Strategic use of analytics can boost business; The key lies not in the quantity of data used but in using the right data and doing a trial run

Strategic use of analytics can boost business
26 Mar 2014
The key lies not in the quantity of data used but in using the right data and doing a trial run.
Within the telecommunications industry, the word “churn” is a dirty word. Churn happens when customers stop doing business with a telco (telecommunications company), often with the assumption that they have switched to a competitor. In an industry where customers have abandoned high-margin services such as voice calls and SMS texts for more cost-efficient ones such as data usage, every customer counts.
“What is common these days are churn prediction models,” says Sutowo Wong, Director of Marketing Analytics at communications agency OgilvyOne. “Telcos will try to predict within a certain time frame whether a subscriber will terminate his or her contract. The telcos will target those who are more likely to churn and give them a retention offer such as a handset discount voucher.”
As the world churns
To run such prediction models, organisations need to have a suitable analytics data set where “every row represents a customer, and every column represents an attribute of the customer, for instance gender and monthly expenditure broken down into categories, etc,” Wong tells Perspectives@SMU.
With the information, a model is built to produce a score between zero and one for every customer. A score of zero means the customer is least likely to engage in the behaviour that you are predicting, and one being most likely.
“The first instinct is usually, ‘I want to do analytics; where’s the data?’ Organisations then pull all the data they can find, which are not all valuable.”
For telcos, examining a customer’s billing history, such as declining voice or data usage or a change of address leading up to the termination of the contract, has yielded certain insights.
“Let’s say there are three mobile phone lines within a household: the wife, the husband, and the husband’s mother, for example. If the wife churns, then the husband and his mother would be four times more likely to churn than if the wife does not. In other words, if one person in the household churns, it would be four times as likely that someone else in the family does so than if nobody churned in the first place.”
Using analytics to spot trends
Churn rates are particularly relevant for businesses with subscriber-based service models, with telcos and pay TV operators being the most visible examples. This number-crunching is often labelled as analytics, which Competing on Analytics author Thomas Davenport defines as “the extensive use of data, statistical and quantitative analysis, explanatory and predictive models, and fact-based management to drive decisions and actions”.
The famous case of American retail chain Target sending baby product coupons to a pregnant teenager – they were intercepted by her furious father who thought Target had made a mistake – unveiled a pattern of purchasing behaviour by pregnant women.
Target statisticians uncovered the following findings after analysing historical buying data of women who had signed up for Target baby registries: they bought larger quantities of unscented lotion at the start of their second trimester; they load up on supplements like calcium, magnesium and zinc in the first 20 weeks of pregnancy; and they bought lots of scent-free soap and extra large bags of cotton balls, hand sanitisers and washcloths close to the delivery date.
If businesses can spot trends like Target did, there would be plenty of money to be made. However, businessmen and salespeople do not always understand how to use analytics.
“The first instinct is usually, ‘I want to do analytics; where’s the data?’,” says Wong. “Organisations then pull all the data they can find, which are not all valuable. You’ll find out the value of a data point only after you’ve done a pilot.”
Churning out profits
Wong provides an example of how to run a pilot to deliver a three-month long “proof of concept” at a telco with a churn prediction model.
“We picked out the customers who were identified by the model as being ‘high risk’ and we ran a retention campaign for them. We then ran the same retention campaign for a group of customers whom we picked out randomly. The lift (customers who responded) that we got from running the campaign on the ‘high risk’ group was 40 percent higher than the control group.”
The results produced by the pilot led to the telco management operationalising the campaign. OgilvyOne gave the telco the prediction model and the IT department was tasked with automating the collection process of data to feed it. At the end of every month, the telco runs the model with the data, and identifies which of their customers they should be targeting to retain.
On implementing analytics into business decision-making, Wong says: “We need to look at high impact areas where managers can quickly demonstrate the value of analytics. One needs to look at a business question that needs answering and those in the organisation who are willing to be early adopters, to run pilot or “proof of concept” exercises. Then they should integrate the findings into everyday processes and automate them.”
Sutowo Wong was the speaker at the Singapore Management University Centre for Marketing Excellence talk, “Solving the equation of customer decision making with data”, held on Feb 14, 2014.