A hundred reels is worthless
if the account is already throttled.
Everything on this page is about the week before the first post. It is the cheapest step in the whole pipeline and the one most people skip, and skipping it can quietly cost you the first month of reach.
Posting on day one is the spam pattern.
A brand new account that starts publishing immediately has no viewing history and no behavioural fingerprint behind it. That combination is what the classifier is looking for, because it is exactly what an automated account looks like. Real people scroll for a while and post later. Spam accounts post within minutes of signup.
The reason this matters more than it sounds: suppression is silent. There is no strike and no notification. You just get a fraction of the reach and have no way to tell whether the content or the account was the problem.
Ten minutes a day, by a person, on the phone.
Five to seven days before anything publishes. It cannot be done from a scheduler, because the point is the behaviour, not the result.
- Watch 10 to 15 reels to completion in the target niche. Completion is the heaviest signal, so watching fully matters more than watching many.
- Like 5 to 10 posts from those accounts.
- Leave 1 or 2 real comments. Full sentences with an opinion. Empty comments actively hurt, so two good ones beat twenty lazy ones.
- Follow 2 or 3 accounts, spread across the week rather than in one burst.
- Share one reel to a friend or to your own story.
Vary the time of day. An identical block at the same minute every evening is itself a pattern.
- Complete the profile first. Picture, bio, name. A finished profile before the first post reads as a real account.
- Switch to a Professional or Creator account. Insights is where outliers are found later, and it does not exist on a personal account.
- New accounts should sit at 5 to 25 actions per hour. The higher numbers quoted in most guides apply to aged accounts only.
Do not buy followers or engagement, and do not join comment pods. Coordinated engagement is now the leading trigger, ahead of hashtags and bots. Tight repetitive cycles of accounts engaging with each other get flagged as inauthentic amplification. It is worse than doing nothing at all.
Slow on purpose, with a stopping rule.
One reel a day for five days, then two for five, then three. The last phase runs knowingly above the usual 4 to 7 per week guidance, so it carries an explicit rule for when to back off.
5 per post. The old 20 to 30 tag habit is now itself a spam signal. Current guidance is 3 to 5 highly relevant tags.
Buffer, on automatic publish. It is an official Meta partner, so it is not the unauthorised automation that gets accounts flagged.
Post times offset so they never publish in the same minute, and hook order deranged so the same line never runs on both on one day.
We are not making a hundred good reels. We are finding the three outliers.
Most of the hundred will do unremarkably. That is the expected outcome, not a failure. The entire value of volume is that it gives enough samples for a real outlier to be visible above the noise.
Which is why the posting order is neither random nor simply best-first. The first 15 are the strongest, because a new account is calibrated by its earliest posts. The remaining 85 interleave strong, middling and weak, so an outlier is measured against reels posted into the same week rather than against a different month's audience.
- 3x the trailing 10-reel median on any single signal.
- Non-follower reach above 50%, which means the algorithm pushed it rather than your followers finding it.
- Completion above 70%.
- Beats the recent average on 2 of 3: completion, saves per view, shares per view.
If the strongest reels all went out first, a good week and a genuine outlier would look identical in the numbers. Spreading strong, middling and weak through the same weeks is what makes the comparison mean anything.
Small paid budget on the exact reel, unchanged. Do not recut the thing that just worked.
Not the topic in general. If the winner is about juggling four apps, the follow-ups are about juggling four apps.
The equivalent reel on the other character, to separate whether it was the hook or the account.
Instagram has a native feature for exactly this. A Trial Reel publishes only to non-followers, stays off your grid, and reports back after 24 hours so you can compare it against your recent average. It is the cleanest outlier test available because it removes your existing audience from the sample entirely. It needs a professional account and 1,000 followers, so it is a phase two tool rather than something available on day one.
Adam Mosseri's position is that “in connected ranking, we do not limit reach”. What creators call a shadowban, Instagram describes as content being non-recommendable. So treat the frequency numbers on this page as practitioner consensus, not documented policy.
Day-zero upload velocity is consistently named across sources. The exact posts-per-week ceiling is much softer, and is why this plan carries a stopping rule based on our own numbers instead of trusting someone else's figure.
Either account can be checked any time under Settings, Account Status, which is Instagram's own diagnostic. Most suppression clears within 14 to 30 days once the triggering behaviour stops.
Everything above, with its source.
- Account warm-up guide 2026, Ssemble
The day-zero upload velocity signal, and the daily consumption actions.
- Instagram limits 2026, Social Champ
Posting caps, the 4 to 7 reels per week guidance, and the 3 to 5 hashtag range.
- Instagram shadowban 2026, ContentStudio
Mosseri on reach, the five confirmed triggers, and the 14 to 30 day recovery window.
- Instagram post limits 2026, DiCloak
Per-hour action bands for new versus aged accounts.
- Instagram algorithm 2026, Dataslayer
The ranking signals Mosseri has confirmed.
- Reels insights guide 2026, Inro
Non-follower reach as the growth signal, and the 50% threshold.
- Trial Reels 2026, brandID
How trial reels work, the 1,000 follower gate, and the graduation rule.
This is one step of a longer pipeline.
Positioning, a synthetic cast built from reference sheets, two stage generation, deterministic assembly and adversarial QC all happen before a single reel reaches a schedule.