Australia Post plans big future with predictive analytics

The government-owned postal service is looking at initiating a wide range of predictive analytics projects.

Australia Post has had a taste of what predictive analytics can achieve, and, as a result, a number of additional analytics projects are currently in the works, according to Australia Post manager of business systems and development Armand Mizan.

In 2011, the government-owned postal company decided to implement IBM's SPSS predictive analytics software, which would work in conjunction with the vendor's Cognos TM1 enterprise planning offering. The software combination was used to help forecast the movement of cash flow at Australia Post's retail branches, and through its online portal.

It marked the first step in Australia Post's predictive analytics journey, and the organisation has since found great success with the new system, which has provided more visibility over cash flow, allowing for more automation of processes that had been extremely time consuming back when they were done manually, Mizan said.

By using old data along with datasets from a variety of external sources, the accuracy of the analytics model was very high, according to Mizan.

Australia Post is now evaluating a number of other ways to use predictive analytics across its organisation. This includes customer sales predictions on a daily basis, predicting profitability of products, and reducing customer churn by gaining better insight into their interactions with Australia Post.

"We are looking at customer churn in the parcels business," Mizan said at IBM's Smarter Analytics Live 2013 conference in Sydney. "As you know, mail is a declining business — parcels is really where growth is in Australia Post, so protecting that market and stopping customers churning out is very high on the priority list."

Using predictive analytics to automate the process of sending Armaguard cars down to specific retail branches to collect cash once the amount reaches a certain threshold is also on the cards, according to Mizan.

"We can then schedule collection as early as possible — there is lot of money out in the [Australia Post retail] network," Mizan said.

Another thing the company is looking at is doing social media analytics, which SPSS is well suited for.

However, Mizan made it clear that nothing is set in stone just yet, and all of these predictive analytics projects will have to be looked at carefully in order to determine the funding required and the timeframe for implementation.

"It takes time to get funding, approval, and to define business requirements," he said.

Depending on where Australia Post decides to go with predictive analytics, the company may have to invest in new hardware and software, Mizan said. In the meantime, the company will be continuously fine tuning its current predictive analytics system, feeding it more information so it can make cash-flow predictions even more accurate.