Teardown, build, run, shoot

Services

We build video, and the production systems behind it, for brands that want a content engine, not a campaign. There are four ways to buy it. Each is scoped on a short call. Prices are below, and no result is promised.

Ways to work together

Teardown, build, run, shoot.

Each one starts with a short call and is scoped in writing before any work begins.

A channel that posts without a plan pays for it every month: videos made without feedback, and no way to tell which ones to repeat. The Teardown puts a number on that gap from your own channel’s figures, so you decide with the cost of waiting in front of you.

Audit

Teardown

You get: A written teardown of one channel and the plan that follows: a dated snapshot of its public numbers, fixes in priority order, what each costs in time, tools and AI usage, and what waiting costs, worked out from your own numbers. Each fix has its evidence, an effort rating and what we would measure.
Price: $150, flat, paid up front. Credited in full toward a Build, a Run or a Shoot if you start within 14 days.
Rough timeline: A few days after we have the link. A guide, not a deadline.
We need from you: The channel link, what you want to change and how you measure success.
Not included: Doing the work, a forecast or a promised result. A teardown names what to change. It changes nothing.

See a sample teardown

System

Build

You get: A production system for short-form or long-form video, set up around your format, with a QA hold and a kill switch built in.
Price: From $3,000, as a fixed scope agreed in writing. Half to start, half on delivery.
Rough timeline: Weeks, depending on the format. The written scope gives the estimate.
We need from you: Access to the channel it will publish to, a written brief for the voice and the look, and one person who can say yes or no.
Not included: Running the channel week to week, which is the Run. A promised audience.

Operation

Run

You get: A channel we operate month to month, with numbers every week: the median, the misses and what changes next.
Price: From $2,000 a month, billed monthly in advance. Either side can stop at the end of a month.
Rough timeline: Month to month, with a review at the end of each month.
We need from you: A channel we may publish to, a written list of what is off limits, and one person who answers questions as they come up.
Not included: A growth target. We report the numbers every week and promise none of them.

On camera

Shoot

You get: I shoot, edit and schedule a batch of short-form video for your channel myself. The Teardown sets the shot list and the posting plan first.
Price: From $2,500 per shoot day, edit and posting plan included. Half to book, half on delivery.
Rough timeline: Booked from the Teardown. I take a limited number each month, so dates are agreed in writing.
We need from you: A location and access, a written brief for the voice and the look, and one person who can say yes or no.
Not included: A promised audience, or a video that goes viral. Nobody can sell either. A shoot day makes good videos; the numbers decide the rest.

Fit

Who this is for, and who it is not.

We would rather say no early than late, so the line is written here.

Who this is for

Brands and creators who already have a product or an audience, and want short-form and long-form video made on a schedule, as a system.
Teams with one person who owns the voice and can say yes or no.
People comfortable with AI-assisted production that is disclosed.

Who this is not for

Anyone who wants a result promised up front.
Anyone who needs a borrowed logo, a track record with other people’s brands that does not exist yet, or proof that we have grown a channel to your size. The largest of our three channels has 351 subscribers on its public About page, read Oct 2, 2026.
Anything that needs a person on camera who is not you.

How an engagement runs

Four steps, in this order.

We write down each one, and we do not start the work before the scope is agreed.

Call
A short call. You say what you want to change and how you would measure it. We say whether we fit, and we say no when we do not.

Ends with a plain yes or no.
Scope
A written scope before any work. It names what you get, what is not included, who says yes or no, and what is off limits.

Nothing starts before you agree.
Work
The work, with numbers shared every week: the median, the misses and what changes next.

A weekly note, in writing.
Review
An end-of-month review: what moved, what did not, and whether to continue, change or stop.

One decision, and it is yours.

What you receive

Sample teardown: Carousel Studios & Lab, as of Oct 2, 2026

This is a teardown as it arrives: a dated snapshot of what the public numbers show, five fixes in priority order, each with its evidence, its effort and what we would measure, and a plain statement that none of it has been tested yet. The channel is ours, so every figure can be checked.

1 The snapshot

50

Public videos published

Carousel Studios & Lab · Sep 18 – Oct 1, 2026 · as of Oct 2, 2026 · YouTube Data API

676

Median views per video

Same 50 videos · Sep 18 – Oct 1, 2026 · as of Oct 2, 2026 · YouTube Data API

1,492

Best single video“King Tut's Tomb Might Have a Secret Door”

Carousel Studios & Lab · Sep 18 – Oct 1, 2026 · as of Oct 2, 2026 · YouTube Data API

4

Videos under 100 views

Of the same 50 · Sep 18 – Oct 1, 2026 · as of Oct 2, 2026 · YouTube Data API

What this proves: our own channel shows the production system can publish in public at up to four Shorts a day (50 public videos in 14 days), and that it has stopped when its sign-ins expired. It does not prove that audience growth will follow for anyone else.

These are channels and systems we own and run. They are not client results and are not a forecast of yours.

2 Five fixes, in priority order

Ranked by how directly the public numbers point at the problem. Effort is our estimate of the work, small, medium or large, not of the result.

01

Widen the look-back on stories already covered.

Evidence

The California sea turtles story ran three times in the window, under titles that differ by a few words and, in one case, by a full stop. The research step is told only what ran in the last two days.

