​Telstra using machine learning to deal with 'distressed' customers

Telstra's general manager for analytics has said the telco giant has been employing machine learning techniques in attempt to reduce the frustration of its distressed customers.

Machine learning is being employed by Telstra behind the scenes in a bid to predict and better deal with distressed customers.

Andrew Condron, Telstra's general manager of analytics, told ZDNet that the telco aims to intervene before a customer's issue causes further frustration, calling the concept a "proactive reach-out initiative".

"We know the customers in any large organisation can get frustrated when they ring in to have an issue resolved," Condron said.

"That issue they think is resolved, but it doesn't turn out to be resolved, so they ring back in again and they get passed around.

"So we're trying to really address that issue in a much more proactive sense."

Condron said the organisation attempts to predict customers that are in distress or are likely to be in distress, and contact them before they contact Telstra, effectively beating them to raising their concern.

"That's an area where we've been using machine learning techniques to predict the customers who are likely to be in distress and address that," he said. "So we actually intervene before the issue gets to a point of causing further frustration."

When it comes to using machine learning to achieve outcomes for Telstra, Condron said the telco giant also employs the technology to determine if upselling a product is an appropriate next step when dealing with a customer.

"We've applied machine learning in determining likelihood to respond to our marketing campaigns and most recently we've been starting to apply these techniques to uplift modelling which is actually a combination of two models to predict your likelihood to be converted or influenced by marketing," Condron said.

Not wasting time on customers that are not likely to buy, Condron said Telstra experiences a better return focusing its upselling resources on a select group of existing customers that are more likely to be influenced by marketing. He said that area is one where Telstra heavily employs machine learning techniques.

"We found that we get much better predictions, much better models from these techniques," he said. "Really, it generates better outcomes."

Additionally, Telstra employs machine learning to drive advocacy, which Condron said essentially means how likely a customer is to recommend the telco to others.

The former incumbent also uses machine learning to direct inbound calls to the appropriate department.

When it comes to network outages, which Telstra customers were subjected to three times over a six-week period, Condon said the organisation is beginning to look at "proactive service recovery", hoping to use machine learning capabilities to identify the indicators that lead to, or are predictive of, an outage on its network.

"These days the network is made up of millions of different devices and each of those devices has sensors on them which is collecting data and they are constantly sending alarms back, or setting off alarms. It is the combination of the alarms that is important as we're trying to understand the patterns that might predict the likelihood of that component of the network going down," he said.

Telstra's first outage of the year was on February 22, which affected prepaid and post-paid mobile services; the second on March 17 involved an hours-long national mobile data and voice outage; and the latest on March 22 was a smaller voice outage.

As a result, Telstra has been conducting a network engineering review, with plans to increase capacity of its signalling channels, add extra traffic management protection, improve capacity for its home location register, and heighten its "awareness plan".

It emerged Telstra's first outage was caused by "embarrassing human error", and saw the telco stump up free unlimited data as compensation. Another free data day, also on a Sunday, was offered to customers following the second outage.

Some Telstra customers were distressed recently by reports claiming the company had been forced by the Catholic Church to quieten its marriage equality campaign.

Initially, Telstra responded to this by saying it still supported marriage equality, but would take a step back from vocalising this until Parliament had held a plebiscite on the matter.

Telstra CEO Andrew Penn then "renewed" the telo's position on same-sex marriage earlier this week, saying it supports marriage equality and will take an "active" role in affirming this.

"Telstra supports marriage equality," Penn said.

"Last week, we advised that out of respect for the individual, our view had been that we would not add further to the debate on marriage equality ahead of a plebiscite or parliamentary debate. It is clear that rather than Telstra stepping back, we should in fact step forward and support our view for marriage equality, and so that is what we will do.

"By renewing our active position, we acknowledge that we are at equal risk of inflaming a new debate, but it is the right thing to do. It also remains very important that we continue to recognise and respect the right of the individual to hold their own view on this issue."