Guide
Running a video pipeline: from brief to published
The stage model that keeps agency video production legible — what each stage means, who owns it, and how to spot stuck work before deadlines do.
Cadenus · 11 June 2026 · 8 min read
Ask an agency where a video is, and you learn everything about their operations. "Marc has it, I think, he said something Friday" is one answer. "Internal review since Tuesday, client review tomorrow, publishes Thursday 18:00" is another. Same agency size, same talent, completely different machine.
The difference is a pipeline: a fixed set of stages every video passes through, where each stage has a meaning, an owner and an exit condition. Here's the model that survived contact with real client work.
The stages
1. Brief. The request exists, written down: what, for whom, by when, with what raw material. The exit condition is that an editor could start without asking a single question. Most "production delays" are actually brief delays wearing a costume.
2. Production. Someone owns it and is cutting it. One owner. "The team has it" means nobody has it. If footage is missing or the brief was wrong, the item goes back to Brief visibly, not into a quiet limbo.
3. Internal review. A second set of eyes before the client ever sees it. This stage exists to protect the client review: every round you'd burn on a typo or a wrong logo internally is a round of credibility saved externally. Exit: someone other than the editor says ship it.
4. Client approval. The version is with the client, with a deadline and a single decision expected (see the approval workflow). This is the stage where videos go to die in most agencies, which is why it needs the most structure, not the least.
5. Scheduled. Approved, captioned, in the calendar with a slot and a channel. Distinct from stage 4 because "approved" and "ready to publish" are different facts: thumbnails, captions and channel choices live here.
6. Published. Out in the world, link recorded. If your system publishes natively, this stage completes itself.
Six stages is the sweet spot we landed on. Fewer, and stages stop meaning anything ("In progress" covering edit, review and approval is how work hides). More, and people stop updating them.
The two metrics that matter
You don't need a dashboard of forty numbers. You need two:
- Stage age. How long has each item sat in its current stage? Anything unusually old is either blocked or forgotten, both findable in seconds when stage age is visible, and invisible forever when it isn't. Set a threshold per stage (say, three days in client approval) and treat crossings as alarms, not information.
- Stage history. When an item ships late, where did the time actually go? Teams guess this wrong constantly: production gets blamed for what was a five-day approval stall. Stamping stage transitions turns the argument into data.
Raluca Hook needs to land in the first second.
Marc Recut, sending for review now.
Who owns what
A pipeline without role clarity is a prettier spreadsheet. The minimum viable split:
- Editors own items in Production and answer for stage age there.
- A producer/strategist owns Brief quality, Internal review and the client relationship during Client approval.
- Whoever runs social owns Scheduled through Published.
- The owner watches stage-age alarms across all clients, not every item, just the exceptions.
The handoffs between these owners are exactly the stage transitions. That's the point: the pipeline is the coordination, replacing the standing meeting where everyone reads their list aloud.
Making it stick
Three failure modes account for nearly every abandoned pipeline:
- The pipeline lives apart from the work. If updating a stage means opening a second tool, it won't happen by Thursday. The board has to be where briefs, files, comments and approvals already live.
- Stages don't match reality. If your agency genuinely has a "Filming" stage, add it. A pipeline people mentally translate is a pipeline people abandon. Make the stages yours.
- Exceptions are silent. A board nobody is alerted by is wallpaper. Stuck work must announce itself.
Start with the six stages, rename them to your vocabulary, wire the two metrics, and assign the owners. The first week feels like admin; the third week, someone asks where a video is and the answer takes four seconds, and nobody goes back.
Cadenus runs this model out of the box: configurable pipelines with custom stages and fields, automatic stage history, stuck-work flags, and native publishing that completes items on its own.