One story, three uploads. Public views as of Oct 2, 2026, inside the window Sep 18 – Oct 1, 2026. Source: YouTube Data API.
Title	Published	Public views
Sea Turtles Just Nested in California. A First.	Sep 25, 2026	1,148
Sea Turtles Laid Eggs in California. First Ever.	Sep 27, 2026	1,042
Sea Turtles Just Nested in California. A First	Sep 30, 2026	1,178

Counted by story, not by title, four more stories ran twice: King Tut’s tomb door, a building-sized rock that hit the Moon, 40 buffalo in New York, and a burned wall under Paris, whose two titles share almost no wording. At least five stories account for 11 of the 50 uploads. These repeats mostly did not flop: 8 of the 11 uploads sit above the Shorts median of 728, and so do 5 of the 6 that came after a story’s first run. Hypothesis: the cost is to the follower, who sees the same story twice. The numbers cannot show whether it loses anyone.

Fix

Tell the research step what ran in the last two weeks, not the last two days.

Effort

Small A change to what the research step is told about earlier stories.

What we would measure

Stories that run more than once per window (at least five here, across 14 days), and the views of a repeat against its first run.

02
Decide what long-form is for.

Evidence

The Night Edition runs 60:04. As of Oct 2, 2026 it has 11 public views and 0 likes. Every other upload in the window is a Short of about a minute, with a median of 728 views. In the same window Shorts earned 33.2k views and the long-form earned 11. It took about two hours of graphics-card time to make.

Hypothesis: a one-hour video gets no shelf space on a channel whose audience arrives for Shorts. The numbers fit that reading and do not prove it.

Fix
Decide what long-form is for, and write the decision down. There are three honest options: a home of its own, a place that Shorts point to, or stop making it. Publishing it into an audience that arrived for Shorts is the one option the numbers argue against.

Effort
Medium A decision, then either a second home to set up or a line to cut.

What we would measure
Public views and likes at 7 days and at 28 days for each long-form upload, whichever option is chosen.

03

Read the lowest Shorts before making more like them.

Evidence

These are the four lowest Shorts in the window. The first three got under 100 views. The fourth video under 100 in the snapshot is The Night Edition, in fix 2.

The four lowest Shorts, lowest first. Public views as of Oct 2, 2026, inside the window Sep 18 – Oct 1, 2026. Source: YouTube Data API.
Title	Published	Public views
He Measured Gravity Blind For 10 Years	Sep 20, 2026	9
New Words Added to Dictionary 2026: Rickroll	Sep 21, 2026	38
This Aztec Book Left Mexico in 1840. It's Back	Sep 21, 2026	84
Bikes Beat Cars On This Street At Rush Hour	Sep 23, 2026	109

The four titles share no topic and no format that we can name. Three of the four went up within about a day of each other. That could be chance. Hypothesis: timing or delivery, not the story, held them down. Public numbers cannot say which.

Fix

Before the system makes more Shorts like these, read the lowest five each week against their neighbors: topic, title, time of day and views at 48 hours. Write down a cause, or write down that it is unknown.

Effort

Small One person, reading five videos a week.

What we would measure

Shorts under 100 views in each window (3 of 49 here), and whether any written cause appears twice.

04
Test cadence against reach.

Evidence

Cadence broke on three days, and not by choice. The channel averaged 3.6 public videos a day across the window, and on 11 of the 14 days, in Pacific time, it posted four Shorts. The three lighter days were Sep 27 (two Shorts), Sep 28 (one) and Oct 1 (two). Volume does not explain which Shorts do well: on 10 of the 11 four-Short days, the best Short of the day had at least three times the views of the worst.

Sep 27 and 28 were a production failure, an expired sign-in written up in Step 6 of the Shorts system guide. Oct 1 has only its first two posts, and nothing went up on Oct 2 before our read, a gap that Step 6 also traces to a fault. So Oct 1 appears to be the start of that failure, though public numbers cannot confirm it.

None of the three days was a cadence setting, so this window cannot show whether posting fewer would raise the level for every Short, or whether posting four lowers it. The three lighter days hold only five Shorts, too few to compare in any case. Hypothesis: reach per Short may fall as the count per day rises. This window cannot confirm it.

Fix
Test cadence on purpose. Hold the schedule at three a day for two weeks, keep the rest steady, and compare against the four-a-day weeks. Mark any day lost to a failure and leave it out of the comparison.

Effort
Medium Two weeks at a different rate, and a person to keep everything else steady.

What we would measure
Median views per Short at 48 hours, by the number of Shorts posted that day, over at least two weeks per setting, with failure days marked and left out.
05
Compare Shorts at equal age.

Evidence

This teardown reads views once, as lifetime views. The 24 oldest Shorts in the window had up to 13 days longer to collect them, yet their median is below the median of the 25 newest. A single read mixes age with quality, and it cannot say why the newer Shorts did better.

Hypothesis: something about the Shorts changed during the window, or the audience grew. Public numbers cannot say which.

Fix
Read every Short at the same ages: views at 48 hours and at 7 days, taken on a schedule. Log each change to the system with its date, beside the numbers.

Effort
Small A scheduled read, and one line in a log for each change.

What we would measure
Median views at 48 hours and at 7 days, week by week, with the change log beside it.

Start with a channel teardown.

Send the link to one channel and tell us what you want to change. You get a written teardown in this format, built from your numbers.

Request a teardown

Or write to hello@grandblademedia.com.