      </div>
       <HealthChartDynamic
         chartKey={chartKey}
         types={[measurementType]}
diff --git a/src/components/insights/metric-range-delta.tsx b/src/components/insights/metric-range-delta.tsx
new file mode 100644
index 000000000..d8ee4e72a
--- /dev/null
+++ b/src/components/insights/metric-range-delta.tsx
@@ -0,0 +1,86 @@
+"use client";
+
+import { ArrowDown, ArrowUp } from "lucide-react";
+
+import { useTranslations } from "@/lib/i18n/context";
+import { cn } from "@/lib/utils";
+import {
+  getTrendSentiment,
+  sentimentColorClass,
+  type TrendDirectionSentiment,
+} from "@/lib/insights/trend-sentiment";
+import type { AnalyticsRangeData } from "@/hooks/use-analytics-range";
+import type { AnalyticsRange } from "@/lib/analytics/range-delta";
+
+/**
+ * v1.9.0 — period-over-period delta caption beside the time-range pills.
+ *
+ * Renders "+3 % vs prior 30d" with the metric-aware sentiment colour the
+ * dashboard tiles use (up-good vs up-bad). When the prior window has no data
+ * it shows a neutral "no prior-period data" line instead of a misleading 0 %
+ * (the `deltaPct` is null in that case). For `target-band` metrics the
+ * sentiment is neutral, so the delta still shows magnitude + direction but
+ * carries no good/bad colour — matching Oura's deviation framing for metrics
+ * with no single good direction.
+ */
+export function MetricRangeDelta({
+  data,
+  range,
+  directionSentiment,
+  isLoading,
+}: {
+  data: AnalyticsRangeData | undefined;
+  range: AnalyticsRange;
+  directionSentiment: TrendDirectionSentiment;
+  isLoading: boolean;
+}) {
+  const { t } = useTranslations();
+
+  if (isLoading) {
+    return (
+      <span
+        data-slot="metric-range-delta-pending"
+        className="bg-muted/60 inline-block h-3.5 w-28 animate-pulse rounded motion-reduce:animate-none"
+        aria-hidden="true"
+      />
+    );
+  }
+
+  const priorLabel = t(`insights.range.priorLabel.${range}`);
+
+  // No comparison available — prior window empty, or current window empty.
+  if (!data || data.delta === null || data.deltaPct === null) {
+    return (
+      <span
+        data-slot="metric-range-delta-empty"
+        className="text-muted-foreground text-xs"
+      >
+        {t("insights.range.noPriorData")}
+      </span>
+    );
+  }
+
+  const direction = getTrendSentiment(data.delta, directionSentiment);
+  const colorClass = sentimentColorClass(direction);
+  const isUp = data.delta > 0;
+  const pct = Math.round(Math.abs(data.deltaPct) * 100);
+  const sign = isUp ? "+" : "−";
+
+  return (
+    <span
+      data-slot="metric-range-delta"
+      data-direction={direction}
+      className={cn("inline-flex items-center gap-1 text-xs", colorClass)}
+    >
+      {isUp ? (
+        <ArrowUp className="h-3 w-3" aria-hidden="true" />
+      ) : (
+        <ArrowDown className="h-3 w-3" aria-hidden="true" />
+      )}
+      <span>
+        {sign}
+        {pct}% {priorLabel}
+      </span>
+    </span>
+  );
+}
diff --git a/src/components/insights/time-range-pills.tsx b/src/components/insights/time-range-pills.tsx
new file mode 100644
index 000000000..f02e69d92
--- /dev/null
+++ b/src/components/insights/time-range-pills.tsx
@@ -0,0 +1,64 @@
+"use client";
+
+import { useTranslations } from "@/lib/i18n/context";
+import { cn } from "@/lib/utils";
+import {
+  ANALYTICS_RANGES,
+  type AnalyticsRange,
+} from "@/lib/analytics/range-delta";
+
+/**
+ * v1.9.0 — `7d / 30d / 90d / 1y` segmented control for the Insights metric
+ * pages. The interaction model Apple Health / Oura users already know (a
+ * range selector across the top of a metric chart). Built as an accessible
+ * `radiogroup` of pills over the existing zinc/shadcn primitives rather than
+ * pulling a new Radix dependency — the project's tab strip uses the same
+ * custom-pill approach. The selected range drives the period-over-period
+ * delta caption and (where wired) the chart window.
