The “saves matter more than likes now” narrative is half-right and badly misleading
Every Instagram growth Twitter thread in the last two years has repeated some version of “likes don’t matter anymore — saves and shares are what the algorithm cares about.” The framing is technically accurate on per-unit weight and operationally backwards on how the algorithm actually scores posts. Likes still affect reach and rankings in 2026 because they’re the upstream volume signal that triggers everything else. A post that fails to clear the first-hour like threshold never reaches the audience that would have produced the higher-weight saves and shares. Treating likes as obsolete is the most common reason creator strategies underperform their content quality.
SMMNut Stacking Signal Model 2026: Instagram’s ranking model doesn’t choose between engagement types — it stacks them. Each post accumulates a composite score from the weighted sum of all engagement types, with higher-weight signals producing more score per unit but lower-weight signals producing more total volume. The model fires distribution decisions when the composite score crosses tier thresholds; below threshold the post stays inside the follower base, above threshold it unlocks broader distribution. Likes contribute roughly 30–45% of the typical post’s composite score because their volume offsets their per-unit weight. Discounting likes entirely as “lower weight” miscounts how the model actually scores — it’s like ignoring 35% of the income on a P&L because it comes from the lower-margin product line.
The math behind why likes still matter
Consider a typical Instagram post that picks up 100 likes, 8 comments, 4 shares, and 4 saves in the first 90 minutes. Applying the standard weight model — likes 1x, comments 2.5x, shares 3x, saves 4x — the composite score breaks down as 100 + 20 + 12 + 16 = 148 units. Likes contribute 100/148 = 67% of the score in this scenario. Even doubling the higher-weight signals to 16 comments, 8 shares, and 8 saves moves the composite to 100 + 40 + 24 + 32 = 196, with likes still contributing 51%. The only way likes drop below half the composite score is on outlier-share content where save and share rates are 4–8x the typical ratio.
The composite score breakdown across content types
| Content type | Likes share of composite | Saves share | Shares share |
|---|---|---|---|
| Entertainment Reel | 55–65% | 10–15% | 15–20% |
| Educational carousel | 35–45% | 30–40% | 10–15% |
| Lifestyle feed post | 60–70% | 8–12% | 5–10% |
| B2B / professional | 40–50% | 25–35% | 12–18% |
| News / commentary | 45–55% | 8–12% | 25–35% |
Across every content category, likes contribute more to the composite score than any other single signal. The narrative that saves dominate is true only for outlier educational and B2B content where the save-rate runs unusually high — and even there, likes still contribute the largest share.
Why the narrative shifted (and why it overshot)
Around 2022, Instagram increased the per-unit weight of saves and shares in its ranking model. Creator strategists picked up the change and amplified it into “saves are the new likes.” The amplification was directionally useful — creators who weren’t optimising for saves at all were under-indexed — but it overshot the reality. The actual change was a weight shift, not a substitution. Likes didn’t lose their role; they kept their role and got a slightly smaller share of the composite. The strategists who took the narrative literally and stopped optimising for likes ended up under-indexing the largest single component of the score.
How the algorithm uses likes specifically
Beyond the composite score, likes play three distinct algorithmic roles that no other signal fully substitutes for.
Velocity threshold trigger
The broader-distribution decision in the first 60–90 minutes uses engagement velocity (engagement events per minute) as one of its two primary inputs. Because likes happen at much higher per-minute rates than saves or shares, they’re the dominant velocity contributor. A post can have great save-to-impression ratios and still fail the velocity bar because saves don’t accumulate fast enough to clear it.
Social proof for organic visitors
Visible like counts directly affect whether new viewers stay and interact with a post. Saves are invisible to non-owners; shares are partially aggregated. Likes are the social proof signal visitors actually see. Posts with high like counts produce 30–60% better stay rates from cold visitors compared to similar posts with low like counts and high save counts.
Profile-level engagement floor for monetization
Brand deal scouts and platform monetization features use like-density as the readable engagement floor. The SMMNut deep-dive on likes and monetization walks through why the like-to-follower ratio dominates scout triage decisions — likes are the only signal visible at scout-scan speed.
SMMNut Signal Stacking Reality Check: The three highest-leverage engagement signals on Instagram in 2026 are likes (highest volume), watch-completion (highest per-unit weight for Reels), and saves (highest per-unit weight for static content). The optimal strategy stacks all three — content that’s interesting enough to like in the first few seconds, watchable enough to finish, and useful enough to save. Strategies that optimise only for the “high-weight” signals end up producing content that’s worth saving but not interesting enough to like — which means it doesn’t pick up the velocity to be surfaced to anyone in the first place. Volume and weight are complements; treating them as substitutes is the most common analytic mistake in Instagram strategy.
For a low-risk way to benchmark first-hour response on a public post, use SMMNut’s public-post likes check before scaling any paid plan.
What the “likes don’t matter” advice actually fixed
The narrative wasn’t wrong — it was responding to a real problem. Creators in the early 2020s over-optimised for likes by chasing easily-likeable but un-saveable content — generic motivational quotes, low-effort meme reposts, vanity selfies. These posts produced like counts but no saves, no shares, no DMs, no profile visits. The algorithm correctly recognised this pattern and penalised single-signal optimisation. The “saves matter” framing was a corrective — diversify your signal profile so you’re not producing the like-only content the algorithm had learned to discount.
Where the framing overshot was treating saves as a replacement rather than an addition. The right interpretation is “make content that picks up likes AND saves” — not “make content that picks up saves at the expense of likes.” Creators who internalised the wrong interpretation produced save-strong but like-weak content that never crossed the velocity threshold in the first place.
SMMNut Velocity-Volume Reality 2026: Likes happen at 5-20x the per-minute rate of saves and shares because the action friction is lower — one tap versus a deliberate decision. This volume advantage means likes dominate the velocity threshold trigger across every content category, even on save-heavy educational content where saves carry more per-unit weight. A post that picks up 60 saves but only 30 likes in the first hour will fail the velocity threshold despite scoring well on the heavier signal; a post with 200 likes and 8 saves clears the threshold and unlocks the broader-distribution cascade. Volume drives velocity; velocity drives distribution; distribution drives every subsequent signal. The order matters and likes sit at the top of it.
The right way to use this in 2026 strategy
The actionable takeaway is to design content for stacked signals, not isolated signals. A post that delivers a quick-likeable hook, a save-worthy value layer, and a share-worthy specific claim or quote produces composite scores 2–4x higher than posts that optimise only for one signal. The SMMNut comparison of likes vs saves vs shares covers the specific content patterns that stack the signal mix.
The three checkpoints in a stacking-friendly post
- First 2 seconds: A hook that’s interesting enough to like before viewing further
- Middle 60–80%: Value content (specific, actionable, demonstrably useful) that triggers saves
- Final 10–20%: A specific claim, statistic, or framing that’s quotable and triggers shares
Bottom line — likes still affect reach because the algorithm hasn’t replaced them
Instagram’s ranking model in 2026 weights saves and shares more heavily per unit than it did in 2020, but likes remain the largest single contributor to the composite score on most content types. The volume advantage of likes — they happen at 5–20x the rate of saves and shares — means they dominate the velocity trigger, the social proof signal, and the monetization-readable engagement floor. Discounting them in strategy is the analytic equivalent of ignoring the largest line item on the P&L because its margin is lower. The SMMNut Instagram likes guide and the underlying Stacking Signal Model are built around this reality.
For the foundational explainer behind all of this, see what Instagram likes are and why they still matter in 2026 — it covers the full lifecycle of a like through the ranking system.