Deployment

The calendar: how the trap is escaped in practice. Each phase is a principle on the clock.

The deadline is not a target

Before any of this makes sense, understand what "deadline" means in film — because it is not what it means in software. In software, quality is fixed and the date slips. In film, the date is fixed and the quality slips: when a production is late, the deadline doesn't move — the movie changes. Shots get cut, scope gets simplified, bodies get thrown at the problem, and the work gets triaged. The deadline eats the film, never the reverse.

The date can't move because it is load-bearing for money that dwarfs production: the release slot is booked, the marketing spend (often larger than the production budget) is committed against it, presale contracts define delivery as the deal, air dates are sold to advertisers a year out, and production loans accrue interest by the week. The institutional proof: an entire insurance industry — completion bonds — exists solely because the date cannot move. A bond company will seize a film and finish it themselves rather than let the date slip. When finance invents an industry to enforce a constraint, the constraint is real.

Everything below follows from this one fact.

Pipeline Precedes Production

The fourth foundation stone is a mandate: infrastructure must precede production, or the money is wasted — because it will never be used. Two facts make it non-negotiable:

The onboarding window

The only free time in all of production is while the newest team is onboarding — two to three weeks, never more. The crew isn't productive yet; no deadline owns those hours. That is the entire installation budget.

The delivery lock

A law of production: when deliveries start, infrastructure locks — and it locks harder as the deadlines grow. Nobody has ever built a good pipeline in the middle of hardcore deliveries. (Manufacturing has known this for a century: you never retool a line mid-run; changes between runs, never during.)

Between those two facts, pipeline-precedes-production is not a preference — it is a theorem. Production is a locomotive: it departs on schedule and moves forward whether you like it or not. If pipeline isn't done when it leaves, you don't have a choice — it is better to have people waiting than to miss the deadline — and unfinished pipeline simply gets left on the platform.

The defection

The human mechanism underneath the lock — perhaps the cause of pipeline failure in this industry. The deadline is the deadline: if the systems aren't there when it arrives, artists don't wait — they give up and build their own systems from whatever they're comfortable with, and chaos ensues. Three properties make this fatal rather than merely messy: the defection is rational (an artist's responsibility is shots, not systems — when the pipeline fails them, protecting their shots by any means is correct behavior); it is one-per-artist (chaos isn't the absence of a system, it's the presence of N personal ones — the garage regenerates itself in a single crunch week); and it is irreversible for the show (nobody migrates off the system carrying their shots mid-delivery, and trust in the official pipeline is spent).

The consequence: the pipeline must already be the comfortable path before pressure arrives — pressure sends everyone fleeing to comfort, and the flight must land on the line. This is why "artists never change tools" is a survival rule, not a courtesy, and why the nail-it loop must finish inside the onboarding window: a workflow still being nailed when deliveries start is a workflow artists will abandon mid-nail.

Fifty percent is zero

A system that is 50% done at its department's point of no return is worth exactly nothing — and every department's window closes on its own date.

Every department has a point of no return — a date past which what you have is what you have. And the value of a pipeline at that date is a step function, not a line: if you have 50%, you have nothing, because the person responsible has to move on with or without your system, and nobody stakes a delivery on half a system. Below complete-enough-to-rely-on, adoption isn't partial — it's zero, the defection is total, and the money spent is vaporized, not discounted.

The lock is per department: production is not one locomotive but a train schedule — modeling's window closes before animation's, animation's before comp's. Each department's system is a separate all-or-nothing bet against its own departure date, which turns "nail it, then scale it" from advice into arithmetic: every lane's nail-it loop must finish before that lane's train leaves. And it explains the trap most studios live in: pipeline built during a show loses to the deadline, gets abandoned, bodies get thrown, and the corpse becomes evidence that "pipeline doesn't work here."

The calendar

Phase	What happens
Day 1	A foundation must exist day one — files, IT, ontology already standing. Nothing that already exists gets built on the clock.
Weeks 1–2	Build what the supervisor wants — nail that workflow with the lead responsible, inside the onboarding window.
Live	From then on it's in production — the system is proven where it counts, because there is nowhere else.
Ongoing	Fixes on the fly as artists discover new things — stop it, fix it, keep the line moving. Discovery never ends; the pipeline keeps up or gets abandoned.

Why adoption doesn't take five years here

Traditional technology adoption in film is political: every department must feel heard, and one supervisor's refusal stops production. Three properties of this framework remove the political cost along with the technical one:

1. Buy-in by subtraction — departments are never asked to adopt a technology; chores they hate are removed. The refusal reflex has nothing to grab.
2. Artists never change tools — the biggest refusal trigger is off the table by principle. The pipeline adapts to the artist.
3. The lead builds it with you — nail it, then scale it means the person with the power to refuse is instead the co-author.