+ */
+export function TimeRangePills({
+  value,
+  onChange,
+  className,
+}: {
+  value: AnalyticsRange;
+  onChange: (range: AnalyticsRange) => void;
+  className?: string;
+}) {
+  const { t } = useTranslations();
+
+  return (
+    <div
+      role="radiogroup"
+      aria-label={t("insights.range.label")}
+      data-slot="time-range-pills"
+      className={cn(
+        "bg-muted/50 inline-flex items-center gap-0.5 rounded-lg p-0.5",
+        className,
+      )}
+    >
+      {ANALYTICS_RANGES.map((range) => {
+        const selected = range === value;
+        return (
+          <button
+            key={range}
+            type="button"
+            role="radio"
+            aria-checked={selected}
+            data-range={range}
+            data-selected={selected}
+            onClick={() => onChange(range)}
+            className={cn(
+              "focus-visible:ring-ring rounded-md px-2.5 py-1 text-xs font-medium transition-colors focus-visible:ring-2 focus-visible:outline-none",
+              selected
+                ? "bg-background text-foreground shadow-sm"
+                : "text-muted-foreground hover:text-foreground",
+            )}
+          >
+            {t(`insights.range.option.${range}`)}
+          </button>
+        );
+      })}
+    </div>
+  );
+}
diff --git a/src/hooks/use-analytics-range.ts b/src/hooks/use-analytics-range.ts
new file mode 100644
index 000000000..dfb38bc3d
--- /dev/null
+++ b/src/hooks/use-analytics-range.ts
@@ -0,0 +1,56 @@
+"use client";
+
+import { useQuery } from "@tanstack/react-query";
+
+import { useAuth } from "@/hooks/use-auth";
+import { queryKeys } from "@/lib/query-keys";
+import type { AnalyticsRange } from "@/lib/analytics/range-delta";
+
+/**
+ * v1.9.0 — client reader for the single-metric period-over-period range
+ * read (`GET /api/analytics/range`). One cache slot per `(type, range)` via
+ * the centralised query-key factory, so switching the time-range pills is a
+ * cheap cache hit after first fetch and never collides with the shared
+ * `["analytics", "summaries"]` slot.
+ */
+export interface AnalyticsWindowAggregate {
+  count: number;
+  min: number | null;
+  max: number | null;
+  mean: number | null;
+  sum: number | null;
+}
+
+export interface AnalyticsRangeData {
+  range: AnalyticsRange;
+  windowDays: number;
+  granularity: string;
+  current: AnalyticsWindowAggregate;
+  previous: AnalyticsWindowAggregate;
+  delta: number | null;
+  deltaPct: number | null;
+}
+
+export function useAnalyticsRange(
+  type: string,
+  range: AnalyticsRange,
+  enabled = true,
+) {
+  const { isAuthenticated } = useAuth();
+
+  return useQuery({
+    queryKey: queryKeys.analyticsRange(type, range),
+    queryFn: async (): Promise<AnalyticsRangeData> => {
+      const res = await fetch(
+        `/api/analytics/range?type=${encodeURIComponent(
+          type,
+        )}&range=${encodeURIComponent(range)}`,
+      );
+      if (!res.ok) throw new Error("Failed to load range");
+      const json = (await res.json()) as { data: AnalyticsRangeData };
+      return json.data;
+    },
+    enabled: isAuthenticated && enabled && type.length > 0,
+    staleTime: 60 * 1000,
+  });
+}
diff --git a/src/hooks/use-insights-layout-prefs.ts b/src/hooks/use-insights-layout-prefs.ts
index af37ce719..c863d24b1 100644
--- a/src/hooks/use-insights-layout-prefs.ts
+++ b/src/hooks/use-insights-layout-prefs.ts
@@ -1,5 +1,7 @@
 "use client";
 
+import { useCallback, useState } from "react";
+
 import { useQuery } from "@tanstack/react-query";
 
 import { queryKeys } from "@/lib/query-keys";
@@ -7,6 +9,59 @@ import {
   resolveDashboardLayout,
   type DashboardLayout,
 } from "@/lib/dashboard-layout";
+import {
+  ANALYTICS_RANGES,
+  type AnalyticsRange,
+} from "@/lib/analytics/range-delta";
+
+const RANGE_STORAGE_KEY = "healthlog.insights.range";
+const DEFAULT_RANGE: AnalyticsRange = "30d";
+
+function isAnalyticsRange(value: string | null): value is AnalyticsRange {
+  return (
+    value !== null && (ANALYTICS_RANGES as readonly string[]).includes(value)
+  );
+}
+
+/**
+ * Read the persisted range once, lazily. Guarded by `typeof window` so the
+ * server render (where `localStorage` is absent) returns the default rather
+ * than throwing; the first client render then hydrates with the stored value
+ * via the lazy `useState` initializer — no `setState`-in-effect needed.
