View velocity: spotting a video that is about to take off

A total view count is a historical record. By the time a video shows 400,000 views, every decision that could have amplified it has already passed. Velocity is the derivative — and it is the only short-form metric that supports acting in time.

Three levels of the same measurement

1. 1Views — cumulative. Answers "how big is this".
2. 2Velocity — views per hour or per day. Answers "how fast is it moving now".
3. 3Acceleration — the change in velocity between consecutive intervals. Answers "is distribution widening or closing".

Acceleration is the signal. A video doing a steady 5,000 views an hour is performing well but is in a stable distribution loop. A video that goes 800 → 2,400 → 7,100 an hour has been picked up by a broader audience, and the next hours are when a boost, a pin, or a paid amplification has the most leverage.

What sampling frequency actually buys you

Velocity cannot be measured more precisely than you sample. Two readings a day give you daily velocity and nothing finer — a video can rise and peak entirely between them.

There is a cost to frequency — every reading is a request against a platform — which is why sensible tools spend their budget unevenly: fast sampling for young or accelerating videos, slow sampling for a settled back catalogue. A flat "check everything hourly" policy spends most of its capacity re-reading videos that have not changed in a month.

A threshold that survives contact with reality

Absolute thresholds ("alert above 10,000 views/hour") fire constantly for large creators and never for small ones. Use relative thresholds instead:

• Versus the video's own recent rate. Alert when this interval's velocity is ≥3× the median of the previous six intervals.
• Versus the creator's normal. Alert when velocity exceeds the creator's median first-24-hour rate by a factor you set.
• With a floor. Require some minimum absolute movement so a video going from 2 to 12 views an hour does not page anyone.

The combination — relative multiple, plus floor, plus confirmation across two consecutive intervals — cuts false positives dramatically, at the cost of one interval of delay. That trade is almost always worth it.

What to do when the alert fires

Detection is only useful if a decision follows. The realistic options:

• Amplify. Put media behind the post while organic distribution is still expanding. This is the highest-leverage move and the most time-sensitive.
• Extend. Ask the creator for a follow-up or a part two while attention is on the account.
• Reallocate. Move budget from flat performers to the format that is working.
• Prepare operations. If the video points at a product, make sure the landing page, stock and support can absorb the traffic.

If none of those are possible in your organisation, alerting on velocity is decoration. Build the decision path first.

Common ways velocity readings mislead

• Publication-time spikes from the creator's own followers look like acceleration but decay within hours. Confirmation across intervals filters most of them.
• Replays inflate short videos. All three major platforms count replays as views, so a 6-second loop shows higher velocity than a 45-second video with the same audience.
• Missing readings look like a stall. If a scheduled read fails and you divide by the wrong elapsed time, velocity collapses artificially. Always compute against the actual interval between readings, not the intended one.
• Time zones. Feeds have daily rhythms. Compare hour 14 to hour 14, not to hour 03.

The reporting habit that makes this work

Store every reading with its timestamp and compute deltas between consecutive readings. That single discipline gives you velocity, acceleration, lifecycle curves and defensible campaign numbers from the same data — and it is the difference between a dashboard that tells you what happened and one that tells you what to do next.