diff --git a/‎.changeset/21995-measure-column-aggregate.md‎
Lines changed: 14 additions & 0 deletions b/‎.changeset/21995-measure-column-aggregate.md‎
Lines changed: 14 additions & 0 deletions
diff --git a/‎content/docs/api/data-api.mdx‎
Lines changed: 8 additions & 1 deletion b/‎content/docs/api/data-api.mdx‎
Lines changed: 8 additions & 1 deletion
diff --git a/‎packages/services/service-analytics/src/__tests__/measure-column-aggregate.test.ts‎
Lines changed: 225 additions & 0 deletions b/‎packages/services/service-analytics/src/__tests__/measure-column-aggregate.test.ts‎
Lines changed: 225 additions & 0 deletions
diff --git a/‎packages/services/service-analytics/src/__tests__/preview-column-enrichment.test.ts‎
Lines changed: 3 additions & 2 deletions b/‎packages/services/service-analytics/src/__tests__/preview-column-enrichment.test.ts‎
Lines changed: 3 additions & 2 deletions
diff --git a/‎packages/services/service-analytics/src/analytics-service.ts‎
Lines changed: 14 additions & 4 deletions b/‎packages/services/service-analytics/src/analytics-service.ts‎
Lines changed: 14 additions & 4 deletions
@@ -0,0 +1,14 @@
+---
+"@objectstack/spec": minor
+"@objectstack/service-analytics": minor
+---
+
+feat(spec,analytics): a dataset answer's measure column states its aggregate, labelled or not (`fields[].aggregate`)
+
+Clause-②: yes (widening)
+
+- **What a renderer can now read.** Each measure column of a dataset answer (`POST /analytics/dataset/query`) carries `fields[].aggregate`: the aggregate its dataset measure declares, in the closed `AggregationFunction` vocabulary (`count`, `sum`, `avg`, `min`, `max`, `count_distinct`). It is there whether or not the author gave the measure a `label`. So a chart can tell a count from a sum, for example to draw whole-number axis ticks for a count instead of 0.75 / 1.5 / 2.25.
+- **What was missing.** The only aggregate on the wire was `builtinAggregate`, and it is present only when the measure has no `label`. A labelled measure, such as a `count` named "Tasks", reached the wire as `{ name, type: 'number', label }`, with nothing to say what kind of number it was.
+- **Where it is set.** `AnalyticsService` writes it in the one step that describes a dataset answer's columns from the dataset's own measures. That step runs for both the live query and the draft-data preview, so the two answers agree. A measure's `__compare` column carries the same aggregate.
+- **Where it is absent.** Dimension columns. Derived measures, which combine other measures and have no single aggregate (a stray `aggregate` written beside `derived` is ignored when the dataset compiles, so it is not stated here either). And a cube query answer (`POST /analytics/query`), which does not run through the dataset column step.
+- **Unchanged.** `builtinAggregate` keeps its meaning: present only on a label-less measure column, to mark a header that is the server's default. No authoring key is added; `aggregate` is a response member only. `AnalyticsResultResponseSchema` and the `AnalyticsResult` contract declare the member, and the REST route relays it as it does every other column key.
@@ -461,7 +461,7 @@ is executed. `POST /analytics/sql` refuses the same keys. To order by a member,
 <Callout type="info">
 **`fields[]` is the resolved presentation surface — read it first.** Each entry carries
 `name` and `type` and, when the producer declares them, `label`, `format`, `currency`,
-`percentScale` and `builtinAggregate`: the optional members
+`percentScale`, `builtinAggregate` and `aggregate`: the optional members
 `AnalyticsResultResponseSchema` declares (`packages/spec/src/api/analytics.zod.ts`),
 mirrored member for member by `IAnalyticsService.query`'s `AnalyticsResult` and bound to
 it at compile time. Optional means a given column may omit them — read them defensively,
@@ -485,6 +485,13 @@ tenant default currency (the `localization.currency` setting), if one is set. A
 `dynamic` field (the default mode) never lends its `defaultCurrency` — only `fixed` does —
 so its column shows the tenant default. The cube query on this page carries no column
 `currency`.