+ */
+function readStoredRange(): AnalyticsRange {
+  if (typeof window === "undefined") return DEFAULT_RANGE;
+  try {
+    const stored = window.localStorage.getItem(RANGE_STORAGE_KEY);
+    return isAnalyticsRange(stored) ? stored : DEFAULT_RANGE;
+  } catch {
+    return DEFAULT_RANGE;
+  }
+}
+
+/**
+ * v1.9.0 — persisted time-range choice for the Insights metric pages.
+ *
+ * The choice sticks across metrics (a user who picks `90d` on weight sees
+ * `90d` on HRV) via `localStorage` — the same lightweight client-pref seam
+ * the comparison toggle conceptually lives alongside, without a server
+ * round-trip or a dashboard-layout schema change. Defaults to `30d` so the
+ * existing fixed-window behaviour is the default and nothing regresses.
+ */
+export function useInsightsRangePref() {
+  const [range, setRangeState] = useState<AnalyticsRange>(readStoredRange);
+
+  const setRange = useCallback((next: AnalyticsRange) => {
+    setRangeState(next);
+    try {
+      window.localStorage.setItem(RANGE_STORAGE_KEY, next);
+    } catch {
+      // Best-effort persistence; the in-memory state still drives the UI.
+    }
+  }, []);
+
+  return { range, setRange } as const;
+}
 
 /**
  * v1.4.25 W4 — shared reader for the dashboard-layout payload across
diff --git a/src/lib/analytics/__tests__/range-delta.test.ts b/src/lib/analytics/__tests__/range-delta.test.ts
new file mode 100644
index 000000000..9f4914f98
--- /dev/null
+++ b/src/lib/analytics/__tests__/range-delta.test.ts
@@ -0,0 +1,120 @@
+import { describe, it, expect } from "vitest";
+
+import {
+  composeDelta,
+  rangeWindowDays,
+  sliceWindowDelta,
+  type WindowAggregate,
+} from "@/lib/analytics/range-delta";
+import type { RollupBucketRow } from "@/lib/rollups/measurement-read-wmy";
+
+/**
+ * v1.9.0 — the period-over-period range delta is two reads composed into one
+ * caption. The composition (slice the buckets into the current vs previous
+ * halves, aggregate each, derive the delta) is the load-bearing logic and is
+ * pure over the bucket rows — pinned here without a DB.
+ */
+
+const DAY_MS = 24 * 60 * 60 * 1000;
+
+function bucket(daysAgo: number, count: number, mean: number): RollupBucketRow {
+  return {
+    bucketStart: new Date(Date.now() - daysAgo * DAY_MS),
+    count,
+    mean,
+    sd: null,
+    slope: null,
+    r2: null,
+    sumValue: count * mean,
+    minValue: mean,
+    maxValue: mean,
+  };
+}
+
+describe("rangeWindowDays", () => {
+  it("maps each range onto its trailing-window length", () => {
+    expect(rangeWindowDays("7d")).toBe(7);
+    expect(rangeWindowDays("30d")).toBe(30);
+    expect(rangeWindowDays("90d")).toBe(90);
+    expect(rangeWindowDays("1y")).toBe(365);
+  });
+});
+
+describe("sliceWindowDelta", () => {
+  it("splits buckets into the current and previous halves by bucketStart", () => {
+    const now = Date.now();
+    // 30d window: current = [now-30, now), previous = [now-60, now-30).
