Customer case · Aprila Bank

37% more applications. Same budget.

Aprila Bank finances Norwegian small businesses. Their marketing team knew the channel mix was not optimal, but last-click reporting could not tell them where to move the money. One model changed that.

Book a demo Read the story

Industry: Fintech, SME lending
Market: Norway
Channels: Digital and offline in one model

A small business owner at a workbench in a Nordic workshop — +37% more loan applications from the same budget

The challenge

Three questions the reports could not answer.

Aprila's team was already data-driven. The budget balanced brand building against performance, and performance was steered by last-click attribution and each channel's own reporting. Some channels were clearly flattening out, but nobody could say which, or by how much.

Which channels actually drove applications, rather than taking credit for demand that was already there
What would happen to sales if budget moved between channels, when the underlying reports were not trusted
How to get a modern model in place without spending scarce data-science time building one in-house

What we built

Every channel in one model, in days.

Odins collected and structured Aprila's marketing data across every channel, digital and offline, and trained a Bayesian marketing mix model on it. The structured data flows back into Aprila's own warehouse, and the model retrains every month.

Years of history collected and structured in days, not a six-month data project
Channel contribution measured against actual applications, not last-click attribution
Saturation curves showing where each channel stopped paying for itself
Specific monthly moves, reviewed by our team before they reached Aprila

The results

What changed, in numbers.

+37% More loan applications From the same total budget, by reallocating out of saturated channels.

~15% Lower cost of data work Less time and spend on building and maintaining integrations.

Days To structured marketing data Years of history collected and structured before the first model ran.

Measured 7 October 2024 to 26 January 2025, adjusted for the change in spend over the period.

37% more applications. Same budget.

Three questions the reports could not answer.

Every channel in one model, in days.

What changed, in numbers.

Want the same answer for your channels?