01) Posting time is niche-dependent (not universal)
Timing
Correlational
Niche-aware
The “best time to post” isn’t global. It shifts by niche, audience geography, and account size.
Open details (what to do, why it works, what to measure)
What to do
- Pick your niche segment first (don’t mix niches when evaluating timing).
- Test 3–4 time windows for 2 weeks and compare distributions, not averages.
Why it works (observed)
Distribution is sensitive to early user availability. Different niches peak at different times,
and the algorithm’s initial test pool reflects who is online then.
What to measure
- First-hour view velocity
- First-hour retention (if available)
- 72-hour views as the stable outcome
PLACEHOLDER — Hour × niche heatmap
Heatmap by niche: probability of surpassing a threshold (e.g., P(>10k views)).
02) Early engagement velocity beats posting hour
Timing
Mixed
Early phase
Two posts at different hours can perform the same—if their first 30–60 minutes show similar signal strength.
Open details
What to do
- Focus on improving the first minute-to-minute signal (hook + pacing + clarity).
- Compare posts by early velocity, not by absolute time.
Why it works (observed)
The early test phase appears to gate expansion. The algorithm reacts to the slope of engagement,
not just the hour of posting.
PLACEHOLDER — Early velocity vs 72h views
Scatter plot: x = engagements/minute (0–30min), y = 72h views.
03) Overposting can reduce distribution coherence
Cadence
Correlational
Account-level
Posting more often doesn’t automatically increase reach. Past a point, it can dilute audience matching and lower consistency.
Open details
What to do
- Choose a cadence you can sustain with consistent niche + format.
- Measure performance distribution over 2–4 weeks before increasing output.
What to measure
- Median views per video (not total)
- Variance of results (spiky vs stable)
- Audience overlap consistency
PLACEHOLDER — Cadence vs stability
Boxplots of views by posting frequency buckets.
04) The hook window is shorter than most creators think
Structure
Tested
Retention
The biggest drop usually happens in the first 1–2 seconds. Your hook must be immediate and unambiguous.
Open details
What to do
- Start with the outcome (not the intro).
- Use the first frame to make the topic obvious (visual + text-on-screen).
Why it works (measured)
Higher first-second retention correlates strongly with longer distribution windows.
Small improvements early can compound dramatically.
PLACEHOLDER — Retention curve (0–5s) before/after
Two curves showing improved first-second retention and downstream lift.
05) Late hooks can work—but mostly for loyal audiences
Structure
Mixed
Account size
“Slow build” formats are viable when the audience already trusts you. For small accounts, it often underperforms.
Open details
What to do
- If you’re small: front-load clarity + payoff signal.
- If you’re established: you can delay payoff, but maintain micro-hooks every 2–3 seconds.
PLACEHOLDER — Retention by account size
Plot retention curves split by small vs large accounts for the same format.
06) Video length has diminishing returns
Structure
Correlational
Format
Longer videos can outperform—but only if retention stays high. Length alone isn’t a growth lever.
Open details
What to measure
- Watch time % (relative to length)
- Drop-off points (seconds)
- Completion rate
PLACEHOLDER — Length vs watch-time%
Scatter plot with a “good zone” band where both length and retention remain high.
07) Comments often predict redistribution more than likes
Engagement
Correlational
Velocity
Likes are lightweight. Comments indicate deeper involvement and can correlate with broader second-wave distribution.
Open details
What to do
- Design “comment triggers”: a question, a choice, or a prediction.
- Reply quickly to early comments to extend the interaction window.
PLACEHOLDER — Comment velocity vs second-wave views
Scatter: early comments/min vs views gained from hour 6–72.
08) Engagement timing can matter more than engagement volume
Engagement
Mixed
Early phase
The same number of likes over 24 hours doesn’t equal the same outcome. Concentration early can correlate with expansion.
Open details
What to measure
- Engagement concentration (0–30min, 0–1h, 0–6h)
- Early slope of views (view velocity)
PLACEHOLDER — Same total engagement, different timing
Two time-series with identical totals but different early concentration and outcomes.
09) Saves & rewatches are silent amplifiers
Engagement
Correlational
Retention
Some signals don’t look flashy but correlate with longer distribution windows.
Open details
What to do
- Create “reference value”: checklists, templates, mini tutorials, before/after comparisons.
- Use tight loops or micro-surprises that naturally trigger rewatches.
PLACEHOLDER — Save rate / rewatch proxy vs long tail
Compare videos with similar likes but different long-tail views (day 3–14).
10) Captions rarely boost reach—but they stabilize targeting
Metadata
Correlational
Classification
Captions help context and audience matching more than they create virality by themselves.
Open details
What to do
- Use clear keywords that match what the viewer just saw.
- Avoid baity captions that misclassify the content.