+
+The dataset query also states each measure column's `aggregate` (`count`, `sum`, `avg`,
+`min`, `max`, `count_distinct`), whether or not the author labelled the measure, so a
+chart can tell a count from a sum. `builtinAggregate` is narrower: it is present only
+when the measure has no `label`, to mark a header that is the server's default. Neither
+is set on a dimension column or a derived measure, and the cube query on this page
+carries neither.
 </Callout>
 
 ### `GET /analytics/meta`
 
@@ -0,0 +1,225 @@
+// Copyright (c) 2026 ObjectStack. Licensed under the Apache-2.0 license.
+
+/**
+ * `fields[].aggregate` — every measure column of a dataset answer states the
+ * aggregate its measure declares, whether or not the author labelled it.
+ *
+ * `builtinAggregate` answers one question only: "is this header the server's
+ * default?". It is absent the moment an author writes a `label`, and every
+ * measured showcase widget labels its measure (a `count` named "Tasks"). So on
+ * exactly the columns a dashboard draws, the answer said `{ name, type:
+ * 'number', label }` and nothing else, and a chart over a count drew 0.75 /
+ * 1.5 / 2.25 axis ticks because it could not tell the count from a sum.
+ *
+ * The aggregate is part of the dataset's own authored measure, so the ADR-0021
+ * column-description seam (`enrichResultColumns`) states it from there, like
+ * `label` / `format` / `currency` / `percentScale`. That seam serves both paths
+ * that produce a dataset answer, the live engine query and the ADR-0037 P3
+ * draft-data preview, so each pin below is asserted on both.
+ *
+ * Reverse verification, direction predicted BEFORE running: deleting the one
+ * producer line that writes `f.aggregate` turns every positive `aggregate`
+ * assertion RED on both paths and leaves the absence assertions and every
+ * `builtinAggregate` assertion GREEN. Deleting only its `!m.derived` guard
+ * turns the stray-aggregate case RED and nothing else.
+ */
+
+import { describe, it, expect } from 'vitest';
+import { DatasetSchema } from '@objectstack/spec/ui';
+import type { ExecutionContext } from '@objectstack/spec/kernel';
+import { AnalyticsService } from '../analytics-service.js';
+
+interface Task extends Record<string, unknown> {
+  status: string;
+  amount: number;
+  due_on: string;
+}
+
+const ROWS: Task[] = [
+  { status: 'open', amount: 100, due_on: '2026-02-03' },
+  { status: 'open', amount: 50, due_on: '2026-02-10' },
+  { status: 'done', amount: 25, due_on: '2026-02-15' },
+  { status: 'done', amount: 10, due_on: '2026-01-20' },
+];
+
+const DATASET = DatasetSchema.parse({
+  name: 'task_ds',
+  label: 'Tasks',
+  object: 'task',
+  dimensions: [
+    { name: 'status', field: 'status', type: 'string', label: 'Status' },
+    { name: 'due_on', field: 'due_on', type: 'date', label: 'Due' },
+  ],
+  measures: [
+    // The showcase shape: a `count` the author named.
+    { name: 'task_count', aggregate: 'count', label: 'Tasks' },
+    // The built-in default: no label, so `builtinAggregate` too.
+    { name: 'count', aggregate: 'count' },
+    // A `sum` over a currency field.
+    { name: 'total_amount', aggregate: 'sum', field: 'amount', label: 'Total Amount', format: '$0,0' },
+    // A derived measure has no single aggregate.
