A worked example
₹3,00,000, twenty-one days, hour by hour.
This is how we would spend it, not a campaign we have run. Every figure below is an allocation — how the budget is divided, how many versions go in, when decisions are made. There are no results on this page: no cost per view, no install counts, no reach. Keyart has not run a client campaign yet, and inventing the numbers from one would be the easiest thing on this website to do and the last thing worth doing. When there is a real one, it will appear under Work with the client's name and their permission on it.
The other one ₹12,00,000 · 45 days · teaser, then trailer
Before day one
What we agree before anything is spent
A trailer, a release date and a budget. The first thing we settle is what counts as a result, because attribution cannot be backfilled and a week of spend without it is a week nobody can reconstruct.
One definition of a result
For a trailer that is usually a completed view of real length, or a ticket page reached. Agreed in writing, wired into the measurement before the first campaign goes live.
Nine angles, written down
Not nine edits — nine ideas about why somebody would watch, each one a sentence you can disagree with before we spend anything making it.
A daily ceiling you set
Enforced hourly in our own system, not only in the ad account, because a platform-side cap is one settings change away from gone.
Days one to four
A fifth of the budget, spread thin on purpose
About ₹60,000 of the ₹3,00,000 goes across all nine angles, in every placement shape. Most of it is spent to find out what does not work, which is what it is for.
₹60,000 across nine angles
Roughly ₹6,500 each. Enough for each one to be seen by enough people that the difference between them means something, and small enough that eight of them being wrong costs a fifth of the budget rather than all of it.
Four shapes each
Square, vertical, tall and wide, in Telugu and English as a pair. The same angle can win in one placement and lose in another, and a single master cut hides that.
We are watching watch-through, not views
A view that stops at two seconds is not an audience. The number that decides which angles survive is how far through people get.
Day five
The cut, and it is ruthless
Whatever is not holding attention is switched off, not nursed. Typically two or three of the nine survive. We will tell you which of our own ideas failed, by name.
Killed, not paused "to give it a chance"
An angle that spent its share without holding anyone is finished. The budget it would have taken is worth more behind something that works.
Fresh variants of the winners
The two or three that held attention get new cuts of the same idea, because the original will fatigue in the auction before the window closes.
You get the list
Every decision records the rule and the numbers behind it. "Why was this stopped" has an answer on the day and a year later.
Days six to eighteen
The rest of the budget, moved nightly
About ₹2,10,000 goes behind what is working, rebalanced every night against the definition agreed on day zero. Channel budgets move weekly, not daily — daily channel-hopping is noise.
Scale, hold, kill, rotate, rebalance, shift
Six rules, with thresholds written down and versioned before they are applied. The system follows them every night without missing one, which is the part that is hard to copy.
Frequency is watched as closely as cost
Rising frequency with falling performance is fatigue, not failure. A fresh variant goes in and scaling is suppressed that night rather than pushing harder at people who have already seen it.
Nothing waits for a weekly call
In a twenty-one day window a week of waiting is a third of the campaign.
After the window
What you are left holding
The last ₹30,000 stays in reserve for the final days, when the release is closest and the auction is most expensive. Then the campaign ends and you get the record of it.
Spend, results, and cost per result
Broken down by platform, by creative and by city, against the one definition agreed before anything ran.
Every automated decision, with its reason
Which rule fired, on which night, against which numbers. Not a summary written afterwards — the record itself.
Frozen when it is sent
The copy you have is the copy we have, permanently. It is the thing the next film starts from.
The honest part
What this example cannot tell you
Whether it works for your film. A budget of ₹3,00,000 against a wide release is a rounding error; against a first feature in one language it is a real campaign. The split above is the shape of the method, not a promise about an outcome, and the first thing we will do on a call is tell you which of those two your release is.