How to Use the YouTube Views Predictor (And What the Score Actually Means)
Most creators look at the 90-day number and walk away. The real value is in the score, the curve shape, and what they tell you about your video before it is too late to act.
The Views Predictor is not a crystal ball. It is a diagnostic tool. The forecast tells you less than the score does — and the score is almost entirely driven by two or three inputs you control.
What the predictor is actually doing
The engine works in five stages. First, it converts your current views and engagement into an algorithm signal score. Second, it calibrates a peak daily view rate from your existing data. Third, it applies a view-decay curve — different for Long Form and Shorts, different at each score level — to project 90 days forward. Fourth, it applies an audience pool cap based on your language market and niche. Fifth, it outputs the curve you see on screen.
The part that surprises most people: the algorithm score, not the view count, determines the shape of the curve. A video with 500 views and a score of 2.4 will generate a more aggressive growth projection than a video with 50,000 views and a score of 0.6. The current view count calibrates the baseline. The score determines whether the curve rises, plateaus, or decays.
The key insight
The predictor does not forecast views. It forecasts the algorithmic trajectory implied by your current performance signals. The difference matters: a trajectory can be changed. A forecast cannot.
The algorithm score — what goes into it
The score is calculated differently for Long Form and Shorts, because the signals YouTube uses for each format are genuinely different. For Long Form, the engine combines CTR, retention, and weighted engagement (likes ×1, comments ×3, shares ×8, subs gained ×5) with an authority multiplier derived from your channel's average view performance. For Shorts, swipe rate replaces CTR and completion rate replaces retention, with additional viral triggers applied when completion exceeds 85% or 95%.
The authority multiplier is the part creators most often miss. It adjusts the score based on how your current video is performing relative to your channel average. A video at twice your usual view count increases the multiplier; a video at half your average decreases it. This means the same CTR and retention numbers produce different scores for a 500-subscriber channel and a 500,000-subscriber channel — which is correct, because YouTube's algorithm treats them differently.
Shares carry the highest engagement weight by a significant margin. A video that generates shares at even 0.5% of view count is sending a distribution signal that comments and likes cannot replicate at the same rate. If your engagement score is low and you can improve one metric, shares are the highest-leverage target — although they are also the hardest to generate intentionally.
Score thresholds and what they predict
The score maps onto four curve families, each with a distinct shape. Below 0.8, the video follows a rapid decay pattern: a strong day-one push from subscribers and notifications, followed by near-complete drop-off by day five. The 90-day number will be close to the 7-day number because almost all traffic arrives in the first week.
Between 0.8 and 1.5, the engine blends toward an average-performance curve: a sharper day-two peak, a more gradual shoulder decay, and meaningful traffic persisting through the 30-day window. This is the range where most videos land. The 30-day number will be meaningfully higher than the 7-day number.
Between 1.5 and 2.2, the curve shifts to a strong-push pattern: peak on day two or three, a secondary plateau around days 12–15, and a slower tail. The 90-day number can be two to three times the 30-day number for strong performers in this range, because the algorithm continues pushing the video into new audience segments after the initial burst.
Above 2.2 — the viral pattern
Above 2.2, the model applies the viral curve: a delayed peak around day four, a strong secondary peak at days 13–16, and a slow decay from there. If your score is in this range, the most important thing you can do is engage with early comments immediately — the algorithm reads creator engagement in the first 48 hours as a quality signal that can extend the initial push.
Long Form vs Shorts — not the same calculation
The two formats use completely different input logic and completely different curve families. For Long Form, CTR and retention are the primary score drivers. A 6% CTR and 50% retention is the threshold above which the engine expects meaningful algorithmic distribution — below it, the curve defaults toward the decay pattern regardless of engagement counts.
For Shorts, swipe rate is the CTR equivalent, but it operates in reverse: a low swipe rate is good. Below 30% is the green zone; above 50% triggers a weak-signal curve regardless of other metrics. Completion rate has a non-linear effect — crossing 85% and 95% activates multipliers (2.5× and 3×, with an additional 5× above 105% for loop-inducing content) that can dramatically change the projected trajectory.
One practical consequence: if you produce both formats, do not compare their scores directly. A Shorts score of 1.8 and a Long Form score of 1.8 are not equivalent predictions — they are scores within different curve families calibrated to different signals. The absolute number only has meaning relative to the format it was calculated in.
The audience pool cap — why projections get bounded
Every prediction is bounded by a theoretical maximum derived from your language market and niche. The engine calculates a 90-day niche audience pool (monthly views in that language × niche audience share × 3) and then applies a reach fraction based on the algorithm score. The reach fraction scales from near-zero at low scores to roughly 0.1% of the pool at elite scores — a ceiling that reflects how even the most viral videos reach a fraction of their theoretical maximum audience.
You will see this cap applied when the raw projection exceeds the calculated maximum. When it triggers, the daily view curve is scaled down proportionally, which is why the curve shape is preserved but the absolute numbers are lower than the uncapped projection would suggest. The cap indicator in the results panel tells you when this has happened.
