Minimum Viable Budget: Costing a Film Backwards From What It Can Actually Return

India releases more feature films than any country on earth — 1,769 of them in 2022 alone — and in a representative recent year only about forty per cent of releases were profitable. That is not a talent shortage. India does not have a shortage of people who can direct, shoot, cut or perform. What it has is a shortage of the boring apparatus that any other capital-intensive industry treats as non-negotiable: a costed plan, a baseline, a change-control log, and someone whose job is to say the number out loud before the money is spent rather than after.

The overrun data says the same thing everywhere the industry has been measured. Studio productions run roughly thirty-one per cent over budget on average; independents run closer to forty. In 2022, seventy-two per cent of films budgeted above one hundred million dollars went over. The industry has known this long enough to have built a financial instrument specifically to price the risk — the completion bond, typically a fee of around two per cent of the budget on top of a mandatory contingency of about ten per cent of direct costs. That is the market's own estimate of how likely a film is to miss its plan. Twelve per cent, in cash, before a frame is shot.

None of that is a criticism of producers. It is a description of a gap. A production house is organised to make films. It is very rarely organised to run a capital programme, and a film is a capital programme wearing a beret.

Almost every film budget in the world is built forward. Someone establishes what can be raised — a slate commitment, a satellite pre-sale, a distributor's advance, a family office — and that figure becomes the budget. The script is then made to fit it, roughly, by a line producer working from a page count and experience. The number that never gets calculated is the one that matters: what the film has to earn before a rupee comes back, and whether anything comparable has ever earned it.

Costing backwards inverts the whole exercise. You start from the realistic gross floor of the film's own comparable set, run the recoupment waterfall down to what the producer actually receives, and the result is a ceiling. That ceiling is the Minimum Viable Budget: the all-in negative cost at which the film still returns capital if it performs at the low end of its band, not the high end. Every rupee above it is a bet that this film outperforms every comparable it has.

The direction of the arithmetic is the entire argument. Forward, the budget is an input and the return is a hope. Backward, the return is the input and the budget is a constraint — which is the only form in which a budget can actually be defended in a room.

The standing rule is that a film needs a gross of roughly two to two and a half times its production budget before it breaks even. The reasoning is sound as far as it goes: the exhibitor keeps around half of net box office, and prints-and-advertising has historically added about half the production budget again on top.

But that rule was formed in a studio context where the studio is also the distributor. Drop an independent Indian producer into the same waterfall and two more lines appear. Admissions carry GST before anything is shared. And a distributor who is not you takes a commission on the distributor's share, typically fifteen to twenty per cent, and recovers P&A off the top before anything reaches the producer. Run those in and the multiple for an independent lands closer to three times, not two and a half.

This is exactly why the rule of thumb is a starting point rather than an answer. It gets you into the right postcode. The actual figure comes from writing that specific film's waterfall out, with its own tax treatment, its own distribution terms, its own P&A commitment and its own cost of money — which takes an afternoon and is almost never done.

What follows is a worked example rather than a client engagement — the structure is the one we run, the rates are published industry averages, and every figure is shown so the arithmetic can be argued with. A mid-budget Hindi-language thriller. One recognisable lead at a defined star tier. A planned wide release. The producer arrives with a settled intention: eighteen crore.

Gate 0 starts with comparables, not with the script. Same genre, same language, same star tier, same release window, last five years, all-India theatrical. That set brackets a realistic gross band of ₹22 to ₹34 crore, with a median of ₹27 crore. We plan against the floor of the band — ₹22 crore — because a budget defended by the median is a budget that fails half the time.

The MVB is a ceiling, not a budget. The next question is what the script actually costs, and it is answered scene by scene rather than by page count. Every scene gets a unit, a location class, a cast day-out-of-days entry, a stunt or crowd flag, and a VFX methodology — not a VFX line, a methodology, because 'we'll fix it in post' is not a method and it is where the money goes.

Costed that way, this script comes to ₹11.4 crore. Not the ₹18 crore that was going to be spent, and not the ₹7.0 crore that can be recovered. The gap is entirely legible once it is broken out.

