Analyzer playbook
Seven questions that decide your next Reel.
These are the routines we run every day on our own Instagram with the Mica Social Media Analyzer. Each one is a question you type into ChatGPT or Claude. The answer comes from your account, read hour by hour, compared against your own history.
What your AI can see
The hours Instagram does not keep for you.
Every Reel and post, hour by hour
Every hour for its first 24 hours, then about every six hours out to 72. Each Reel reading keeps skip rate, saves and saves per hundred views, shares, views and reach at that moment.
Stories while they are live
Read every hour while Instagram still serves their numbers.
Account and audience
Account totals and your audience, read once a day and kept as a history.
What it showed us
On our own account, the first hours called it.
Five views apart at hour one. 1,716 against 121 at hour six.
Two of our Reels opened almost level. By hour two their skip rates were 44 and 58 percent, and by hour six one had pulled away. A day-one total tells you this a day late.
DeepSeek costs 1/100th
0views
38% skip at hour 2
3 Claude Code fixes
0views
44% skip at hour 2
A new AI model, reacted to
0views
58% skip at hour 2
Hour
Views in the first 24 hours. Hover or drag across the chart to read any hour.
See the readings as a table
| Nearest hour | DeepSeek costs 1/100th | 3 Claude Code fixes | A new AI model, reacted to |
|---|---|---|---|
| 1 | 68 | 118 | 113 |
| 2 | 166 | 144 | 120 |
| 3 | 191 | 192 | 121 |
| 4 | 265 | 533 | 121 |
| 5 | 340 | 1,384 | 121 |
| 6 | 495 | 1,716 | 121 |
| 8 | 1,298 | 1,919 | 121 |
| 12 | 1,908 | 2,070 | 122 |
| 18 | 3,384 | 2,199 | 122 |
| 24 | 4,637 | 2,458 | 122 |
Skip rate split 31 Reels.
All 8 at 41 percent skip or under cleared 1,200 views in 24 hours. None of the 22 at 47 percent or over reached 1,000. One at 43 percent stalled at 112.
Each dot is one Reel
See all 31 Reels as a table
| Skip rate | Views at 24 hours |
|---|---|
| 32% | 2,472 |
| 33.5% | 1,319 |
| 35.5% | 5,079 |
| 37.6% | 4,759 |
| 37.8% | 1,272 |
| 39.6% | 1,658 |
| 40.5% | 2,018 |
| 41% | 2,447 |
| 43.3% | 112 |
| 47% | 580 |
| 48.6% | 155 |
| 48.8% | 978 |
| 50.7% | 805 |
| 51.7% | 160 |
| 51.9% | 149 |
| 53.3% | 481 |
| 54.7% | 227 |
| 57.1% | 291 |
| 57.4% | 122 |
| 58.8% | 353 |
| 58.8% | 229 |
| 58.9% | 563 |
| 60.7% | 276 |
| 62.4% | 129 |
| 63.3% | 130 |
| 67.3% | 123 |
| 68.1% | 124 |
| 68.4% | 147 |
| 71.9% | 35 |
| 75.9% | 62 |
| 85% | 113 |
- 97.5%
- of the accounts we reached over 28 days did not follow us.
- 2 to 1
- saves over likes across the same 28 days.
Numbers are from our own account, read through the Analyzer on 27 September 2026. Your account will have its own pattern.
The routines
What we ask, and what it changes.
Check a new Reel at hours 1, 3, 12 and 24
You see whether a Reel is speeding up or stalling while you can still act on it, long before the totals settle.
Ask ChatGPT or Claude
How is my newest Reel doing at hour 3 compared with my last 20?
Read skip rate first
On our account it called the outcome Reel after Reel. It tells you whether the opening held people before anything else matters.
Ask ChatGPT or Claude
Rank my last 30 Reels by skip rate and show views at 24 hours next to each.
Gate the opening at hour 3, decide at hour 12
A high skip rate at hour 3 points at the opening. We wait for hour 12 before recutting, because we have watched a Reel sit flat for hours and then take off.
Ask ChatGPT or Claude
Did my last Reel pick up again after hour 6, and when did it start?
Tag your hooks and let the results sort them
Label each Reel's opener, experiment and variant as you post. After a few weeks you know which kind of opening works for your audience, from your own numbers.
Ask ChatGPT or Claude
Tag my newest Reel's hook as a statement opener, then compare statement openers with question openers.
Judge a new format the same day
The hour-by-hour curve shows a flop within hours, so you fix the opening while the idea is still fresh instead of waiting a month.
Ask ChatGPT or Claude
My new format posted this morning. Is it on track against my usual Reels at the same hour?
Compare trial Reels with trial Reels
Trial Reels go only to people who do not follow you. Tag a Reel as a trial and it is measured against your other trials, so you know whether to share it to your followers.
Ask ChatGPT or Claude
Is my latest trial Reel worth sharing to my followers?
Make more of what people save and send
Saves and shares per hundred views show what people kept and passed on. Likes told us the least.
Ask ChatGPT or Claude
Which of my Reels from the last month got the most saves and shares per 100 views, and what do they have in common?
Good to know
- Read only. The Analyzer never posts, replies, or changes anything on Instagram.
- Hourly readings cover each post's first 72 hours.
- It works with an Instagram professional account (Business or Creator).
- Views are Instagram views. Plays from a Facebook crosspost are not counted.
- During the private beta, connecting takes one extra step: you accept a tester invitation inside Instagram. We email you when it is waiting.
Run the playbook on your own account.
Seven days free with a card on file, then $29 a month. Cancel before day eight and pay nothing.
30-day money-back guarantee on your first month. Email guru@graniteai.co and we refund it, no questions.