+    const rows = [
+      bucket(5, 2, 80), // current
+      bucket(20, 3, 82), // current
+      bucket(40, 2, 70), // previous
+      bucket(55, 2, 74), // previous
+    ];
+    const { current, previous } = sliceWindowDelta(rows, 30, now);
+    expect(current.count).toBe(5);
+    expect(previous.count).toBe(4);
+    // Count-weighted means compose linearly across buckets.
+    expect(current.mean).toBeCloseTo((2 * 80 + 3 * 82) / 5, 6);
+    expect(previous.mean).toBeCloseTo((2 * 70 + 2 * 74) / 4, 6);
+  });
+
+  it("excludes buckets outside the 2N span", () => {
+    const now = Date.now();
+    const rows = [
+      bucket(5, 1, 90), // current
+      bucket(40, 1, 80), // previous
+      bucket(200, 1, 50), // outside both halves
+    ];
+    const { current, previous } = sliceWindowDelta(rows, 30, now);
+    expect(current.count).toBe(1);
+    expect(previous.count).toBe(1);
+  });
+
+  it("yields empty aggregates when no bucket falls in a half", () => {
+    const now = Date.now();
+    const rows = [bucket(5, 4, 100)]; // only current half populated
+    const { current, previous } = sliceWindowDelta(rows, 30, now);
+    expect(current.count).toBe(4);
+    expect(previous.count).toBe(0);
+    expect(previous.mean).toBeNull();
+  });
+});
+
+describe("composeDelta", () => {
+  const win = (count: number, mean: number | null): WindowAggregate => ({
+    count,
+    min: mean,
+    max: mean,
+    mean,
+    sum: mean === null ? null : count * mean,
+  });
+
+  it("computes the signed delta and percentage", () => {
+    const { delta, deltaPct } = composeDelta(win(5, 103), win(5, 100));
+    expect(delta).toBeCloseTo(3, 6);
+    expect(deltaPct).toBeCloseTo(0.03, 6);
+  });
+
+  it("returns null delta when either window is empty", () => {
+    expect(composeDelta(win(0, null), win(5, 100))).toEqual({
+      delta: null,
+      deltaPct: null,
+    });
+    expect(composeDelta(win(5, 100), win(0, null))).toEqual({
+      delta: null,
+      deltaPct: null,
+    });
+  });
+
+  it("returns null deltaPct (not Infinity) when the prior mean is zero", () => {
+    const { delta, deltaPct } = composeDelta(win(5, 4), win(3, 0));
+    expect(delta).toBeCloseTo(4, 6);
+    expect(deltaPct).toBeNull();
+  });
+
+  it("captures a negative move", () => {
+    const { delta, deltaPct } = composeDelta(win(5, 90), win(5, 100));
+    expect(delta).toBeCloseTo(-10, 6);
+    expect(deltaPct).toBeCloseTo(-0.1, 6);
+  });
+});
diff --git a/src/lib/analytics/range-delta.ts b/src/lib/analytics/range-delta.ts
new file mode 100644
index 000000000..2132e4778
--- /dev/null
+++ b/src/lib/analytics/range-delta.ts
@@ -0,0 +1,180 @@
+/**
+ * v1.9.0 — period-over-period range deltas for a single metric.
+ *
+ * Background
+ * ----------
+ * The Insights metric pages render fixed windows (`avg7`, `avg30`,
+ * `slope7/30/90`). A user-selectable time-range with a "vs prior period"
+ * delta needs two reads of the same metric: the current window and the
+ * previous comparable window, composed into a delta. This is exactly the
+ * two-window pattern `computeAvg30LastYearForType` already uses for the
+ * year-ago baseline (`summaries-slice.ts`) — current = the trailing window,
+ * previous = the equally-sized window immediately before it.
