Performance review

Best practices are averages over other people's accounts. This skill looks at the user's own posts and numbers to see which of those practices hold for them, and where their audience behaves differently.

1. Collect the numbers

With Saypage connected (get_account first; insights need can_read_insights):

• list_instagram_posts, following next_cursor until you have about 30 posts or 60 days. Leave stories out: they are listed only while live, and their insights end with them.
• get_instagram_post_insights for each post: reach, views, likes, comments, shares and saves, with average watch time for Reels, and profile visits and new followers for feed posts. Instagram gives Reels neither of those last two.
• get_instagram_insights over 30 days for the account as a whole.
• get_page_stats over 30 days: visits per page and channel, and clicks on each page's links.
• list_automations: which posts have a keyword automation behind them.

Without Saypage, ask for screenshots of the insights of the user's last posts.

State only numbers that a response returned. When a metric is in unavailable, pass on its reason (fewer than 100 followers, still processing, expired story) instead of a zero or an estimate. The API has no retention graph and no follower/non-follower split: the user can screenshot those from the app if they want them in the review.

2. Compute ratios, not totals

A big view count can simply mean a big following. Work out these, showing how:

Report page views from instagram_dm separately, with their page and 30-day window: they combine every automation sending that page and count page loads, not people, so do not attribute them to one post here. funnel-review follows one keyword from reach to page visits, and says when those numbers can be set side by side.

Order the posts by breakout ratio first and shares second. Keep formats apart: Reels against Reels, carousels against carousels.

3. Find the pattern

Split the posts into a top third and a bottom third and list what differs: subject, hook type (see reel-script), format, length, a face or hands in the opening frame, the ask, how fast the user answered comments. Posting day and hour come at the end of the list, since they seldom explain much.

Attach the evidence and a confidence level to each finding. Under about fifteen posts, present patterns as hunches to test, not conclusions.

4. Diagnose before advising

• Low reach: the hook and the opening frame (reel-script).
• Good reach, few new followers (feed posts, or the account's total, since Reels report none): visitors reach the profile and leave; look at the bio and its link (profile-review).
• Good reach, few shares or saves: people watch but do not find it useful or personal enough to keep or pass on.
• Comments but no sales, or DMs missing: that is one funnel's story, not the post's: use funnel-review.
• Reach with nowhere to go: a Reel that broke out without a keyword ask, or whose keyword no automation answers (comment_keyword in list_instagram_posts, nothing in list_automations), left its viewers only the bio link. Say what it could have brought, without inventing a number, and offer reel-funnel: on that Reel if it still draws comments, otherwise on the next one.

Output

REVIEW  ·  28 posts  ·  1 Sep - 25 Oct  ·  median views: Reels 3,900, carousels 1,700

TOP BY BREAKOUT RATIO
  9.1x  Reel      "I salted pasta water wrong..."  35,600 views  41 shares/1k  14 s watched
  4.3x  Carousel  5 weeknight sauces               7,300 views   58 saves/1k
  ...

BOTTOM
  0.3x  Reel      new menu announcement            1,200 views   2 shares/1k
  ...

WHAT THE NUMBERS SAY
1. Mistake-and-fix Reels: 4 of the top 5, none in the bottom 5 (fairly sure, n=9).
2. Announcements get reach but no shares: move them to stories.
3. The salt Reel reached 90,000 people with no keyword ask: nowhere to go from there.

KEEP   mistake-and-fix Reels, recipe carousels
STOP   menu announcements as Reels
TRY    a carousel version of the top Reel, with a keyword for the full recipe
       the next mistake-and-fix Reel closes on "Comment SALT": its page sent by DM

Then offer to build next week on it (content-calendar).