Transforming research data with AI to build a predictive financial needs model.

Project Description

We applied AI to traditional research data to uncovered far more powerful determinants of financial needs than Sorted’s original research had identified.We turned this into a predictive model that fused multiple data sets to inform successful media targeting at scale: reaching those that needed Sorted most.

KEY Outcomes

0%

Increase in site visitation

0%

Increase in total interactions on site

The Challenge

Sorted.org.nz is the consumer website of Te Ara Ahunga Ora Retirement Commission. It’s a trusted source of free, impartial, and independent financial information for New Zealanders.

Together was tasked with increasing site visitation and engagement amongst those that needed help the most. Existing financial needs research suggested focusing on specific demographics to find those most in need, but our analysis of the same data showed demographics were just 38% accurate in predicting someone’s financial need.

The Solution

We trained a machine learning model to re-analyse the research data to find more accurate determinants of financial needs. Our approach found that answers to just 45 questions of the 583 in the original survey had very high accuracy in assigning respondents to future financial need clusters.

We then turned this insight into targetable media audiences through data enrichment and custom programmatic audience builds.