Instagram's "Sends Per Reach" Algorithm — How to Design Content for the 2026 Ranking Signal

In a January 2025 explainer video posted to his own account, Adam Mosseri named the three signals that decide where your post lands: watch time, likes per reach, and sends per reach. Four months later, in a May 2025 interview that DNyuz headlined “the paradigm shift from posting in public to sharing in private,” he said the quiet part out loud — on Instagram, way more photos and videos get shared into DMs than into Stories, and way more into Stories than into the feed. The platform now ranks accordingly. This isn't a tweak. Sends per reach is the new top-of-stack signal, weighted 3–5x heavier than likes for any reach outside your follower base. If your content isn't engineered to be forwarded, you don't have a distribution problem — you have a content problem.

What “sends per reach” actually measures

Sends per reach is per-post, not per-account. The numerator is the number of times a single post (Reel, carousel, or static) was forwarded via Instagram DM. The denominator is unique reach — the number of distinct accounts the post was shown to. The metric surfaces inside Instagram Insights under the paper-plane “Shares” icon on any post or Reel, and it's the same number Mosseri keeps telling creators to track first.

Three distinctions worth knowing:

• Sends vs. saves. A save is a private bookmark — a signal that the post has personal utility. Mosseri has described saves as useful but secondary; saves indicate the post matters to one person. A send indicates it matters enough to recommend to another human. That's a different intent class entirely, and the algorithm treats it as such.
• Sends vs. story shares. Resharing a post to your own Story is a public broadcast to people who already follow you. A DM send is a private, one-to-one (or one-to-group) recommendation. Both count toward the “Shares” tile in Insights, but creator-side analysis consistently flags DM sends as the heavier signal because Stories rank on a separate “maintain existing relationships” surface, not on the discovery engine that decides Reels distribution.
• Sends vs. DM-forwards-from-automations. If a creator runs a comment-to-DM funnel, the DM the bot sends to a commenter does not count toward sends per reach on the parent post. Only a human tapping the paper-plane and forwarding the post itself counts.

The weighting: how much heavier is a send than a like?

The most cited number across creator-side teardowns is 3–5x — one DM send is worth three to five likes in the distribution score Instagram uses to decide whether to push your post to non-followers. Some sources push the ratio further (Buffer's and GOSO's 2026 algorithm write-ups quote roughly 1 DM share = 15 likes on the Reels distribution score specifically). The honest read: nobody outside Meta has the exact coefficient, and it isn't fixed — Mosseri himself has said likes matter slightly more for connected reach (i.e., your existing followers' feeds) and sends matter slightly more for unconnected reach (Explore, the Reels tab, suggested feed). 3–5x is the conservative number, 15x is the aggressive read for Reels-into-Explore specifically. Either way, the gap is at least one order of magnitude on the surface that actually matters for growth.

Why sends carry the most weight, mechanically:

1. Friction. A like takes one tap. A send takes four — tap the paper plane, pick a contact, optionally type a message, tap send. The friction itself is the signal. Mosseri has repeatedly framed this as “commitment” — high-friction actions reveal real preference.
2. Network propagation. A like dies inside the algorithm. A send produces a new viewer who didn't see the post organically — that viewer is more likely to watch, send onward, or follow. The algorithm gets free distribution data each time it happens.
3. Social stake. Sending a post to a friend is a micro-reputation bet. You don't forward bad content to people you like. The act selects for genuinely good content in a way that no public engagement metric can.

What the algorithm is actually rewarding

Read together with Mosseri's 2024–2025 originality posts (the April 2026 PetaPixel piece confirmed Instagram extended the original-content distribution penalty platform-wide; accounts that post 10+ unmodified reposts in a 30-day window are excluded from recommendations), the picture is consistent. Instagram isn't rewarding engagement. It's rewarding three layered things: relationship strength (whose DMs is your post propagating through), originality (was the post made by you, or scraped/reposted), and retention (did people watch past the first three seconds and finish the loop).

Sends per reach is the single metric that captures all three at once. A send-worthy post is, by definition, original enough to be worth forwarding, watched enough to reach the share button, and relevant enough to a specific person that the sender is willing to spend social currency.

Seven content patterns that actually get sent

From the post-mortems on reels that broke past 7.5x follower count (the median viral threshold Meta itself cites for reels), seven patterns keep appearing. Each one earns the share for a specific, mechanical reason.

