Most Indian marketing teams have heard of the 70/20/10 split: seventy percent on what works, twenty on scaling what shows promise, ten on genuinely unproven bets. Most also apply it wrongly in the same two ways.
Mistake one: the ten percent is not experimental
Ask a team what their ten percent is and the answer is usually a new ad format on an existing platform, or a channel their competitor already uses. That is not experimental, it is the twenty percent wearing different clothes.
Real ten-percent work has a genuine chance of returning nothing. A research report before you know anyone will cite it. A tool before you know anyone will use it. A channel with no local playbook. If every bet in the bucket pays off, the bucket is set too conservatively and you are not learning anything.
Mistake two: cutting it first
When quarterly numbers slip, the ten percent gets cut. It is the easiest line to defend cutting, because by definition it has not produced results yet.
This is why so many Indian marketing teams have identical channel mixes to their competitors. Everything that might have differentiated them was cut during a soft quarter, repeatedly, over years.
The fix is to protect it structurally rather than by willpower. Treat the ten percent as fixed cost, and if the quarter is bad, take the cut from the seventy percent instead. That hurts more in the moment and is why almost nobody does it.
A more useful split for small teams
Under about two crore annual marketing spend, 70/20/10 fragments into pieces too small to matter. Two experiments a year at ten percent of a small budget is not a learning programme.
Better: pick one channel to be genuinely good at, one to maintain, and one bet per year that gets real resource rather than a token slice. Depth in one channel beats presence in five, particularly against larger competitors who can outspend you everywhere at once.