+    { name: 'open_share', derived: { op: 'ratio', of: ['task_count', 'count'] }, label: 'Share' },
+  ],
+});
+
+const sourceFieldMeta = (object: string, field: string) => {
+  if (object !== 'task') return undefined;
+  if (field === 'amount') return { type: 'currency' };
+  if (field === 'due_on') return { type: 'date' };
+  return undefined;
+};
+
+const CTX = { tenantId: 'org_A', currency: 'USD' } as ExecutionContext;
+
+/** Enough of a GROUP BY for this fixture: the LIVE path's engine. */
+function evaluateAggregate(opts: { groupBy?: unknown; aggregations?: unknown }) {
+  const groupBy = (opts.groupBy ?? []) as Array<string | { field: string }>;
+  const aggs = (opts.aggregations ?? []) as Array<{ field: string; method: string; alias: string }>;
+  const buckets = new Map<string, { key: Record<string, unknown>; rows: Task[] }>();
+  for (const r of ROWS) {
+    const key: Record<string, unknown> = {};
+    for (const g of groupBy) {
+      const f = typeof g === 'string' ? g : g.field;
+      key[f] = r[f];
+    }
+    const id = JSON.stringify(Object.values(key));
+    let b = buckets.get(id);
+    if (!b) { b = { key, rows: [] }; buckets.set(id, b); }
+    b.rows.push(r);
+  }
+  return [...buckets.values()].map(({ key, rows }) => {
+    const row: Record<string, unknown> = { ...key };
+    for (const a of aggs) {
+      row[a.alias] = a.method === 'sum'
+        ? rows.reduce((s, r) => s + Number(r[a.field] ?? 0), 0)
+        : rows.length;
+    }
+    return row;
+  });
+}
+
+/** Two services that differ ONLY in whether a pending seed draft exists. */
+function svc(preview: boolean) {
+  return new AnalyticsService({
+    sourceFieldMeta,
+    queryCapabilities: () => ({ nativeSql: false, objectqlAggregate: true, inMemory: false }),
+    executeAggregate: async (_object: string, options: Record<string, unknown>) => evaluateAggregate(options),
+    ...(preview ? { draftRowsResolver: async () => ROWS as Record<string, unknown>[] } : {}),
+  });
+}
+
+type Field = Awaited<ReturnType<AnalyticsService['queryDataset']>>['fields'][number];
+const byName = (fields: Field[]) => Object.fromEntries(fields.map((f) => [f.name, f]));
+
+async function bothPaths(dataset = DATASET, selection: Record<string, unknown> = {
+  dimensions: ['status'],
+  measures: ['task_count', 'count', 'total_amount', 'open_share'],
+}) {
+  const live = await svc(false).queryDataset(dataset, selection as never, CTX);
+  const preview = await svc(true).queryDataset(dataset, selection as never, CTX, { previewDrafts: true });
+  return { live: byName(live.fields), preview: byName(preview.fields) };
+}
+
+describe('fields[].aggregate — a dataset answer states each measure column\'s aggregate', () => {
+  it('a labelled `count` measure states `count`, and an unlabelled one still carries builtinAggregate', async () => {
+    const { live, preview } = await bothPaths();
+    for (const [path, fields] of [['live', live], ['preview', preview]] as const) {
+      // The showcase column: the author's label, and now the aggregate beside it.
+      expect(fields.task_count?.label, path).toBe('Tasks');
+      expect(fields.task_count?.aggregate, path).toBe('count');
+      // `builtinAggregate` keeps its label-only meaning: absent under a label…
+      expect(fields.task_count?.builtinAggregate, path).toBeUndefined();
+      // …and present, beside the new member, on the label-less default.
+      expect(fields.count?.builtinAggregate, path).toBe('count');
+      expect(fields.count?.aggregate, path).toBe('count');
+      expect(fields.count?.label, path).toBeUndefined();
+    }
+  });
+
+  it('a `sum` over a currency field states `sum`', async () => {
+    const { live, preview } = await bothPaths();
+    for (const [path, fields] of [['live', live], ['preview', preview]] as const) {
+      expect(fields.total_amount?.aggregate, path).toBe('sum');
+      // The column's other descriptors are untouched by the new member.
