Which Film Experiment Is Worth Doing Before the Next Pitch?
Which Film Experiment Is Worth Doing Before the Next Pitch?
You can fund one development experiment before the next pitch. The question is not which test would look best on a screen in a small room. It is which answer would change something.
The best next experiment is the one that can change an important decision, and especially one that other decisions are already leaning on. A test whose different outcomes all lead you to the same place may still be worth making — but it is a persuasive asset, not an experiment, and it should be briefed and judged as such.
That distinction is easy to blur because a promotional sample feels like progress. It is progress; it just isn't evidence about the thing you don't know yet. A finished scene can sell a tone, give an actor something to answer, put a face on a project. None of that tells you whether an unresolved approach holds. Those are different commissions, and conflating them is how a budget gets spent proving something no one doubted.
Four moves sort the candidates: list the decisions before the shots, follow the dependency that other choices assume, write the result branches before approving anything, and bound the test without stripping out what it has to reveal. The comparison that runs through the rest of this piece is constructed, not reported — two candidate tests for a fictional feature, neither of them shot.
List decisions before listing things to shoot
Begin with the unresolved choice, not the shot list. "We need an impressive scene" is an intention. "We do not know whether this form can carry this interaction" is a decision: it has more than one live answer, and guessing wrong costs something specific.
For each candidate, write down three things in ordinary sentences. What are the alternatives still actually available — not the version you've quietly settled on, but the routes you could still take? If you assume wrong, what changes? And which commitment does that assumption move?
Then name who will use the result and when they need it. Information that arrives after the relevant commitment cannot help that decision. If the pitch is the commitment, a test that lands the week after it, however good, has changed nothing about what you pitched. It might still shape what you develop next, which is a legitimate job — just not the one you thought you were buying.
A useful way to test a candidate: state the question so that both a yes and a no are plausible and consequential. If you can't finish that sentence, you probably have a demonstration in mind rather than an uncertainty.
Follow the dependency that other choices assume
Some unknowns sit upstream. Answering them changes the shape of a lot of work that came later. Other unknowns sit downstream: demonstrating them already assumes the upstream question is settled. The two feel similar in a proposal — both are "tests" — but they occupy different positions in the stack, and the position determines what a result can move.
Take a fictional feature whose stated form is long, uninterrupted dinner scenes. Two candidates are on the table.
The first is a sustained overlapping exchange: one continuous take of the kind of group conversation the film depends on, several people talking over each other, no cut available to rescue the rhythm.
The second is a polished, effects-heavy opening image, finished to the standard you would actually show.
Neither has run. The comparison is a paper exercise, built to make the choices visible.
Now trace what each result would touch. For the sustained exchange, the answer feeds camera planning, scene construction, rehearsal time, lighting, and the schedule — and it feeds the writing, because how much a scene leans on simultaneous talk depends on whether the take can hold. If the form works, the film gets built around it. If it doesn't, you cut, and a great deal downstream of that decision changes.
For the opening image, the answer touches the opening image. If it lands, you keep it; if it doesn't, you design another. The film's approach to its dinner scenes does not move either way. That is what makes it downstream: it demonstrates something rather than deciding something the rest of the film is waiting on.
Position alone doesn't settle it, though. Ask whether each test can actually distinguish the alternatives under credible conditions. For the exchange, the conditions have to include the essential pressure — the overlapping dialogue and the absence of a cut. Rehearse it down to polite turn-taking and cut every eight seconds for safety, and you have tested a different form and learned little about yours. For the opening image, a single hero shot may confirm the look while leaving the real question untouched. If what you need to know is whether that world can be produced at the quantity and consistency the film requires, one image is not a test of it.
Then weigh the cost of false reassurance. The opening image will produce usable material. Usable material feels like evidence, and a striking frame in a pitch room makes everyone — you included — more confident. The confidence is real. Its calibration is not. An easy test that comes back beautiful is the easiest kind to over-read, because nothing in the room is arguing with it.
There is a countercondition worth holding. If the visual effect actually enables the story world — if every dinner scene necessarily plays inside it or against it, so the world is the premise rather than the backdrop — then the opening image is no longer downstream. It is the dependency, and it moves to the front of the queue. That is also true if the effect has to work at a scale the film can't proceed without. The position comes from the project, not from a rule that performance tests always win. The test still has to be built so it can answer the world-scale question instead of producing one good frame.
Compare the candidates in plain language: what each result could change, whether the test can tell the routes apart, what it costs to get useful information, and when the information lands. No scores, no invented odds. The ordering comes from consequence and dependency, which you can usually argue about out loud.
Write three result branches before approving the experiment
Before you approve either test, write three outcomes: positive, negative, and inconclusive. Not the mood you expect. The action.
For the sustained overlapping exchange:
- Positive. The unbroken take holds and the overlapping exchange stays legible and tense under the tested conditions. This supports developing the long-take approach — under those conditions, at that length, with that preparation. It does not graduate into proof that the feature works.
- Negative. The take doesn't hold; the overlap turns to mush, or the tension needs the cut. This prompts an edited-coverage alternative. You write and stage for the cut, and scene length, lighting, and schedule shift accordingly. A negative result here redirects an approach. It doesn't disprove the project.
- Inconclusive. The footage neither settles the question nor refutes it — the wrong room, a stand-in cast, too little rehearsal, a scene that wasn't representative. The form stays unresolved, and the result names the preparation or evidence that was missing. This is not a soft positive. Usable images did come out. They are not an answer.
For the opening image:
- Positive. Supports that particular visual proposal, which can go into the pitch as a realized thing.
- Negative. Prompts a replacement image. The film's form is untouched.
- Inconclusive. Defers the visual claim pending a narrower technical question — often about repeatability rather than the look.
