The hierarchy of signals, in order
Through 2024 and into 2026, Meta has been quietly rebalancing what the Reels ranker weights. Based on creator data, ad account performance, and the bits Meta has shared publicly, the current hierarchy looks like this:
- Watch time. Still the dominant signal. A 30-second reel watched fully outperforms a 60-second reel skipped at 20 seconds, every time.
- Shares. Risen sharply in importance. Instagram now treats shares as the strongest indicator of cultural value.
- Saves. Steady — high signal for utility content (recipes, tutorials, how-tos).
- Comments with replies. A comment thread (creator replies, others engage with the reply) is much stronger than a single comment.
- Profile visits triggered by the reel. A new signal that’s been climbing through 2026.
- Likes. De-emphasised. Still counted, but secondary.
The 3-second rule has tightened to 2
The first 2 seconds determine whether you keep the viewer. The hook needs to either visually disrupt or verbally state the payoff in the first frame. “Today I’m going to show you” is dead. “Stop putting conditioner on your roots” works because the value is upfront.
Length is now context-dependent
Through 2023, the ideal reel length was 7–15 seconds. Through 2025, it pushed to 15–30. In 2026, the sweet spot is bimodal: very short (5–10s) for one-idea content, or longer (45–90s) for utility/educational content where viewers commit to watching all the way through. The dead zone is 20–40 seconds — too long for a hook-only reel, too short to deliver real value.
What organic distribution looks like now
Three viewer pools your reel is shown to, in sequence:
- Followers (small initial test). If they engage, it goes wider.
- Adjacent audiences (similar interests, similar behaviour). The expansion zone. Engagement here decides whether you go viral.
- Broad explore feed. Only if shares + saves cross a threshold.
Most reels die in pool 2. The ones that break through usually do so via shares specifically — the algorithm reads shares as social validation that justifies wider distribution.
The format that’s outperforming everything else
Across our client accounts and the wider data we have visibility into, the single highest-performing format in 2026 is what we call the “frame the problem in 5 seconds, deliver the answer in 15” format. Hook with a relatable problem. Pause. Deliver the answer or transformation. Quick CTA. Total length 20–25 seconds.
It hits short enough for full-completion, long enough to register as substantive, and the structure encourages shares because it has clear payoff value.
The trends trap
Trending audio still helps reach, but the gap has narrowed. A reel with original audio that hits on watch time and shares now outperforms a reel using a trend with weaker engagement. This is a shift from 2023, when trending audio was almost a guaranteed reach multiplier.
Use trends when they fit naturally. Don’t bend the content to fit a trend — the algorithm increasingly recognises forced trend-chasing.
What this means for paid
If you’re amplifying organic reels via paid (the boost bank model), the same hierarchy applies. A reel that won organically on watch time + shares will outperform as an ad. A reel that won on likes alone often won’t translate. Pick the right boosts.