Summary of "Спасибо, Адам! Взлом алгоритмов от директора инстаграм | Читаем статьи МЕТА и учимся снимать Reels"

Main ideas / lessons

Instagram Reels distribution is driven by “retention” and early performance metrics, not just initial reach.
Your video is evaluated in phases:
1. Automatic compliance checks (AI + community rules)
2. Analysis of visuals/audio/text (computer vision)
3. An early “test” distribution to a small set of relevant viewers (loyal + interested non-followers)
4. Scaling or stopping based on engagement and watch behavior
Key mindset: You don’t “get good results and then fly”—you “fly because the results are good.” In other words, metrics determine distribution.
Content wins when it feels understandable and attention-holding, especially by:
- subtitles (and ideally on-screen titles),
- a strong hook in the first 3 seconds (avoid quick skips),
- editing that increases clarity and momentum (e.g., dynamics; don’t bore viewers).
Algorithms keep testing and retraining with new data over time, meaning posts can get a “second wind” later, and trends can cause clusters of similarly performing content.
“Freshness” matters alongside quality: experimenting with multiple formats increases your odds of finding what works now, as audience interests shift.
To protect your author rating, avoid “red flags” that harm distribution.

Methodology / structured workflow (as presented)

A) How Instagram evaluates and distributes Reels (step-by-step)

Initial screening (immediately after upload)
- AI checks compliance with community guidelines.
- If violations are detected, the video is blocked.
Content analysis (still early)
- Computer vision and audio/text analysis evaluate:
  - Visuals: objects/scenes/actions
  - On-screen text: captions/subtitles/words
  - Audio: track quality and content
  - Geotags
  - Theme/genre classification
Stage 2: shown first to subscribers (while systems “load”)
- The system prioritizes early interactions and watch behavior:
  - people who actively interact (likes/comments/saves),
  - people who recently engaged with your stories/posts,
  - people who often watch Reels on your topic.
- What is measured:
  - viewing duration (watch time / watch to the end),
  - reactions shortly after posting (likes/comments/saves),
  - reposts / sending in DMs,
  - “negative” early behavior: skipping within the first ~3 seconds.
Stage 3: test impressions for cold audiences
- If loyal audience results look promising, Instagram tests the reel with non-followers who might be interested.
- Instagram builds a “digital fingerprint” of your video and compares it to successful Reels in the niche.
- If it performs well for this new group, it moves to bigger distribution.
Stage 4: scaling vs stopping (the “hit” decision)
- If key metrics are weak:
  - impressions stop after the first wave; it may remain on your profile but won’t be recommended.
- If metrics are strong:
  - the reel moves into recommendation placements:
    - Reels feed,
    - main feed near posts,
    - “Interesting” tab/search-like discovery (and uses covers as a click signal).
Ongoing learning
- Models are retrained hourly; the reel can gain additional views on day 2–3 if key metrics remain strong.
- Super-brilliant reels can also be distributed geographically and tested in different countries.

B) Practical actions to improve performance (implied recommendations)

Optimize for the first 3 seconds
- Use strong hooks so people don’t swipe/skip immediately.
Use subtitles (mandatory) + ideally titles
- Many viewers watch without sound.
- Subtitles improve comprehension and act as quality signals for algorithm understanding.
Boost “retention” and engagement signals
- Aim for longer watch time, sound-on behavior, and interactions (save/share/send).
Use covers designed for clarity and clicks
- Covers are treated as an extra quality signal.
Improve technical quality
- Sharp visuals, stabilization, lighting; better sound via microphones or post-processing.
Edit for dynamics and clarity
- Keep momentum; add subtitles/visualizations to help algorithms interpret context and keywords.
Experimentation + format testing
- Learn base Reels formats and test at least 6 of 12 in practice.
- What worked last month may stop working later—keep iterating.
Use Direct + Stories to revive/prepare distribution
- If reviving an older “blog,” the speaker recommends strengthening communication in Direct/Stories to increase subscriber interactions.

C) “Red flags” (what algorithms don’t forgive) — detailed list

Watermarks and platform branding
- TikTok/YouTube icons, captcha intro at the end, logos of other platforms.
Skips / fast swiping (within ~3 seconds)
- Interpreted as the algorithm’s signal that content is uninteresting.
- You’re competing for attention with what your target audience sees in recommendations—not only direct niche competitors.
Re-uploading the same video repeatedly
- Don’t delete and re-upload over and over.
- Principle stated: “post and forget.”
- Deleting and re-uploading can lead to critically lower ratings; algorithms may “remember” performance.
- Re-uploading after a long period (months/years) is described as normal.
User “negative recommendations”
- If users signal disinterest, it lowers author rating.
Behavior that causes unsubscribes and engagement/loyalty drops
- Example pattern: inconsistent posting/vanishing and returning, which reduces trust from both algorithms and people.
No fundamental content base / no strategy
- “Random actions” without a step-by-step plan create incomprehensibility for both the platform and the audience.
- Lack of strategy is claimed to prevent consistent results.

“Author rating” (how it’s framed)

Author rating is presented as a hidden health metric (medals/score) that Instagram uses.
It increases when:
- viewers watch longer,
- sound is turned on (stated as a valuable metric),
- negative behavior is avoided.
It influences future performance:
- a successful reel can “lift” previous and next content,
- the speaker gives an example of increasing the “minimum view bar” from hundreds to a higher baseline.

The overall idea: your distribution depends heavily on how early viewers behave (watch time, sound-on, skips avoided), and your “author rating” reflects that health over time.

Who is featured / speakers and sources

Speakers / on-camera or narrated voices

Adam Mosari
- Described as a former Facebook executive and current CEO of Instagram.
- Quoted as the source of claims about audition/testing and distribution behavior.
Alexey Nekrasov
- Main speaker.
- Shares his growth experience, teaches the algorithm breakdown, and promotes the “BD club.”
Vika (Alexey Nekrasov’s wife)
- Mentioned as a blogger who creates concepts and helps produce the video.
- Described as more “complex” in content thinking, and responsible for translating and structuring materials for the content base.

External sources / referenced organizations

META / Instagram
- Referenced in the context of banned resources in Russia (speaker claims Instagram is part of META, recognized as extremist in the Russian Federation).
YouTube / TikTok
- Mentioned indirectly as examples of watermark content that harms Reels performance.