Social Growth Analyst Complete
What this does
You send four screenshots per platform. You get a diagnosis built entirely from your own numbers — not generic advice that treats your account like everyone else's.
Most creators optimise for the wrong metric. The posts with the most views are usually not the posts bringing new followers, and that gap stays invisible until someone compares the two columns side by side. That comparison is the first thing this agent does.
The three editions
- Social Growth Analyst — one platform of your choice: https://capafy.ai/en/agent/social-growth-analyst/5496559849
- Social Growth Analyst Duo — two platforms, compared side by side: https://capafy.ai/en/agent/social-growth-analyst-duo/4604900740
- Social Growth Analyst Complete — all four platforms in one session
Supported platforms
Instagram · TikTok · Facebook · YouTube Shorts
Analyse one platform or all four in the same upload. When several are present the agent compares them directly: which one converts new viewers into followers, where a format that works on one dies on another, and whether your effort is going where your return is.
What you send
Four screenshots per platform, from your phone. Three minutes.
- Your analytics overview — the screen with period totals: views, interactions, new followers
- Your posts sorted by views
- The same posts sorted by new followers — same list, different sort. This is the important one
- Your profile as a stranger sees it — bio, picture, link
Keep the date range visible in every screenshot. It's usually a row of buttons near the top. A screenshot without its period can't be analysed.
Screen-by-screen guide with visual maps: https://capafy.psdt.pro/help
On a computer? Add a CSV export on top of the screenshots for a deeper read — every post in the period instead of the visible top few. It never replaces the screenshots: the follower ranking and TikTok's traffic split exist only in the app.
What you get back
Your baseline — the median your account actually performs at, and how far your outliers sit from it. Most creators have never seen this number.
Six diagnostics
- Reach composition — how much of your reach comes from people who don't follow you yet, whenever your data carries that split. It's the growth engine, and almost nobody looks at it.
- Follow conversion — which posts turn new viewers into followers. Frequently not your biggest posts.
- Retention — where viewers drop off, and whether that points to your opening, your pacing, or one specific moment. Video formats only, and only when watch time is in your data.
- Signal quality — shares predict reach, saves predict value, likes are mostly noise. Which signals is your content actually generating?
- Your profile — how many of the people who land on your profile end up following. Read as a ceiling, since a viewer can also follow straight from the feed, it still tells you when the profile is the thing losing them, and what in the bio is causing it.
- Time behaviour — whether your posts die in 48 hours or keep earning views for weeks. This decides whether you chase trends or build a library.
A plan for the next 30 days — how many posts and of which kind, plus one experiment with its metric and its duration. Concrete enough to start tomorrow, and every item names the metric that will show whether it worked.
A continuity card — a short block at the end of the report. Keep it and paste it back next time: the following report opens by telling you whether the plan worked.
An honest list of what the data can't tell you, and which screen would close the gap.
The check-in
Around the middle of your cycle, send just the posts published since the last report. The agent checks whether those actions moved the metrics they were meant to move — and says so plainly when they didn't.
That's what makes this compound. A report nobody follows up on is a document. A report that gets checked is a process.
How it's different
Everything comes from your data. No invented industry benchmarks, no "accounts like yours typically see" — those numbers are made up and they're wrong for you specifically.
It says what it can't determine. Small sample, short period, missing metric: you'll be told, instead of having the gap filled with a confident guess.
It checks the boring explanations first. Trending audio, a repost from a bigger account, seasonality, or simply a normal week following a viral one — the most common false alarm in social analytics. Your content gets blamed only after the alternatives are ruled out.
It will contradict you. When your data disagrees with what you believe about your own account, that contradiction is usually the most valuable line in the report.
What it does not do
- No account connection. It never asks for a password, a login or an access token. You export your own data and upload it yourself.
- No growth promises. It reads numbers and proposes experiments. Results are yours to earn.
- No competitor spying. Private analytics belong to their owner. Upload only accounts you own or manage.
Requests
6 requests per month on the monthly plan, and 6 per month on the annual plan too — 72 over the year. The annual discount is in the price, not in the quota.
A request is one message you send, not one analysis — the platform meters it that way, so a follow-up question costs the same as an upload.
This is why the agent delivers the whole report in a single reply and asks for anything missing all at once. Send every platform together in one message: it's cheaper for you, and it's also the better analysis, because comparing platforms surfaces things a single-platform view never shows.
Suggested rhythm: one full analysis and one check-in per month. That leaves room to spare.
Good to know
Reports are written in your language — just write to the agent in it.
The analysis is produced by a language model and can be incomplete or wrong. Check figures against your platform dashboards before acting on them.
Questions: [email protected]