1. Niche-internal in-jokes. The reel a SaaS PM sends to three other SaaS PMs because nobody else would get it. The send happens because identity is at stake — forwarding it says “we are the same kind of person.” Example: account managers forwarding any “client said the brief was clear” reel to their group chat.
2. Hyper-relatable observations with a named recipient.Content that describes a specific human in the viewer's life so accurately that not forwarding it feels rude. The format is usually framed as “send this to your [gym partner / sister / chaotic coworker].” Crunch Gym's “never let your gym bro look bad” reel ran this pattern and crossed a million views; the share rate did the work.
3. Single-screenshot wisdom. One screenshot, one line of text, no edit. Quotes, micro-essays, screenshots of tweets or DMs. These get sent because they're zero-effort to consume on the receiver's end — the sender pays no social tax because nobody has to watch a 47-second video. Static posts in this format regularly outperform Reels on share rate for non-entertainment accounts.
4. POV scenarios. “POV: your boss asks how your weekend was on Monday at 9:02am.” Mirror neurons fire, the viewer recognizes themselves, and the impulse to forward is immediate. POVs work because the share doubles as a confession — the sender doesn't have to say “this is me,” the format says it for them.
5. Polarizing takes (not rage bait). A confident opinion that splits the room — “wedding photographers should never use flash indoors,” “you can't learn React by reading docs.” The send happens because the receiver either agrees violently or disagrees violently, and both reactions are worth texting about. The line between this and rage bait: polarizing takes assert a real position; rage bait is content engineered to be wrong on purpose to farm angry comments. Meta's engagement-bait transparency policy explicitly suppresses the second one.
6. Useful saves that double as sends. Tutorials and checklists that the viewer saves and forwards to one person who'd use it. “Save this and send to whoever's planning a trip to Tokyo.” The dual-action framing pulls both signals out of the same viewer — saves for retention, sends for reach.
7. Fascination loops (“I cannot stop watching this”). Visually arresting, slightly inexplicable content: ASMR, satisfying process videos, optical illusions, oddly mesmerizing handcraft. The share happens because the sender wants someone else to feel the same minor astonishment. Watch time and send rate both spike — these are the reels most likely to break 10x follower-count reach.

What does NOT get sent (and Meta is suppressing)

Mosseri has been explicit, multiple times, that the manipulative engagement formats from 2018–2022 are now actively demoted. Meta's own transparency center lists engagement bait as a content-distribution guideline violation. The specific patterns that cost you reach in 2026:

• “Tag a friend” / “Tag 3 people” CTAs. These produced comment count, never sends. Algorithm now scores comment count as a weak signal at best, and the explicit tag-bait pattern triggers the engagement-bait demotion.
• “Comment YES if you agree” or any prompt that asks for a literal one-word comment. Same demotion bucket.
• “Send this to 5 friends” — numerical send-bait. The format itself trips the bait classifier; situational send prompts (“send to someone who needs this today”) do not.
• Reposted / aggregator content. If 70%+ of the visual frame matches an existing post, the repost-detection system flags it and caps distribution. Aggregator accounts saw 60–80% reach collapse through 2025.
• Generic comment-farming hooks (“controversial opinion below” with no actual opinion). The algorithm reads comment quality, not quantity — short, emoji-only, or repetitive comments don't count.

Where comment-to-DM funnels fit in this picture

A comment-to-DM funnel (someone comments a keyword, your account auto-DMs them a link, gift, or thread) does not directly generate sends per reach on the parent post. The DM thread is between you and the commenter — no third party sees the post forwarded. But the reel that triggered the comment gets a comment (which counts toward watch time because composing a comment takes seconds on the post) and, more importantly, the keyword prompt itself functions as a content design constraint that pushes you toward send-worthy formats. A reel that's good enough to make someone comment a specific word is usually a reel good enough to be forwarded too. Creator Lane runs that comment-to-DM layer for any creator who wants it without the ManyChat watermark — the funnel captures the high-intent commenter while the algorithmic upside comes from the share rate the same reel earns in parallel.

How to measure if your content is “send-worthy”

Open Instagram Insights on any post (the paper-plane icon under Interactions). The aggregated number is what Mosseri wants you to read first. Divide it by reach. That ratio — your share rate — is the single number to optimize.

Benchmark thresholds the creator-side analytics platforms have converged on:

• Under 0.5% — your content is consumed and forgotten. Reach is capped by your existing follower base. This is the median for most creators.
• 0.5–1% — average. The post will perform inside your follower base but won't break out.
• 1–2% — strong. This is the band where reels start picking up unconnected reach from Explore and the Reels feed.
• 3%+ — viral territory. Posts in this band are consistently the ones that hit 7.5–10x follower-count reach. Meta's public number is that reels are reshared 4.5 billion times a day; a 3% share rate on your own post means you're inside the top fraction of that traffic.

A/B testing for it: post the same idea in two formats a week apart (POV vs. talking-head, single screenshot vs. carousel) and compare share rate, not likes. Likes scale with reach; share rate is the leading indicator of future reach. The pattern that wins on share rate is the one to run again.

Closing

Designing for sends per reach is the only content strategy that scales in 2026, because it's the one signal the algorithm rewards heavily and engagement bait can't fake. Build for the four-tap send, not the one-tap like. Comment-to-DM tools like Creator Lane handle the downstream conversion when a sendable reel produces real comments — the upstream content work is yours. Here's the full algorithm breakdown for 2026, and if your numbers cratered earlier this year, these are the nine reasons your reach dropped. Start Creator Lane free — Meta-approved Tech Provider, no watermark, official Graph API.