Key takeaways
- We pulled sixty uploads from each of five channels we measure ourselves against, three hundred videos for about nineteen cents, and scored every one against its own channel's median rather than against the others.
- That data corrected a rule we had followed for a year. The mistake is not naming your tool in the title. It is making the tool the payoff.
- A brief now runs before any script gets written, and it asks three questions it is allowed to answer no to. Being allowed to say no is the only part that saves a day.
- The beat we had been skipping entirely is the one that says what the thing is worth, what an agency charges for it, and who you would sell it to.
- Five channels are a better hypothesis, not proof. Our own sample is eleven videos deep, so the plan is to run it and check it on Friday.
The expensive part of making videos is not the filming. It is the videos you should not have filmed. An idea sounds strong in the morning, you block out an afternoon, and somewhere around the third take you realise nobody outside your own head was looking for this.
So we built a step that runs before the camera does, and what makes it worth having is that it is permitted to fail the idea.
What three hundred videos actually said
Before writing any of it we bought data. Sixty long form uploads from each of five channels in our lane, scraped through the same runner we use for prospect research, about nineteen cents for the lot. Every video was then divided by its own channel's median, because a big channel's flop beats a small channel's hit and raw view counts across channels tell you nothing.
The first finding was not about titles at all. The channel closest to us in size and subject, just over forty thousand subscribers in the same software niche, posts nearly eight times a week and holds a median under two thousand views. Another posts less than once a week, keeps its videos under ten minutes, and holds one past a million. Cadence is not the lever most people assume.
The rule we had been following backwards
Our standing title rule was never name the tool. It came from our own numbers, and they looked convincing: the two videos we shipped with a product name in the title pulled nineteen views and four views.
Against three hundred competitor videos that rule does not survive. Naming the tool won on four of the five channels. The strongest shape measured anywhere in the sample was a tool paired with an outcome, which ran at roughly one and nine tenths times its channel median on one account and over twice the median on another. The channel with the cleanest single formula runs at about a third of his own median when he leaves the tool out.
So we went back and looked at what our two failures actually said. One was a tool plus a feature. The other was a tool plus a task. Neither one named an outcome anywhere. The rule was never about hiding the tool. It was that the tool must never be the payoff, and a title that stops at a feature has no payoff in it at all.
One honest exception showed up in the same data. On the pure news channel drama beat outcome, because in news the outcome is the news.
The gate that is allowed to say no
The brief opens with three questions and all three have to pass.
Is somebody outside our bubble already searching for this? We have to be able to name the phrase a stranger types. If the only way to describe the topic is with a product name, it fails.
Do we have receipts nobody else has? It has to be running in production here, with real data, and it has to be showable. A tutorial about a feature we have never shipped against is a video the channel posting eight times a week already made better.
Is the payoff filmable as one held shot? If the finished thing cannot be filmed clean and still, the opening has nothing to open on. We name the shot before we agree to the video.
When something fails, the brief says so in two sentences and offers the nearest idea that passes, almost always a narrower cut of the same thing. Every automation I run fails the third question. The one that texts me when the fleet dies passes it. Same subject, one is a shoot day and the other is a shrug.
News topics swap the receipts question for a different one: can the money angle be written in one sentence right now? If not, it banks rather than dying, and gets picked up when it can be.
The beat we were skipping
The teardown found something we had simply left out. On the best performing build video in the sample, the last third is not the build. It is what to charge and who to pitch. That channel's other strong video has no build in it at all.
So a beat got added. Sixty to ninety seconds, straight after the proof lands, saying what the thing is worth, what an agency would charge for it, and who buys it. It is the beat most likely to get cut when a video runs long, which is exactly why it is written into the brief in actual words rather than left as a note to improvise.
What this means for your business
You probably are not making five videos a week. You are making proposals, or landing pages, or offers, and the same shape applies.
Put a gate in front of the work that is allowed to fail. Most businesses have a process that decides how to do a thing and no process that decides whether to. Three questions, answered honestly before anything gets built, will kill more waste than any productivity system you install afterwards.
Then check whether your own rules came from a sample worth trusting. Ours came from eleven videos and was confidently wrong. Any marketing automation San Jose operation runs into the same thing: a rule built on your own small numbers is a guess in a confident voice, and the cheapest fix is buying a bigger sample.
Want this built for you
We build the systems that decide what is worth doing, and the websites and CRM automation underneath them. Start at optechsol.llc.