Insights

Which kinds of posts perform best, for whom, and where the under-exploited opportunities are — by topic, format, visual style, or referenced expert. Organic-only by default so conclusions measure content quality, not ad spend.

Dimension
Period
Reach
Boosted
Accounts

No classified posts for this selection. Run npm run classify:backfill to label history.

AI trend narrationSign in to generate ↗

Grounded summary from aggregates only (no raw posts), via claude-opus-4-8.

How to read this
  • Organic vs. paid are kept separate. A "best post" list that silently included boosted reach would just be measuring ad spend — defaults are organic-only, boosted excluded.
  • Rank by median, not average — one viral post badly skews the mean. Trust the median and the sample size.
  • Where an account lacks per-post impressions for a category, its engagement rate shows N/A and it's ranked by raw engagement only — don't compare its raw counts against the others' rates.
  • Categories with <5 posts are flagged "insufficient" and 5–10 "low-confidence" — they'll be noisy until more posts accumulate.
  • Topic / visual-style labels are LLM estimates with a confidence score; low-confidence ones (and any manual overrides) are meant for human review.
  • Format, topic, and visual style correlate (podcast ≈ video; company_personal ≈ photo_people). When "video wins", check whether it's really topic or author driving it.

Boost effectiveness

Organic baseline vs paid lift vs spend for boosted posts — the highest-value question this rebuild answers: is boosting buying reach, or resonance? Paid reach is attributed to the real author (import an Ads CSV on the Ad import page to populate this).

loading boost data…

Engagement by media format

Average engagement per post by media format — now using the full taxonomy (video, carousel, document, image, link, text-only), not the old image/text-only split. Click a format to see those posts.

CEO

FormatPostsAvg engagement
text_only830
text_image2229
text_video2223

Company

You