+      expect(fields.total_amount?.currency, path).toBe('USD');
+      expect(fields.total_amount?.format, path).toBe('$0,0');
+      expect(fields.total_amount?.type, path).toBe('number');
+    }
+  });
+
+  it('the preview path states the same as the live path, column for column', async () => {
+    const { live, preview } = await bothPaths();
+    expect(Object.keys(preview).sort()).toEqual(Object.keys(live).sort());
+    for (const name of Object.keys(live)) {
+      expect(preview[name]?.aggregate, `column "${name}"`).toBe(live[name]?.aggregate);
+    }
+  });
+
+  it('is absent on a dimension column and on a derived measure, on both paths', async () => {
+    const { live, preview } = await bothPaths();
+    for (const [path, fields] of [['live', live], ['preview', preview]] as const) {
+      expect(fields.status, path).toBeDefined();
+      expect(fields.status?.aggregate, path).toBeUndefined();
+      expect(fields.open_share, path).toBeDefined();
+      expect(fields.open_share?.aggregate, path).toBeUndefined();
+    }
+  });
+
+  it('a derived measure that also declares a stray `aggregate` states none: the compiler ignores it', async () => {
+    const stray = DatasetSchema.parse({
+      ...DATASET,
+      name: 'task_ds_stray',
+      measures: [
+        { name: 'task_count', aggregate: 'count', label: 'Tasks' },
+        { name: 'count', aggregate: 'count' },
+        { name: 'open_share', derived: { op: 'ratio', of: ['task_count', 'count'] }, aggregate: 'sum', label: 'Share' },
+      ],
+    });
+    const { live, preview } = await bothPaths(stray, { dimensions: ['status'], measures: ['task_count', 'count', 'open_share'] });
+    for (const [path, fields] of [['live', live], ['preview', preview]] as const) {
+      expect(fields.open_share, path).toBeDefined();
+      expect(fields.open_share?.aggregate, path).toBeUndefined();
+      // Control: the base measures beside it are still described.
+      expect(fields.task_count?.aggregate, path).toBe('count');
+    }
+  });
+
+  it('a `__compare` column states its measure\'s aggregate, on both paths', async () => {
+    const selection = {
+      dimensions: ['status'],
+      measures: ['task_count', 'total_amount'],
+      timeDimensions: [{ dimension: 'due_on', dateRange: ['2026-02-01', '2026-02-28'] }],
+      compareTo: { kind: 'previousPeriod', dimension: 'due_on' },
+    };
+    const { live, preview } = await bothPaths(DATASET, selection);
+    for (const [path, fields] of [['live', live], ['preview', preview]] as const) {
+      expect(fields.task_count__compare, path).toBeDefined();
+      expect(fields.task_count__compare?.aggregate, path).toBe('count');
+      expect(fields.total_amount__compare?.aggregate, path).toBe('sum');
+    }
+  });
+});
+
+describe('fields[].aggregate — absent on a cube query answer', () => {
+  it('`query()` (POST /analytics/query) never passes through the dataset column seam, so it states none', async () => {
+    const service = new AnalyticsService({
+      cubes: [{
+        name: 'task_cube',
+        title: 'Tasks',
+        sql: 'task',
+        measures: { task_count: { label: 'Tasks', type: 'count', sql: '*' } },
+        dimensions: { status: { label: 'Status', type: 'string', sql: 'status' } },
+      } as never],
+      sourceFieldMeta,
+      queryCapabilities: () => ({ nativeSql: false, objectqlAggregate: true, inMemory: false }),
+      executeAggregate: async (_object: string, options: Record<string, unknown>) => evaluateAggregate(options),
+    });
+    const result = await service.query(
+      { cube: 'task_cube', measures: ['task_count'], dimensions: ['status'] },
+      CTX,
+    );
+    const measure = result.fields.find((f) => f.name === 'task_count');
+    expect(measure).toBeDefined();
+    expect(measure?.aggregate).toBeUndefined();
+  });
+});
@@ -48,6 +48,7 @@
  * | `label`            | `measure.label` / `dimension.label` + `ctx.locale` (#6761) |
  * | `format`           | `measure.format`                                 |
  * | `builtinAggregate` | `measure.aggregate` + `measure.label == null` (#14492) |
