376
Reels analyzed
It happened to us — with receipts. Our most shared reel (30 shares per 1,000 views) gained exactly one follower, while a reel about a £9 ice cream recruited 205 followers per 100K views. We didn't guess why — we built Deep Core, our own AI analysis engine, and fed it every reel we ever posted. 376 reels. 182.3 million views. 53,724 follows, traced one by one. This page is the autopsy report.
Open the file ↓ No course. No $997 webinar. The numbers are free — we sell shovels at the bottom.Deep Core, our in-house AI engine, ingested the native Instagram insights of every reel we posted — views, watch time, likes, shares, saves, profile visits, follows — and correlated all of it. Not someone else's account. Ours. The findings below cost us six months and 182 million views. They cost you a scroll.
Source: native IG insights — not vibesA high save rate tells the algorithm your reel is worth keeping.
It pushes you to non-followers — the only audience that can grow you.
Reach explodes past 75K views.
Follows happen. You control exactly one link in this chain: the save.
| Signal | Correlation with follows per 100K views |
|---|---|
| Save rate | +0.403 — the strongest controllable signal we measured |
| Like rate | +0.394 |
| Engagement | +0.345 |
| Share rate | +0.186 — and against raw follow volume, just +0.112 |
Our narrated mini-docs hit 5–12 saves per 1,000. Memes die under 1. A save is a viewer telling Instagram: this account keeps things I want to keep. That's the whole game.
We declassify one finding like this every other Thursday → see the open cases ↓
| Action | What it really is | Reach | Does it recruit? |
|---|---|---|---|
| ♥ Like | A flick of the thumb | Feeds your existing followers | Weakly |
| ✈ Share | The paper plane — sent to a friend, then they forget you | Maximum — DM sends drive discovery | Almost never (+0.112). Sharing satisfies the urge; it doesn't create loyalty. |
| 🔖 Save | The bookmark — an act of planning to return | Algorithmic trust + cold reach | Yes (+0.403) — the strongest controllable signal |
| ＋ Follow | A promise to come back tomorrow | Nothing — follows are NOT a ranking signal | It IS the conversion — earned only by a serialized promise |
Receipt: our most SHARED reel — 30 paper-planes ✈ per 1,000 views — gained 1 (one) follower. Shares ≠ saves. Different button, different machine. We checked twice.
Is there a precise, verifiable, surprising fact in this video that someone would SAVE — to find again, or to show a friend?
If the answer is no: bury it.
— Buried News · Viral Forensics · Case File No. 376
This single filter killed our meme formats, our celebrity clips and our shock reposts. Combined recruiting power of everything it killed: 0–3 followers per reel.
Every Sunday, one buried story for your brain. Every other Thursday, one growth finding from the lab. Same dataset, next organ.
Check your inbox. If nothing shows up, dig through your spam folder. Fitting.
[CONFIRMED] = stated by Instagram or measured in our own data · [BENCHMARK] = large third-party studies · [DIRECTIONAL] = strong signal, imperfect source
| Metric | 2026 reality | Confidence |
|---|---|---|
| Watch time | #1 ranking signal — viewers decide in ~1.7 seconds | CONFIRMED |
| Skip rate | Healthy: 30–40%. Above 50%, distribution dies | CONFIRMED |
| Hook length | 8–10 words max on screen | CONFIRMED |
| Reels ER (all) | 0.50–0.52% average (35M posts studied) | BENCHMARK |
| ER 10–100K accts | 2–5% | BENCHMARK |
| View → profile visit | 1.5–3.5% | BENCHMARK |
| Profile → follow | 13.5% average · 20%+ elite niche pages | BENCHMARK |
| Healthy growth 100K+ | 2–3% per month | BENCHMARK |
| Organic reach | Down ~18% year over year, for everyone | BENCHMARK |
Reach is shrinking for everyone. Conversion efficiency is the lever left on the table. That's the lever this entire page is about.
Anyone promising you a guaranteed viral formula is lying — including to themselves. What the data DOES prove: every single one of our top 16 reels was the same format (a narrated 15–60s mini-documentary), and reels passing the field test above were dramatically overrepresented in the winning 3%.
You can't pick the winning ticket. You can refuse to buy losing ones. The strategy is volume at the winning format — and letting the distribution do the sorting.
We drill your account the way we drilled ours. You export your native insights yourself — we never ask for your password, ever — and Deep Core runs the full forensics on your last 90 reels: your save rate against the 1.8 threshold, your format autopsy, your three highest-leverage fixes.
One declassified dossier, delivered in 72 hours. Every claim stamped CONFIRMED or DIRECTIONAL, so you know exactly how far our data carries into your niche. If we ever dig together, the $199 is credited to your first month.
We don't guarantee virality. Anyone who does is lying to you. We guarantee the delivery, the process, and the data — and we put that in writing.
We reply within 24 hours with a secure payment link and the insight-export walkthrough. The 72-hour clock starts when your data lands.
4–6 samples accepted per month · first come, first drilled
We script narrated mini-docs built to pass the field test — and at the top tier, we produce them end to end. Every script ships with the coroner's note: why this fact passes, with the correlations to prove it. One operator, a very busy AI, and a hard cap: we take a handful of accounts at a time. The waitlist hears first.
8 scripts a month, you post. Hook ≤10 words, Save Test annotated, serialized CTA placed outside the dead zones.
We script, voice, cut and schedule. Two seats. That's the whole roster.
Waitlist only. We open a few slots at a time and the list hears first.
Waitlist only · We never touch your password — Meta partner access or you post, period
Every number we trust, tagged by confidence.
Declassifies July 2026Instagram makes the call in 1.7 seconds.
Declassifies July 2026Above 50%, your reel is already dead. You have 10 words.
Declassifies August 2026The intent ladder, expanded, with the receipts.
Declassifies August 20261.5–3.5% of viewers check your profile. Most profiles waste them.
Declassifies September 2026Two zones of your video are dead pixels.
Declassifies September 2026Why episodes beat bangers, per Instagram itself.
Declassifies October 2026Test on strangers before you spend your audience.
Declassifies October 2026Archive, never delete.
Declassifies November 2026One new finding every other Thursday, plus the Sunday story. No course upsell at the end. There is no course.
10 of the strangest true stories we've ever dug up — free, in your inbox in 60 seconds. The lab notes follow.
Check your inbox. If nothing shows up, dig through your spam folder. Fitting.