How RaidCut learning snapshots work

Proposals, not truth

Each snapshot summarizes what seems to be working and what isn't, based on the performance data you've imported. It is deliberately framed as a set of proposals. RaidCut never silently rewrites your strategy; it hands you observations and lets you decide.

Why sample size matters

A snapshot built from three clips is a guess; a snapshot built from fifty is evidence. RaidCut shows the sample size behind each snapshot so you can weigh it. Small samples are noisy — one viral or flopped clip can dominate — so early snapshots should be read lightly.

Everything is "proposed"

Insights appear as proposed changes: a title pattern to try, a posting window to test, a topic that's overperforming. Treat them as experiments. The language is intentionally tentative because audiences shift and correlation isn't causation.

Adopt what you trust

The right workflow is to scan a snapshot, adopt the one or two insights that match your own read of your channel, and leave the rest. You stay in control of your brand and voice; RaidCut just makes the patterns easier to see.

Example insights

• "Clips with a clear stakes hook in the first 2 seconds averaged more views this period."
• "Tuesday evening posts outperformed weekend mornings on TikTok."
• "Shorter titles (under 40 characters) correlated with higher reach."
• "A specific game tag is overperforming relative to its share of uploads."

A caution on low confidence

When a snapshot is marked low confidence, the underlying sample is too small to be reliable. Don't restructure your whole strategy around it — note it, keep posting, and revisit once more clips have real numbers attached.

Snapshots improve as clips link to performance

The single best thing you can do to improve learning is to keep your performance data fresh and to link unmatched videos to their clips. Every link adds signal, and future snapshots draw on a larger, more representative picture of what actually works for you.