+ * | `aggregate`        | `measure.aggregate`, labelled or not; never on a `derived` measure |
  * | `currency`         | `measure.currency` → the source field's FIXED currency (`sourceFieldMeta().defaultCurrency`, relayed only under `currencyMode: 'fixed'`) → `ctx.currency` |
  * | `percentScale`     | `measure.derived.op === 'ratio'`, else `percentScaleOf(sourceFieldMeta())` (objectui#3136) |
  * | `type`             | `measureResultType(measure.aggregate, sourceFieldMeta().type)` (#16101) |
@@ -238,10 +239,10 @@ async function bothPaths() {
   return { live: by(live.fields), preview: by(preview.fields) };
 }
 
-/** The six keys the card tabulates, absent ones dropped so they read as absent. */
+/** The keys the table above lists, absent ones dropped so they read as absent. */
 function descriptor(f: Field | undefined): Record<string, unknown> {
   const out: Record<string, unknown> = {};
-  for (const k of ['label', 'format', 'currency', 'percentScale', 'builtinAggregate', 'type'] as const) {
+  for (const k of ['label', 'format', 'currency', 'percentScale', 'builtinAggregate', 'aggregate', 'type'] as const) {
     const v = (f as Record<string, unknown> | undefined)?.[k];
     if (v != null) out[k] = v;
   }
 
@@ -2492,8 +2492,8 @@ export class AnalyticsService implements IAnalyticsService {
         const previewResult = await new DatasetExecutor(previewService).execute(compiled, selection, context);
         // ADR-0021 result-column enrichment runs on this path too. Every key it
         // writes describes the dataset's OWN authored columns — a measure's
-        // `label` / `format` / `currency` / `percentScale` / `builtinAggregate`
-        // and the `type` its aggregate really returns, plus a dimension column's
+        // `label` / `format` / `currency` / `percentScale` / `builtinAggregate` /
+        // `aggregate` and the `type` its aggregate really returns, plus a dimension column's
         // header `label` — all read off the dataset definition and
         // `sourceFieldMeta`, never off the rows. #16097: this early `return`
         // used to sit ~250 lines ahead of that block, so the same dataset in the
@@ -2777,8 +2777,8 @@ export class AnalyticsService implements IAnalyticsService {
   /**
    * ADR-0021 — describe the result's COLUMNS from the dataset's own authored
    * definition: a measure's `label` / `format` / `currency` / `percentScale` /
-   * `builtinAggregate` and the `type` its aggregate really returns, then a
-   * dimension column's header `label`.
+   * `builtinAggregate` / `aggregate` and the `type` its aggregate really
+   * returns, then a dimension column's header `label`.
    *
    * **Every key here is read off the DATASET** (the authored measure or
    * dimension) **and `sourceFieldMeta`** (the source object's declared field
@@ -2864,6 +2864,16 @@ export class AnalyticsService implements IAnalyticsService {
         // #14492: it would catch an author who really named a field `Count`,
         // and break the moment the default is spelled in another language.
         if (f.builtinAggregate == null && m.label == null && m.aggregate) f.builtinAggregate = m.aggregate;
+        // The aggregate itself, stated whatever the header says. The
+        // discriminator above answers only "is this header the server's
+        // default?", so a LABELLED `count` ("Tasks") reached the wire as a bare
+        // `type: 'number'` and a chart could not tell it from a `sum`: it drew
+        // 0.75 / 1.5 / 2.25 ticks on a count axis. Read off the authored
+        // measure like every other key here, so the live and preview paths
+        // agree by construction. ⛔ Not on a derived measure: the compiler
+        // ignores a `derived` measure's stray `aggregate` and computes it from
+        // its `of` measures, so that `aggregate` describes nothing on the wire.
+        if (f.aggregate == null && !m.derived && m.aggregate) f.aggregate = m.aggregate;
         if (f.format == null && m.format) f.format = m.format;
         // Currency chain. A MONETARY measure resolves its display currency
         // from: explicit measure `currency` → the source field's FIXED