Measure video performance

What this is about

The number a platform shows first is the number it wants shown, which is rarely the one that answers the client's question. A recruiting film is measured by applications, a product film by qualified enquiries, an internal film by whether people can describe the change afterwards. None of those are visible in a view counter. The useful work is deciding the measurement before the film is published, because retroactive metrics get chosen to match whatever the result happened to be. Add a retention curve and the film starts telling you where the edit lost people, which is the only feedback that improves the next production.

The order matters more than the tools. Skip a step and you usually notice two steps later — by which point the fix costs several times as much.

What runs differently here

Where this differs from the general case:

• Define the success metric in the brief, before anyone knows what the number will be
• Views measure distribution, watch time measures interest, and only a downstream action measures effect
• Read the retention curve as an edit note: the point where viewers leave is a place in the timeline you can name
• Compare a film with itself over time and with your own previous films, not with someone else's screenshot
• Separate paid reach from organic reach before drawing any conclusion about the film itself

Step by step

This order gets you there fastest:

1. Write the goal of the film as a sentence with a verb before production starts: more qualified applications, fewer support calls, faster onboarding. Vague goals cannot be measured afterwards.
2. Choose one primary metric and at most two supporting ones. A dashboard with fifteen numbers guarantees that some of them will look good regardless of what happened.
3. Set up the measurement before publication: tracked links, a landing page, a form field asking how people found you, or an internal survey. Retrofitting tracking loses the first weeks.
4. Record a baseline. Without knowing the level before the film, any number afterwards is just a number, and every interpretation of it is a story someone wanted to tell.
5. Read the retention curve at a fixed point after publication and mark the timecodes where viewers leave. Those drops correspond to specific decisions in the edit.
6. Separate the channel effect from the film effect. The same cut performs differently in a paid feed, an email and an embedded player, and only comparable placements can be compared.
7. Turn findings into concrete production changes: shorter opening, earlier product, a different first frame. A measurement that does not change the next brief was an entertainment expense.
8. Report in the client's language, not the platform's. Say what the film achieved against the agreed goal, what it did not, and what you recommend doing differently next time.

Common pitfalls

What most often goes wrong in practice:

• Choosing the metric after the results are in, so the film always succeeded at something
• Comparing organic reach with a paid campaign and concluding the second film was better
• Reporting platform vanity numbers to a client who asked about applications
• Reading a retention drop as an audience problem instead of as an edit note

With TillyGen

The goal and target audience agreed in TillyGen at briefing stage are the reference the results get measured against — so the evaluation compares the film with what it was commissioned to do, not with the platform's favourite number.

Change one constraint and the consequences travel through the whole plan: affected shots are flagged, the call sheet is regenerated, and nobody keeps working from yesterday's version.