PLACEHOLDER — Classification stability
Example: distribution of audience segments across posts with consistent vs inconsistent caption keywords.
11) Hashtags help classification, not virality
Metadata
Correlational
Niche
Hashtags tend to help the system “understand” your topic. They don’t replace retention and velocity.
Open details
What to do
- Use a small set of consistent niche tags + a couple of specific tags per post.
- Don’t spam broad tags unrelated to the content.
PLACEHOLDER — Hashtag relevance vs audience match
Show higher retention when hashtags align with content topic.
12) Text-on-screen impacts retention more than captions
Structure
Mixed
Clarity
People decide fast. Text-on-screen makes your topic instantly legible and can reduce early drop-off.
Open details
What to do
- Show “what this is about” on frame 1 (1 line, not a paragraph).
- Keep text stable long enough to read, but change often enough to maintain movement.
PLACEHOLDER — Early retention vs text-on-screen clarity
A/B example: same video with clear vs unclear on-screen text.
13) Algorithmic memory is real (your account has inertia)
Account
Correlational
Long-term
Consistent topics and formats tend to build more predictable distribution. Random pivots can reset audience matching.
Open details
What to do
- Cluster content into 1–3 recurring pillars (not 12 unrelated themes).
- When pivoting, do it gradually—transition formats and topics over several posts.
PLACEHOLDER — Before/after niche pivot
Time series: median views and variance before pivot, during transition, after pivot.
14) Niche switching can reset distribution probability
Account
Mixed
Transition
Switching niche isn’t “bad”, but it often reduces short-term performance until the system rebuilds audience understanding.
Open details
What to measure
- Audience geography shifts
- Retention shifts (topic mismatch creates early drop)
- Consistency of comment themes
PLACEHOLDER — Audience-match entropy
Show increasing “entropy” (spread of audience segments) during a pivot, then stabilization.
15) Inorganic engagement is a risk factor, not a growth hack
Risk
Correlational
Policy
Artificial likes/views/followers can create abnormal ratios and timing patterns that correlate with weaker distribution over time.
Open details (education + risk model)
What this section is (and isn’t)
This is educational risk analysis. It does not provide instructions to bypass platform rules.
The goal is to explain why abnormal patterns can harm performance.
Observed risk patterns
- Ratio incoherence (views, likes, comments don’t “fit” typical distributions)
- Timing anomalies (engagement arrives in unnatural bursts)
- Low-quality follower graphs (inactive-looking accounts)
PLACEHOLDER — “Coherence zone” visualization
Plot typical ranges for like/view and comment/view by niche (percentile bands).
16) Unrealistic ratios can downshift future distribution
Risk
Correlational
Coherence
When metrics look statistically implausible, the system may treat them as lower-quality signals.
Open details
What to do (safe, organic)
- Focus on improving retention first; it stabilizes ratios naturally.
- Drive real comments via questions and clear prompts.
PLACEHOLDER — Ratio outliers vs median reach
Show lower median distribution for ratio outliers compared to “coherent zone”.
17) “Views without depth” tends to decay faster
Risk
Correlational
Long tail
If distribution isn’t backed by retention and meaningful engagement, growth often fails to compound.
Open details
What to measure
- Day 3–14 views (long tail)
- Repeat-view signals
- Comment quality (not just volume)
PLACEHOLDER — Long-tail decay curves
Compare 2 videos with similar day-1 views but different long-tail behavior.
18) Stability beats spikes (especially for small creators)
Strategy
Mixed
Account growth
Consistent performance builds a clearer audience model than occasional viral spikes with no follow-up pattern.
Open details
What to do
- Standardize 1–2 formats that reliably get above your median.
- Iterate on the hook and pacing more than chasing novelty every post.
PLACEHOLDER — Stability vs follower conversion
Show that consistent median views correlate with steadier follower growth than rare spikes.
19) Viral videos don’t “save” weak accounts
Strategy
Correlational
Long-term
A one-off hit doesn’t automatically make the next posts perform. What matters is the follow-up structure and niche clarity.
Open details
What to do
- After a hit, ship 3–5 posts in the same topic cluster and format.
- Turn the viral theme into a series with consistent framing.
PLACEHOLDER — Post-viral regression
Time series: views of posts after a viral event; compare accounts with structured follow-up vs random follow-up.
20) Consistency beats occasional brilliance
Strategy
Correlational
Compounding
Growth compounds when the algorithm can predict who will enjoy your next post—because your content is coherent.
Open details
What to do
- Define 1 niche + 1 signature format + 1 recurring promise.
- Measure retention improvements week over week (not just view spikes).
What to measure
- Median performance (views, retention)
- Variance (stability)
- Follower conversion per 1,000 views
PLACEHOLDER — Compounding model
A simple chart: stable median improvements → larger test pools → higher probability of breakouts.