Insight, Impact, Innovation
HomeOur WorkCase StudiesAttrition model enables incremental home loan balance savings of $250-300 million

Attrition model enables incremental home loan balance savings of $250-300 million

Download PDF

Attrition model helped client take appropriate proactive action


The Business Challenge

The client identified four key reasons why their Australian home loan customers closed their accounts: internal refinancing, property retention, property selling and external refinancing.

In 2013, 20% of this client’s home loan closures were due to external refinancing with another institution, amounting to nearly $6 billion (Australian).

To reverse this downward trend, the client wanted to proactively identify which customers were most at risk of refinancing their home loans with other institutions so that they could take steps to retain those accounts.


The Solution

Fractal Analytics developed an attrition solution, based on the Cox proportional hazards model, to identity which accounts were likely to externally refinance their home loans in the coming year. The model:

  • Used “survival analysis” to compute the monthly probability of attrition for each account over the next 12 months.
  • Calculated the “distance to attrition,” the time it takes for an account to close.
  • Measured the “distance to attrition” to rank order the bank’s home loan accounts to determine the urgency of contacting the customer in order to ensure retention.


The Results

Fractal’s attrition model helped our client identify which customers were most likely to refinance with another institution each month so that they could take appropriate proactive action.


The customer’s life stage, loan balance and additional interest paid were all key predictors impacting their decision to refinance with a different institution.



  • Fractal determined that targeting the top 20% of customers identified by the attrition model could retain an additional 1,000 home loan accounts.
  • This resulted in an incremental balance retention of $250-300 million (Australian) in fiscal year 2014-15.


  • Fractal Analytics determined that a Cox proportional hazards model could be used to measure a customer’s likelihood of loan attritoin.
  • Specifically, the model also determined the month-by-month probability of attrition and calculated the expected life of a home loan account.