Verified against Claude · 2026-08-07
Turn your LinkedIn post analytics into one specific next-post hypothesis
Feed in the numbers LinkedIn's own post analytics panel gives you and get a plain-language read on what actually happened, plus exactly one testable change for the next post — not a re-explanation of the numbers you already have.
The prompt
Ready to copy — highlighted parts are example details you can swap.
The post, summarized: Text-only post, personal story about a missed deadline with a client, no link. The analytics as shown in LinkedIn's post-analytics panel — impressions, reactions, comments, reposts, and the follower vs. non-follower split from Discovery if you have it: Impressions: 14,200. Reactions: 310. Comments: 62. Reposts: 8. Discovery: 71% non-follower, 29% follower.. A comparable previous post's numbers, if I have one: Last month's carousel post: 9,800 impressions, 190 reactions, 14 comments, 44% non-follower. What I want the next post to do better: More comments from people I could plausibly work with, not just raw reach. Treat every number in Impressions: 14,200. Reactions: 310. Comments: 62. Reposts: 8. Discovery: 71% non-follower, 29% follower. as correct — do not recalculate, round, or estimate anything I didn't give you. Give me: 1. One-sentence verdict on whether this beat or underperformed relative to Last month's carousel post: 9,800 impressions, 190 reactions, 14 comments, 44% non-follower — if I didn't give you a comparison, say so instead of inventing a baseline 2. The specific signal that explains the result — for example, a high non-follower percentage suggests the algorithm pushed this beyond my existing network (often via a topic or hashtag), while a high comment-to-reaction ratio suggests the content itself triggered discussion rather than passive scrolling-by likes. Name which signal is actually present here, don't just say "engagement was good." 3. One concrete, testable change for the next post — a single variable (hook style, or format, or posting time, or topic), not three at once, so I can actually attribute the next result to something 4. One thing that's already working and shouldn't be changed, so I don't accidentally undo it while testing the other variable If the data given doesn't support a confident read on any of these, say what additional number you'd need instead of guessing.
Customize the highlighted detailsoptional — the prompt above already works
Why this works
LinkedIn's native analytics panel breaks impressions into a follower vs. non-follower split under Discovery, and reports reactions, comments, and reposts separately — these are different signals that call for different next moves, and collapsing them into a single "engagement was good" verdict wastes the one piece of real data available. A high non-follower percentage points at topic or hashtag-driven algorithmic push beyond your existing network; a high comment-to-reaction ratio points at the content itself provoking discussion rather than passive liking — this prompt is scoped to reading those signals honestly rather than recalculating anything, the same discipline used for interpreting ad performance data elsewhere in this library. Restricting the recommendation to one changed variable is what makes the next post an actual test instead of a guess with three simultaneous changes and no way to know afterward which one mattered.
What you get back
Verdict: this outperformed the comparison post on both reach and depth — impressions up 45%, but the more telling number is comments per impression, which is roughly 3x the carousel post's rate. Signal: 71% non-follower reach plus a high comment count together suggest this spread via genuine discussion, not just algorithmic hashtag push alone — the story format is doing real work here, not just topic luck. One variable to test next: try the same personal-story structure on a topic outside your usual niche, to see if the discussion-driving effect is about the format or about this specific story. Don't change: keep it text-only and link-free — you have no evidence a link or carousel format would have done better here, and changing it now would confound the next test.
Verified against
Claude Sonnet 5 · 2026-08-07
ChatGPT GPT-5.1 · 2026-08-07
Changelog
- 2026-08-07 — Initial version, scoped to interpretation-only after testing showed models will otherwise "helpfully" estimate missing metrics instead of asking for them.
Need this built into your business?
If a prompt isn't enough — Google Ads management, built and maintained for you — that's Scult's day job.
EXPLORE GOOGLE ADS MANAGEMENT