What the cap is telling you
A capped projection does not mean the model thinks your video is bad. It means the raw signal score is high enough that an uncapped projection would exceed what the niche audience pool supports. The cap is a reality check on exponential growth assumptions, not a penalty on good content.
The practical implication: niche and language selection matter for the ceiling of what is achievable, not just for revenue. A video in a small niche in a small language market with a very high score may be capped lower than a video in a large niche in a large market with a moderate score. If you consistently see capped projections, it may be worth evaluating whether your niche selection is limiting your growth potential.
How to read the 90-day curve
The curve has two segments: the past line (green, solid) and the future line (purple, dashed), divided by the 'Today' marker at your video age. The past segment shows the trajectory implied by your current data — not necessarily the exact day-by-day history, but the curve shape that your current cumulative view count and score project backward. If the past segment looks unrealistic given what you know happened, the most likely explanation is that the video age input is off.
The future segment is the projection. Three things to look for: the peak position (when does the curve peak, and is that in the past or future?), the shoulder shape (how steeply does it decay after the peak?), and the difference between the 30-day and 90-day numbers. A large gap between 30 and 90 days indicates a video with long-tail potential — likely driven by search or browse surface traffic. A small gap indicates a front-loaded video where most value has already arrived or will arrive in the first month.
The two modes
Switch between Cumulative and Daily view modes. Cumulative shows total views over time — useful for understanding the overall trajectory. Daily shows the view rate per day — useful for identifying the peak day, the decay rate, and whether there is a secondary push expected in the second or third week. For videos with a score above 1.5, the daily view mode often reveals a secondary peak that the cumulative view hides.
The most common input mistakes
Video age is the input that produces the most distorted results when wrong. If you enter '3' for a video that is actually 12 days old, the engine calibrates the peak to day three of a video with your current view count — which produces a dramatically different curve than the same data at day twelve. Enter the actual number of days since upload, even if it is uncomfortable because the video has fewer views than you expected by now.
Entering zero or leaving out engagement metrics collapses the engagement score component of the algorithm calculation. The model can still project from CTR and retention alone, but the result will systematically understate the trajectory for videos with strong engagement relative to views. This is especially relevant for Shorts, where engagement ratios (likes per view, comments per view) tend to be higher than long form and carry real weight in the score.
Average previous views
This is the input most people skip. It controls the authority multiplier — the component that adjusts the score based on how your current video is performing relative to your channel baseline. Leaving it blank defaults the multiplier to 1.0, which is correct for a brand-new channel but systematically overstates the score for established channels whose current video is underperforming their average, and understates it for channels whose current video is an outlier above their baseline.
The subscribers field is used to calculate a notification floor — a minimum peak daily view rate derived from the estimated fraction of subscribers who see and click notifications. It does not drive the score, but it ensures the projection does not fall below what your subscriber base alone could plausibly deliver. For large channels with small videos, this floor can be the binding constraint. For small channels with strong signals, it rarely matters.
What the predictor cannot do
It cannot account for external events. A news cycle, a celebrity mention, a community post from a large creator, or a Reddit thread picking up your video can cause view spikes that are completely invisible to any model built on your video's own metrics. When these happen, the actual curve will diverge from the projection — not because the model is wrong, but because an external variable it had no data on was introduced.
It cannot predict YouTube's own distribution decisions. The model estimates the trajectory implied by your performance signals, but YouTube's algorithm can choose to push or withhold a video based on internal factors — A/B tests, policy considerations, advertiser suitability signals, or simply competing inventory — that no external tool can observe. Think of the projection as the expected outcome absent surprises, not the guaranteed outcome.
It cannot substitute for your YouTube Studio data. The predictor is most useful in the first three to seven days of a video's life, when you have real engagement data but not yet enough historical performance to see where the curve is heading. After 30 days, your actual Studio analytics tell you more than any projection can. Use the predictor to inform early decisions — whether to promote, whether to iterate on the hook, whether to revisit the thumbnail — not to replace the data you already have.
The score tells you where the algorithm currently sees your video. The curve tells you where it is likely to go. What you do in the first 48 hours determines which curve you end up on.
How to act on the results
If the score is below 0.8 and the video is less than three days old, the most high-leverage intervention is the thumbnail. CTR changes in the first 48 hours can still move the score enough to shift the curve family — which is the difference between a decay pattern and an average-performance pattern. Retention and engagement are harder to change retroactively; packaging is not.
If the score is between 0.8 and 1.5 and the video has been live for four to seven days, the projection is showing average reach with a realistic long-tail. The most useful action here is cross-promotion — pushing existing traffic sources (community post, other videos, external links) to maintain enough velocity that the algorithm continues sampling the video into new audiences beyond the first week.
If the score is above 1.5, the model expects a secondary push around days 12–15. The highest-leverage action at this stage is engagement with early comments — which signals creator responsiveness to the algorithm — and avoiding any rapid changes to the video's metadata during the initial push window, as these can interrupt the distribution cycle before it has fully run.