+ *
+ * Single metric, two reads, no fan-out
+ * ------------------------------------
+ * The route this backs is single-metric (the metric page is single-metric),
+ * so the cost is one `readBestGranularityRollups` call covering the trailing
+ * `2N` days, sliced into the current and previous halves. No per-type fan-out,
+ * no Prisma-pool burst — the opposite of the deleted 15-way live walk.
+ *
+ * Compositional contract
+ * ----------------------
+ * `count / min / max / mean / sum` are linearly composable across buckets
+ * (the rollup tier's compositional contract — `measurement-read-wmy.ts`), so
+ * a window aggregate built from WEEK / MONTH buckets equals the per-row
+ * aggregate over the same span. SD / slope / r² are intentionally NOT part of
+ * the delta — they do not compose, matching the rest of the WMY reader tier.
+ */
+import type { MeasurementType } from "@/generated/prisma/client";
+
+import {
+  aggregateWmyBuckets,
+  readBestGranularityRollups,
+  type RollupBucketRow,
+} from "@/lib/rollups/measurement-read-wmy";
+
+const DAY_MS = 24 * 60 * 60 * 1000;
+
+/** The user-selectable ranges. */
+export const ANALYTICS_RANGES = ["7d", "30d", "90d", "1y"] as const;
+export type AnalyticsRange = (typeof ANALYTICS_RANGES)[number];
+
+/** Trailing-window length, in days, for each range. */
+const RANGE_WINDOW_DAYS: Record<AnalyticsRange, number> = {
+  "7d": 7,
+  "30d": 30,
+  "90d": 90,
+  "1y": 365,
+};
+
+export function rangeWindowDays(range: AnalyticsRange): number {
+  return RANGE_WINDOW_DAYS[range];
+}
+
+/** A composed window aggregate. `mean`/etc. are null when the window is empty. */
+export interface WindowAggregate {
+  count: number;
+  min: number | null;
+  max: number | null;
+  mean: number | null;
+  sum: number | null;
+}
+
+export interface RangeDeltaResult {
+  range: AnalyticsRange;
+  windowDays: number;
+  /** The granularity the underlying rollup read resolved against. */
+  granularity: string;
+  current: WindowAggregate;
+  previous: WindowAggregate;
+  /**
+   * `current.mean - previous.mean`, or null when either window has no mean
+   * (no data in the current or prior window) — never a misleading 0.
+   */
+  delta: number | null;
+  /**
+   * `delta / previous.mean` as a fraction (0.03 = +3 %), or null when the
+   * prior window has no mean / a zero mean (no divide-by-zero, no misleading
+   * 0 %). The client renders "no prior-period data" in that case.
+   */
+  deltaPct: number | null;
+}
+
+/**
+ * Slice the resolved buckets into the current half (`[now-N, now)`) and the
+ * previous half (`[now-2N, now-N)`) and compose each. Pure over the bucket
+ * rows + the window boundaries so the composition is unit-testable without a
+ * DB. `bucketStart` is the slice key — the same conservative overlap filter
+ * `computeAvg30LastYearForType` uses (a bucket counts toward the half its
+ * `bucketStart` falls into).
+ */
+export function sliceWindowDelta(
+  rows: RollupBucketRow[],
+  windowDays: number,
+  now: number,
+): { current: WindowAggregate; previous: WindowAggregate } {
+  const currentStart = now - windowDays * DAY_MS;
+  const previousStart = now - 2 * windowDays * DAY_MS;
+  const currentRows: RollupBucketRow[] = [];
+  const previousRows: RollupBucketRow[] = [];
+  for (const row of rows) {
+    const t = row.bucketStart.getTime();
+    if (t >= currentStart && t < now) {
+      currentRows.push(row);
+    } else if (t >= previousStart && t < currentStart) {
+      previousRows.push(row);
+    }
+  }
+  return {
+    current: aggregateWmyBuckets(currentRows),
+    previous: aggregateWmyBuckets(previousRows),
+  };
+}
+
+/**
+ * Compose the period-over-period delta from the two window aggregates.
