Meta Insights & Analytics

Meta Insights surfaces the analytics that Meta computes for your WhatsApp Business account — messaging volume, conversation pricing, template performance, and calling stats. Whatomate proxies these live from the Meta Graph API and caches the results in Redis so repeated dashboard loads don't re-hit Meta on every view.

Note

These are Meta's own aggregate analytics, distinct from Whatomate's internal dashboards (message counts, agent performance, chatbot sessions). See API Reference → Analytics for the internal analytics endpoints.

Analytics types

GET /api/analytics/meta proxies four different Graph API insight sets, selected with the analytics_type query parameter:

analytics_type	What it reports
analytics	Messaging analytics — sent/delivered message volume over the range.
pricing_analytics	Pricing analytics — conversation counts and cost broken down by pricing category.
template_analytics	Template analytics — per-template sent/delivered/read/clicked. Limited to a 90-day lookback.
call_analytics	Call analytics — WhatsApp Business Calling volume and outcomes.

Caution

Template analytics have a 90-day lookback limit. A start date older than 90 days is rejected with 400. For template analytics you can optionally pass template_ids (a JSON array); if omitted, Whatomate auto-selects the account's synced templates. The response includes a template_names map so raw Meta template IDs render as human-readable names.

Query parameters

Param	Notes
analytics_type	Required. One of the four values above.
start, end	Required. YYYY-MM-DD. end must not be before start.
account_id	Optional. A specific WhatsApp account UUID, or omit to query all accounts in the org.
granularity	Optional. HALF_HOUR, DAY, or MONTH. Defaults to DAY.
template_ids	Optional, template analytics only. JSON array of Meta template IDs.

Use GET /api/analytics/meta/accounts to list the accounts available for analytics (id, name, phone_id).

Auto-adjusted granularity

Whatomate normalizes granularity to what the Graph API will actually accept for the given range, avoiding Meta-side errors:

• MONTH requires at least 30 days — a shorter range is downgraded to DAY.
• HALF_HOUR only makes sense for short ranges — anything over 7 days is downgraded to DAY.

When an adjustment happens, the response echoes both adjusted_granularity and original_granularity so the UI can explain why the resolution changed.

Caching & manual refresh

Results are cached in Redis under the meta:analytics: key prefix. The cache key is scoped per organization, account (or all), analytics type, date range, and granularity — so different views don't collide.

The TTL scales with granularity, since coarser data changes more slowly:

Granularity	Cache TTL
HALF_HOUR	1 hour
DAY	3 hours
MONTH	6 hours

Each response carries a cached boolean so you can tell a fresh Graph API fetch from a cache hit. To force fresh data, call:

POST /api/analytics/meta/refresh

This clears every cached analytics entry for the current organization (all accounts, types, and ranges). Reading analytics requires analytics:read; the refresh endpoint requires analytics:write.

Response shape

{
  "status": "success",
  "data": {
    "accounts": [
      {
        "account_id": "<uuid>",
        "account_name": "Support Line",
        "data": { /* Meta Graph API analytics payload */ },
        "template_names": { "<meta_template_id>": "order_update" }
      }
    ],
    "cached": false
  }
}

When an account has no data (e.g. no templates for template analytics, or a Graph API error on that one account), its data is null and the other accounts still return — one failing account never sinks the whole request.

What these insights do NOT include

Caution

Per-phone quality rating and messaging-limit tier are not part of Meta Insights. They are not returned by these analytics endpoints. Those values live on the account detail — Whatomate reads a phone number's quality_rating (and verification status) when validating account credentials against Meta, and surfaces them under Settings → Accounts → [account]. Look there, not here, for a number's current quality rating or tier.