Consolidating two cities into one removes six company moves and four shooting days without touching a single line of dialogue — the second city was a location note, not a dramatic requirement. The crowd sequence plays identically with a controlled unit of 120 and generative crowd extension on three wide shots; nobody in the room can tell, because the coverage was always close.

The VFX pass is the largest single change and the one that most needs technical people in the room while the shooting script is still soft. Of 47 shots, 18 are achievable in camera once the blocking is adjusted, and 13 more can be built as 2.5D from generative plates using a reconstructed depth and matte stack rather than a full CG rebuild. That leaves 16 genuine VFX shots with an agreed methodology and a named vendor, budgeted rather than deferred.

Night exteriors drop from eleven to six by moving two sequences to dusk-for-night, which the cinematographer had already suggested and which had never been costed as a saving.

The film lands at a 38-day schedule, three company moves, and ₹7.3 crore of production budget with ₹0.73 crore of governed contingency — ₹8.03 crore all-in. Inside the MVB, with the twenty per cent return band intact, telling the same story with the same beats and the same ending.

Barry Boehm published the observation in 1981 and it has survived four decades of argument: the cost of correcting a mistake rises steeply the later it is found. Requirements, design, code, test, production — each step costs meaningfully more than the last. Boehm and Basili's later work put the ratio at around a hundred to one between requirements and post-delivery on large systems, and around five to one on small ones.

It is worth being precise about what is durable there, because the figure gets quoted more confidently than the underlying data supports. The reliable claim is the shape of the curve, not a specific multiplier. But the shape is enough, and film has a steeper version of the same curve than software ever did. A structural note on a screenplay costs a fortnight of a writer's time. The same note after the unit has wrapped costs a reshoot, a cast availability window, a location that has changed season, and a delivery date that has already been sold.

Where the analogy stops matters too, and we say so to producers before they ask. Software can ship an increment and learn from it; a film cannot release act two and iterate. So we do not run film production as Scrum. We run stage gates for the irreversible decisions — script, budget, schedule, cast, methodology — and sprint cadence only inside a gate, where prep tasks, VFX shots and post deliverables genuinely can be planned, burned down and re-forecast weekly. Gates for the things you cannot take back. Sprints for the things you can.

A gate is not a meeting. It is an artefact, an approval and a decision recorded with its reasoning, so that six months later the question 'why did we agree to this' has a written answer rather than four conflicting memories.

Script analytics is real and it is useful, and it is oversold constantly. Systems trained on thousands of produced screenplays with known outcomes report accuracies in the low-to-mid eighties on financial outcome; Warner Bros. moved to analytics-assisted greenlighting and reported material improvement in selection. Those are meaningful numbers and they beat a room full of instinct.

They are also not an oracle, and we will not present them as one. An eighty-five per cent accurate classifier is wrong about one film in seven, and the film it is wrong about might be yours. Prediction is used here for two things only: narrowing the plausible return band so the waterfall has an honest input, and pointing at specific structural weaknesses — a flat midpoint, a protagonist who stops wanting anything in act two, a genre promise the third act does not pay — that a human writer can then actually go and fix.

Anyone selling certainty about a film's box office is selling something else. The value is not in the forecast. It is in having a written, defensible number early enough that changing the plan is still cheap.

Gate 0 is deliberately small and deliberately separable. A script, a target scale, and a fortnight produces the audit, the comparable band, the modelled return and a written greenlight memo. At that point a producer has the one thing the industry systematically lacks — a defensible number, early — and is free to walk away with it.

If the film continues, Gate 1 produces the Minimum Viable Budget and the scene-level costing, which is where the money is actually saved. Gates 2 through 5 are a retained engagement across prep, shoot, post and delivery, sized to the production.

The order matters more than the tooling. Comparables before budget. Budget before schedule. Schedule before dates. Methodology before plates. Every one of those is reversible on paper and expensive in the field, and the entire discipline is nothing more than insisting they happen in that sequence.