+ * Pure — pinned by unit test. Guards both the missing-data case (either
+ * window empty → delta null) and the divide-by-zero case (prior mean zero or
+ * null → deltaPct null) so the caption never paints a misleading 0 %.
+ */
+export function composeDelta(
+  current: WindowAggregate,
+  previous: WindowAggregate,
+): { delta: number | null; deltaPct: number | null } {
+  if (current.mean === null || previous.mean === null) {
+    return { delta: null, deltaPct: null };
+  }
+  const delta = current.mean - previous.mean;
+  const deltaPct = previous.mean !== 0 ? delta / previous.mean : null;
+  return { delta, deltaPct };
+}
+
+/**
+ * Read the current vs previous window for one `(userId, type)` and compose
+ * the delta. Reads a single `2N`-day window through the granularity router so
+ * both halves resolve at the same granularity (the comparison stays
+ * apples-to-apples). Returns a zeroed result on a coverage miss — the route
+ * still answers 200 with empty windows so the UI shows "no prior-period
+ * data" rather than erroring.
+ */
+export async function computeRangeDelta(
+  userId: string,
+  type: MeasurementType,
+  range: AnalyticsRange,
+  now: number = Date.now(),
+): Promise<RangeDeltaResult> {
+  const windowDays = rangeWindowDays(range);
+  // Read the full 2N span in one go so the current and previous halves share
+  // a granularity; the router picks the coarsest tier that resolves 2N.
+  const resolved = await readBestGranularityRollups(userId, type, windowDays * 2);
+  if (!resolved) {
+    const empty: WindowAggregate = {
+      count: 0,
+      min: null,
+      max: null,
+      mean: null,
+      sum: null,
+    };
+    return {
+      range,
+      windowDays,
+      granularity: "none",
+      current: empty,
+      previous: empty,
+      delta: null,
+      deltaPct: null,
+    };
+  }
+  const { current, previous } = sliceWindowDelta(resolved.rows, windowDays, now);
+  const { delta, deltaPct } = composeDelta(current, previous);
+  return {
+    range,
+    windowDays,
+    granularity: resolved.granularity,
+    current,
+    previous,
+    delta,
+    deltaPct,
+  };
+}
diff --git a/src/lib/insights/trend-sentiment.ts b/src/lib/insights/trend-sentiment.ts
new file mode 100644
index 000000000..5933f0b14
--- /dev/null
+++ b/src/lib/insights/trend-sentiment.ts
@@ -0,0 +1,66 @@
+/**
+ * v1.9.0 — shared metric-aware trend sentiment helpers.
+ *
+ * Extracted from `trend-card.tsx` so the range-delta caption (the time-range
+ * pills feature) paints the exact same colour for the same signal as the
+ * dashboard tiles. The mapping is the v1.5 phase-5 audit's: an upward move is
+ * good for some metrics (mood, HRV, steps), bad for others (resting HR,
+ * weight, body fat), and value-judgement-free for the rest.
+ */
+
+/**
+ * Maps a metric's "up means" direction to colour sentiment.
+ *
+ *   - `up-good`   — higher value is better (mood, HRV, steps). ↑ green, ↓ orange.
+ *   - `up-bad`    — higher value is worse (resting HR, weight, body fat).
+ *                   ↑ orange, ↓ green.
+ *   - `neutral`   — direction carries no value judgement (the change renders
+ *                   muted).
+ */
+export type TrendDirectionSentiment = "up-good" | "up-bad" | "neutral";
+
+/**
+ * The resolved sentiment of a signed change under a metric's direction.
+ *
+ *  - `positive` — the change moves the metric toward its goal. Renders green.
+ *  - `negative` — the change moves away from the goal. Renders orange.
+ *  - `neutral`  — the metric's direction is neutral, or the change is below
+ *    the noise floor (|change| < 0.05). Renders muted.
