Where your numbers come from

Four data types

Every metric says who produced it

The table above answers one question about a figure: what kind of output it is. This section answers a different one: where its value came from. Retensis records both, because knowing that a number was predicted does not tell you whose data it was predicted from.

Every metric Retensis publishes records where it came from: reported by the platform, calculated by Retensis, or read from a screenshot you supplied. The three are kept distinct and never averaged into one figure.

Wherever a calculated metric appears, it carries a note saying how it was worked out. You never have to guess whether a number came from the platform or from us, and you never have to take our arithmetic on trust.

That includes the readings that look wrong until somebody explains them. Your average watch percentage can go above 100%, because a replayed section is counted each time it is watched. It is a real reading of real behaviour, and we would rather explain it than quietly cap it at a rounder number.

Every analysis is checked before it reaches you

Every completed analysis passes automated quality checks before it is saved.

We compare against your real data

When your channel is connected, Retensis scores its own predictions against the retention the platform actually reported for that video. Not against an estimate, and not against a figure read off a screenshot.

It is also why a curve you import from a screenshot stays labelled as one everywhere it appears. It is worth showing you, and it is not the same kind of fact as a measurement the platform returned, so the two are never averaged into a single number.

Responsible use

• Treat predictions as hypotheses.
• Prefer repeated patterns over one isolated score.
• Use measured platform data to validate important decisions.
• Review grounding sources when recency matters.
• Keep brand, legal, medical, financial, and safety judgment with a qualified human.