How Dipana grew sales by 118% with no historical data

The brand had neither media history nor benchmarks: Dipana entered digital from scratch. The usual route — forecasting from past campaigns — had nothing to build on.

We built a model on category data, derived the reach required from target sales and launched. Four months in, we saw that the creative behaved differently from what the model assumed, and refined the coefficients.

The idea: from sales to reach, not the other way round

The model converts reach into sales through the chain 'reach → attention → awareness → sales'. Then it works backwards: from target sales we get the reach we need, and build the media plan from it.

A model we had to correct

That is why the accuracy is 9.8% and not 4%: the first version of the model relied on the category, not the brand. Updating coefficients mid-campaign is not a failure, it is part of the approach.

What we did

• Built the model 'reach → attention → awareness → sales' on category data.
• Derived the required reach from target sales and assembled a media plan with per-channel settings.
• Launched in digital and tracked actual sales as the campaign ran.
• After four months, refined the conversions to match how the creative actually performed and rebuilt the forecast.
• Carried the campaign through to the planned sales volume.

Results

• Sales in January—August 2025: 46,114 packs against 21,090 a year earlier — up 118%.
• Over twelve months: 64,657 against 28,564 — up 126%.
• Market share grew from 0.3% to 0.8%.
• Forecast for the campaign period — 35,229 packs, actual — 38,693. Accuracy 9.8%.

What it means

The case shows that a model is not a document written before launch but a working tool. The first forecast may be off, and that is fine: what matters is that the coefficients get refined against reality and the campaign still reaches planned sales.