+ */
+export type TrendSentimentDirection = "positive" | "negative" | "neutral";
+
+export function getTrendSentiment(
+  change: number | null | undefined,
+  sentiment: TrendDirectionSentiment,
+): TrendSentimentDirection {
+  if (change == null || Math.abs(change) < 0.05) return "neutral";
+  if (sentiment === "neutral") return "neutral";
+  const isUp = change > 0;
+  const isGood =
+    (sentiment === "up-good" && isUp) || (sentiment === "up-bad" && !isUp);
+  return isGood ? "positive" : "negative";
+}
+
+export function sentimentColorClass(
+  direction: TrendSentimentDirection,
+): string {
+  if (direction === "positive") return "text-dracula-green";
+  if (direction === "negative") return "text-dracula-orange";
+  return "text-muted-foreground";
+}
+
+/**
+ * v1.9.0 — map a `MetricDirection` (the metric-status registry's
+ * `higher-better` / `lower-better` / `target-band`) onto a
+ * `TrendDirectionSentiment` for the range-delta caption. `target-band`
+ * metrics (body temperature, respiratory rate, …) have no single good
+ * direction, so a period-over-period delta carries no sentiment colour —
+ * `neutral`, matching Oura's deviation framing for baseline metrics.
+ */
+export function sentimentFromMetricDirection(
+  direction: "higher-better" | "lower-better" | "target-band",
+): TrendDirectionSentiment {
+  if (direction === "higher-better") return "up-good";
+  if (direction === "lower-better") return "up-bad";
+  return "neutral";
+}
diff --git a/src/lib/openapi/registry.ts b/src/lib/openapi/registry.ts
index 8026500e0..762dc90c8 100644
--- a/src/lib/openapi/registry.ts
+++ b/src/lib/openapi/registry.ts
@@ -60,6 +60,7 @@ const openApiBase: Pick<
     { name: "Measurements" },
     { name: "Medications" },
     { name: "Mood" },
+    { name: "Analytics" },
     { name: "Insights" },
     { name: "Dashboard" },
     { name: "Notifications" },
diff --git a/src/lib/openapi/routes.ts b/src/lib/openapi/routes.ts
index 5e5fa6e90..b963b64b5 100644
--- a/src/lib/openapi/routes.ts
+++ b/src/lib/openapi/routes.ts
@@ -47,6 +47,7 @@ import { medicationExtractionSchema } from "@/lib/ai/coach/medication-extract-pr
 import { ACCEPTED_INSIGHTS_TILE_IDS } from "@/lib/insights-layout";
 import { exportSelectionSchema } from "@/lib/validations/health-record-export";
 import { METRIC_STATUS_IDS } from "@/lib/insights/metric-status-registry";
+import { ANALYTICS_RANGES } from "@/lib/analytics/range-delta";
 
 /**
  * Common envelopes — every HealthLog API response wraps payload in
@@ -1156,6 +1157,76 @@ const metricStatusResponse = z
       "Generic per-metric assessment envelope. Identical shape to the seven specialised `*-status` cards so the `InsightStatusCard` consumes it unchanged. Read-only + stale-while-revalidate: a cache miss warms a generation out of band and serves the last-good text meanwhile.",
   });
 
+const analyticsRangeQuery = z
+  .object({
+    type: measurementTypeEnum.describe(
+      "The measurement type to read (single metric — no fan-out). Closed enum: an unknown type 422s.",
+    ),
+    range: z
+      .enum(ANALYTICS_RANGES)
+      .describe(
+        "Trailing window: `7d` / `30d` / `90d` / `1y`. The previous comparable window is the equally-sized span immediately before it.",
+      ),
+  })
+  .meta({ id: "AnalyticsRangeQuery" });
+
+const analyticsWindowAggregate = z
+  .object({
+    count: z.number().int().describe("Reading count composed across buckets."),
+    min: z.number().nullable().describe("Window minimum; null when empty."),
+    max: z.number().nullable().describe("Window maximum; null when empty."),
+    mean: z
+      .number()
+      .nullable()
+      .describe("Count-weighted mean across buckets; null when empty."),
+    sum: z
+      .number()
+      .nullable()
+      .describe(
+        "Cumulative total for cumulative metrics (steps, energy, distance); null when no bucket carries a sum.",
+      ),
+  })