Now compare the columns. Divergence alone is not the test. The opening image's three branches do diverge — you keep it, you replace it, you defer the visual claim — but every one of those outcomes moves the image and nothing else. The dinner-form question sits exactly where it was. What separates the two candidates is which decision the branches can move: a real experiment diverges on the decision it was commissioned to settle, whereas an asset diverges only on the asset. So ask what each set of branches touches, not merely whether the branches differ. If positive, negative, and inconclusive all lead to the same next action, the work cannot change anything, whatever it produces — and if you are making the opening image no matter what the result, that is plainly a promotional asset, to be commissioned as one. But divergent branches do not make it evidence either. What it shouldn't do is occupy the single slot you had for the test that could have moved the dinner-form question.
Timing belongs in this comparison too. Suppose the opening image consumes most of the available weeks and lands days before the pitch. Even a strong result arrives too late to change how you describe and budget the film. The exchange might come back fast enough to do exactly that. Two tests can be equally revealing in principle and very different in practice, because only one of them finishes before the decision it was meant to inform.
Bound the test without removing what it must reveal
An experiment needs an edge. Define the smallest conditions under which the result would mean anything, the evidence you will inspect, and the point at which you stop.
For the sustained exchange, those conditions probably include performers who can handle the overlap, enough rehearsal to be fair to the form, and a scene at roughly the length and density the film proposes. For the opening image, if the real question is quantity, the smallest credible test includes the repeat; one perfect frame tests the frame.
Decide what you will look at before you shoot, and make it observable. Not "does it look good," but "can I follow the exchange without the cut," or "does the tension hold past the sixth minute," or "does the second unit of this effect cost the same as the first." Pick the thing you can point to in the footage.
Name the stopping point out loud. Otherwise a discouraging middle turns into an open-ended shoot, and the experiment quietly becomes a small production. Say what you will not conclude, too. A three-minute test cannot validate a feature's form, and the pitch should say so in the same breath as the result: the approach held under these conditions. It is not de-risked. Keep what the result can justify at the pitch separate from the feature-scale responsibility it cannot carry, because the pitch is exactly where that line tends to get blurred.
Watch for the two standard ways a test becomes cheap and useless. The first is substituting an easier interaction — the calmer conversation, the shorter scene, the safer room — until you've answered a question nobody asked. The second is stripping essential preparation: the rehearsal, the real density, the actual overlap. Both reduce cost by removing the pressure that made the test worth running.
One more step costs almost nothing. Have the people who would shoot it say whether the proposed conditions could actually answer the stated question. That conversation is cheaper than the test and now and then reveals that the test cannot distinguish what you hope it will.
The choice, with the branches attached
For the fictional feature above, the sustained overlapping exchange is the experiment worth doing first. The decision it can change is the one the rest of the film is already leaning on: whether this story is built on unbroken coverage or on the cut. Its credible boundary is the tested conditions — those performers, that scene length, that preparation. And the actions are fixed in advance: a positive result means developing the long-take approach and budgeting the rehearsal and camera planning around it; a negative result means an edited-coverage alternative, with the rewriting and scheduling that follows; an inconclusive result means naming the missing preparation and either narrowing the question or telling the pitch plainly that the form is unresolved.
The opening image may still deserve to be made — as a promotional asset, briefed and paid for as one, rather than as the test that was supposed to settle the form.
If the visual effect genuinely enables the story world, the choice flips, and the method stays the same. That is the point. What you're choosing is not the impressive test or the cheap one. It's the question whose answer you can act on, in time, with the branches already written. A pitch that says "here is what we didn't know, and here is what we did about it" is a different and stronger object than one that says "we made a nice picture." The inconclusive branch gets the same respect as the positive, because pretending an ambiguous result settled something is the most expensive move available — you'll pay for it at feature scale, where the evidence was never strong enough to support the claim.
Frequently asked questions
How do I choose which experiment to fund before a pitch?
Start with unresolved decisions, not a shot list. Ask what alternatives are still live, what changes if you assume wrong, and which commitment that assumption moves. Name who will use the result and when they need it. Information that arrives after the relevant commitment cannot help that decision. A test whose outcomes all lead to the same place may be a promotional asset, not an experiment.
Why can a polished effects shot be less useful than a rough performance test?
If the effects shot only moves the opening image while the performance question sits upstream and other choices depend on it, the performance test can change more. But position comes from the project. If the visual effect genuinely enables the story world and must work at the required scale, it is the dependency and moves to the front. It still has to be built to answer the world-scale question, not just produce one good frame.
What is the point of writing positive, negative, and inconclusive result branches in advance?
The branches name actions, not moods. A positive result supports an approach under the tested conditions, not proof that the feature works. A negative result redirects the approach. An inconclusive result names missing preparation or evidence and leaves the question unresolved, not as a soft positive. Divergent branches alone do not make a test; what matters is whether the branches can move the decision the test was commissioned to settle.
How do I bound an experiment without stripping out what it must reveal?
Define the smallest conditions under which the result would mean anything, the evidence you will inspect, and the stopping point, all before shooting. For a sustained overlapping exchange, that likely includes performers who can handle the overlap, enough rehearsal, and a scene at roughly the proposed length and density. If quantity is the real question for an effect, the smallest credible test includes the repeat. Avoid substituting an easier interaction or stripping essential preparation; both remove the pressure that made the test worth running.
What if the opening effect is the story world itself?
Then it is not downstream. The choice flips, and the effect moves to the front of the queue. The method stays the same: state the question so both yes and no are plausible and consequential, write the result branches, bound the test, and say what you will not conclude. If you are making the image no matter what the result, that is plainly a promotional asset to be commissioned as one.