Retention Prediction
A Content Spy retention tool that estimates hold strength for content you have not published yet — concept, opening hook, optional pacing and payoff — then returns a 0–100 Retention Score. Six credits. Compare mode included.
Overview
Retention Prediction sits in Content Spy under Retention & Engagement. It is built for unpublished plans: you describe the concept and the opening (and optionally pacing, payoff, and channel context). Open Go Viral → Content Spy (creator account, desktop), or open the dedicated form. Each run costs 6 AI credits. Compare mode supports 2–3 videos: cost = number of videos × 6.
How to run
- 1Sign in as a creator. Open Go Viral → Content Spy, then Retention Prediction — or open the dedicated pre-publish form.
- 2Required: content type (YouTube Shorts vs long-form), content concept, and opening hook. Optional: pacing / structure, promised payoff, and a short creator-history note.
- 3From Content Spy with a loaded public or unlisted video, the tool can prefill concept and hook from title and description. That is a shortcut, not a transcript audit. Private videos are not supported.
- 4Run it (6 credits). Credits deduct only after a successful result. Read the overall band, the five weighted sub-scores, Detected Issues, then Optimization Recommendations.
What it uses
Unlike Retention Analysis, this engine does not pull a transcript, duration class, or thumbnail. It scores the plan you type.
- Content type: YouTube Shorts (short-form) or long-form video. This switches the pacing bar inside the model, not a duration cutoff.
- Content concept — what the video is, and the promise it makes.
- Opening hook — the planned first moments. The model quotes short phrases from this field when it critiques the open.
Optional — they change the evidence, not a hard cap
- Pacing / structure: how beats arrive. If omitted, the model still scores pacing from the concept and hook — with less evidence.
- Promised payoff / conclusion: what the viewer is owed at the end. If omitted, payoff alignment is inferred from the concept alone.
- Creator / channel context: audience fit. If omitted, the run still completes; there is no forced 50 on a hidden “history” pillar.
Scoring framework
Five integer sub-scores (0–100) combine into the overall Retention Score with fixed weights. Opening is worth more than everything else. These are not the five equal 20% pillars on Retention Analysis.
Opening friction
30%- 100 = lowest friction / strongest open. 0 = a fatal first stretch.
- High-risk: logo sting, slow context, “hey guys welcome back” with no stake.
- Low-friction: immediate tension, a bold claim, a pattern interrupt, concrete stakes.
Curiosity sustainment
20%- Open loop planted early and kept alive.
- Unresolved question vs a structure that answers everything in the first minute.
Pacing stability
25%- Shorts: new stimulus every 2–4 seconds.
- Long-form: re-hooks every 60–90 seconds. Dead air, energy collapse, and overload all cut this score.
Pattern density
15%- Micro-surprises, re-hooks, narrative resets, twists.
- Penalty: a flat tour with no variation — a settings walkthrough with no new question.
Payoff alignment
10%- Does the described ending pay the opening promise before patience runs out?
- Penalty: misleading promise, or a payoff parked after a long slog.
What you get
The result is Markdown, formatted for the Content Spy results panel and the dedicated form.
- Retention Score: XX / 100 plus a band — STRONG (90–100), HEALTHY (70–89), FRAGILE (50–69), HIGH DROP-OFF RISK (below 50).
- Sub-scores: Opening friction (30%), Curiosity (20%), Pacing (25%), Pattern density (15%), Payoff alignment (10%) — each 0–100 with a one-line why.
- Opening Analysis, Curiosity Mechanics, Pacing Assessment, and Pattern Density — short diagnostic sections.
- Detected Issues — each item tagged (critical), (moderate), or (minor).
- Optimization Recommendations — numbered, ranked by impact.
Case study
Scenario
A productivity creator has not filmed yet. The plan is an 8-minute “I rebuilt my entire workflow” video. The opening names the cost of the old stack. The middle is described as a settings tour. They skip the pacing field to see how thin structure is scored.
Context (available inputs)
- Content type: long-form
- Concept + opening hook (required)
- Pacing / structure: not provided — scored from concept and hook only
- Promised payoff: “the new stack on screen”
- Cost: 6 AI credits
Tool result
Score breakdown
0–100
- Opening86/100
- Curiosity78/100
- Pacing58/100
- Pattern52/100
- Payoff64/100
Retention Score
70 / 100
HEALTHY — the open earns the next ten seconds, but the settings tour has no re-hook plan. Weighted average 70 (opening 30% + curiosity 20% + pacing 25% + pattern 15% + payoff 10%).
Detected issues
- Settings tour with no new question (critical).
- Pacing field left blank — rhythm inferred, not described (moderate).
- Payoff does not return to the cost named in the open (moderate).
Recommendations
- Keep one surprising default. Move the rest of the tour to a pinned comment.
- Plant a re-hook every 60–90 seconds: next tool, failed attempt, or time saved.
- Close by returning to the cost from the open.
Takeaways
- This scores a plan you type — not Studio retention on a live video.
- Opening is 30% of the overall. A strong hook cannot fully rescue a middle with no re-hooks.
- Use Retention Analysis on a published (or unlisted) video when you have a transcript. Use Retention Strength in Content Lab when the draft lives in the Lab, not Spy.
Frequently Asked Questions
Is this my YouTube retention graph?
No. YouTube Analytics shows what happened. Retention Prediction scores whether the described plan is built to hold. It does not invent drop-off points, average view duration, or a completion percentage.
How is this different from Retention Analysis?
Analysis scores a published video from the library (transcript up to 8,000 characters, five equal 20% pillars, Short-form ≤ 180 seconds). Prediction scores a pre-publish form with weighted pillars (opening 30%, pacing 25%, curiosity 20%, pattern 15%, payoff 10%). Lab vs live: Prediction is the plan. Analysis is the cut.
How is this different from Retention Strength in Content Lab?
Same job, different product. Retention Strength scores a draft inside Content Lab. Retention Prediction is the Content Spy form: Shorts vs long-form, opening hook, optional pacing and payoff. Spy also has Compare mode.
What counts as Shorts here?
The content type you select: YouTube Shorts vs long-form. That is not Retention Analysis’s 180-second line. Pick Shorts if you are planning a Short. Pick long-form if you are planning a long video.
What if I skip pacing or payoff?
The run still completes. There is no hard cap like CTR Prediction’s missing-thumbnail 55. The model scores those pillars from the concept and hook with less evidence — and will usually say so in Detected Issues.
Can I compare two plans?
Yes. Compare mode is on. Same cost: 6 credits per video in the run.
How much does one run cost?
6 AI credits per prediction. Compare mode multiplies by the number of videos (2–3).
Next Steps
Score hold design on a published or unlisted video (transcript + equal five pillars).
The Content Lab version of retention scoring — use it on a draft inside the Lab.
See all 10 competitive intelligence tools, including Compare mode.
How credits work, free balance, and pack pricing.