+  .meta({ id: "AnalyticsWindowAggregate" });
+
+const analyticsRangeResponse = z
+  .object({
+    range: z
+      .enum(ANALYTICS_RANGES)
+      .describe("The range that was read (echoes the request)."),
+    windowDays: z
+      .number()
+      .int()
+      .describe("Trailing-window length in days for the chosen range."),
+    granularity: z
+      .string()
+      .describe(
+        "Rollup granularity the read resolved against (`DAY` / `WEEK` / `MONTH` / `YEAR`, or `none` on a coverage miss).",
+      ),
+    current: analyticsWindowAggregate.describe(
+      "Aggregate over the current window `[now-N, now)`.",
+    ),
+    previous: analyticsWindowAggregate.describe(
+      "Aggregate over the previous comparable window `[now-2N, now-N)`.",
+    ),
+    delta: z
+      .number()
+      .nullable()
+      .describe(
+        "`current.mean - previous.mean`; null when either window has no data (never a misleading 0).",
+      ),
+    deltaPct: z
+      .number()
+      .nullable()
+      .describe(
+        "`delta / previous.mean` as a fraction (0.03 = +3 %); null when the prior window has no / zero mean (no divide-by-zero). The client shows 'no prior-period data' in that case.",
+      ),
+  })
+  .meta({
+    id: "AnalyticsRangeResponse",
+    description:
+      "Single-metric period-over-period aggregate. Reads the current and previous comparable windows from the WMY rollup tier and composes a count-weighted-mean delta. `count/min/max/mean/sum` are linearly composable across buckets; SD/slope/r² are intentionally excluded (not composable).",
+  });
+
 const insightsPregenerateRequest = z
   .object({})
   .meta({
@@ -2461,6 +2532,31 @@ export const openApiPaths: NonNullable<ZodOpenApiObject["paths"]> = {
       },
     },
   },
+  "/api/analytics/range": {
+    get: {
+      tags: ["Analytics"],
+      summary: "Single-metric period-over-period range delta",
+      description:
+        "v1.9.0 — returns the current-window aggregate, the previous comparable window, and the composed delta for ONE metric type over a `7d` / `30d` / `90d` / `1y` range. Single-type by construction (the metric page is single-metric), so the read is one rollup-tier call covering the trailing 2N days sliced into the two halves — no per-type fan-out. Additive route; the `/api/analytics` envelope is unchanged. Auth via cookie or Bearer.",
+      requestParams: {
+        query: analyticsRangeQuery,
+      },
+      responses: {
+        "200": {
+          description: "Current + previous window aggregates and the delta.",
+          content: {
+            "application/json": {
+              schema: dataEnvelope(
+                analyticsRangeResponse,
+                "AnalyticsRangeResponseEnvelope",
+              ),
+            },
+          },
+        },
+        ...stdResponses,
+      },
+    },
+  },
   "/api/insights/metric-status": {
     get: {
       tags: ["Insights"],
diff --git a/src/lib/query-keys.ts b/src/lib/query-keys.ts
index d9dd2c23b..92ed727bd 100644
--- a/src/lib/query-keys.ts
+++ b/src/lib/query-keys.ts
@@ -52,6 +52,16 @@ export const queryKeys = {
    */
   analytics: (slice?: "summaries") =>
     (slice ? (["analytics", slice] as const) : (["analytics"] as const)),
+  /**
+   * v1.9.0 — single-metric period-over-period range read
+   * (`GET /api/analytics/range`). A dedicated cache slot per `(type, range)`
+   * so switching the time-range pills is a cheap cache hit after the first
+   * fetch and never collides with the shared `["analytics", "summaries"]`
+   * slot the dashboard tile-strip subscribes to. `["analytics"]` is a prefix
+   * so a blanket `queryKeys.analytics()` invalidation still reaches it.
+   */
+  analyticsRange: (type: string, range: string) =>
+    ["analytics", "range", type, range] as const,
   moodAnalytics: () => ["mood-analytics"] as const,
   /**
    * v1.8.5 — pre-computed mood-insights aggregates (heatmap, distribution,