Post History Insights

Mines a brand's already-scraped post history to extract concrete performance signals: best posting times, winning formats, hashtags that worked, and posting cadence. Pure code, zero AI cost.

What it extracts

Signal	How	Example
Best times	Weekday + hour slots weighted by engagement	Tue 18:00, Thu 09:00
Top formats	Media type buckets (video/carousel/story/image) weighted by engagement	video, carousel
Top hashtags	Hashtags from captions weighted by engagement (unicode-aware)	#branding #design #sustainable
Cadence	Posts per week/month from timestamp span	~4 posts/week

Engagement weight: prefers real engagementRate when available. Otherwise: likes + comments × 2 (comments are a stronger signal).

Caveat: times are derived in UTC (we don't know the audience timezone). Treat them as approximate.

The digest

Serialised into a compact block injected into ai_context:

WHAT WORKS HERE (mined from the brand's own 47 past posts):
- Best times to post (UTC): Tue 18:00, Thu 09:00, Sat 12:00
- Formats that perform here: video, carousel
- Hashtags that have worked: #branding #design #sustainable
- Typical cadence: ~4 posts/week

This flows into every planner and content batch — schedules at proven best times, reuses winning formats and hashtags, matches the brand's cadence.

Refreshed during the flywheel rebuild (every autopilot run).