Sunsuper’s new data scientist wowed a large audience at CMSF with a case study of how predictive modelling helped successfully allocate staffing to achieve a member campaign to roll over legacy accounts.
Kirill Eremenko, a trained mathematician with a background in analysis at Deloitte, told delegates how Sunsuper has conducted five separate targeted member SuperMatch campaigns since 2013, but had encountered problems with response rates overwhelming staff allocated to process such requests.
In March 2014 it had a 27 per cent response rate to emails sent to 55,000 members which led to a 40 day backlog processing applications. The campaigns and the backlogs were also a source of tension between marketing and the processing centre for roll overs.
Prior to launching its August 2014 campaign Eremenko analysed the response rates to the March campaign. He found requests to roll over legacy accounts peaked eight days after the message was sent, that half of all requests came in the first 13 days and 80 per cent came in the first 31 days.
Eremenko said the first impulse was to increase the number of staff from 20 to 40-50 during the 12 busiest days, but that this was not a feasible option for Sunsuper or almost any other fund.
The optimal proposal deduced from the data was to split the August campaign into five separate sends over three weeks. This only needed 11 staff to fulfill and reduced the maximum backlog for each roll over request to three days.
This method was used for the August campaign which gained a 32 per cent response rate from messages sent to 65,000 members.
When Eremenko told delegates all funds had the ability to do such analysis from their data, he was asked what software he used. He revealed that for the database he used Oracle’s SQL software and to display graphs to help visualise the campaign he used Tableau software.
