|Anil Nayar, President (Mobility), Bharti
Tele-Ventures: In telecom, life's a blur
time a customer calls to complain about network congestion, Anil
Nayar knows that she's only a drop-call away from switching over
to a rival cellular services provider. That's a big reason why Nayar,
Bharti Tele-Ventures' 51-year-old President (Mobility), doesn't
wait for complaints before starting to troubleshoot. Instead, Nayar
and his team pick up the truck load of data that they maintain on
their 7.37 lakh subscribers in Delhi circle, crunch it to find out
who's most likely to dump Bharti's AirTel in favour of a rival operator
(Delhi has four cellular operators, including the latest entrant
idea). That done, the next step is to figure out how to keep these
potential "churners" from leaving.
But does churn management, that's what this
data crunching exercise is called, work? Sure, says Nayar, pointing
out that prior to its implementation, the churn ratio at Bharti
touched a peak of 3 per cent, but now it's just a little over 2
per cent. Still, that's not really the point Nayar wants to make.
Says he: "If we hadn't done anything about it, the rate could
have gone beyond 3 per cent." Considering that the average
cost of customer acquisition in the mobile business is as high as
Rs 3,000, retention directly impacts the bottomline.
Bharti cottoned on to churn management way
back in November 1999, when in one of its quality meetings, it was
noticed that the single-biggest factor in opportunity cost (or non-conformance
in telecom-speak) was churn. Immediately, the company set about
pulling in all the data it had on its customers. The idea was to
meld discrete bits of data into an intelligent whole; something
that would betray the churner. Was it poor service, network congestion,
or ill-suited tariff plans that the customer was most complaining
about? If the probability of churn could be predicted accurately,
then not only could the glitches in service be fixed, but the bottomline
| MAPPING THE CHURN
The customer attributes considered in a
churn analysis by telcos are...
and Usage Data
...and the most commonly used historic
* Events like customer address change Source: SAS
This is a datamining technique used to predict a customer's
likelihood to cancel service or his propensity to churn. The
probability scores range from zero to one. If a customer has
a churn score of 0.73, it can be interpreted as "this customer
has a 73 per cent likelihood of canceling service."
One of the analytic techniques used to estimate customer's
value is referred to as Life Time Value (LTV). A simplistic
method for calculating LTV score is based on revenue and tenure.
The LTV score is then put into discrete categories-high value,
medium value and low value based on a set of business rules.
A number of data mining techniques like cluster analysis or
self-organising maps can be used to analytically detect segments
that exist based on patterns in customer data. For example,
a high value high risk customer. Segmentation allows companies
to prioritise their churn targets.
As Bharti started dialoguing and looking for
ways and means to make sense of the customer data, it was led to
American datamining major SAS' churn management customer relationship
management (CRM) solution. Its software tools allow the user to
sift through enormous quantities of data that the business generates
to find hidden patterns and trends that will help in customer retention.
Explains Gourish Hosangady, CEO and Managing Director, SAS India:
"The online application enables a company to build a 360-degree
view of its customers, and create customer-specific strategies for
Customer retention, however, is only one part
of churn management. It also makes the overall organisation much
more effective by identifying potential problems and opportunities.
For example, in one of its markets in West Delhi, Bharti was able
to optimise its coverage by studying customer complaints and usage
behaviour. In another busy commercial market, it was prompted to
set up extra-powerful transmitters because an analysis of the complaints
revealed that there were more users operating out of basement offices.
The end result of such analysis is that it allows Bharti to manage
its network investment much more effectively.
So how do the churn tools work? It starts with
a search database, where information is stored in a structured manner.
Data mining software pulls together all the raw data in whatever
form it is held into one system (also called data warehouse), and
combs through it using artificial intelligence techniques or complex
mathematical models. The SAS datamining techniques predict a customer's
likelihood of cancellation or switchover by scoring them on a scale
of 0 to 1. If a customer scores 0.73 it means there's a 73 per cent
chance of her churning. Ergo, the lower the score, the more contented
the customer. Once you know the scores, it is easy to figure out
which customers to go after first, or which customers (like defaulters)
to let go.
Bharti did not stop at churner identification.
It went a step ahead and used data warehousing tool to launch new
products. For example, when statistics showed that a number of pre-paid
subscribers in Delhi were not locals, but business visitors who
subscribed to AirTel back in their hometowns, Bharti launched regional
roaming for pre-paid subscribers. Says Nayar: "Data is a powerful
resource, and it is up to you to find business insights in it."
From Retention To Cross-Selling
Another good thing about data is that every
kind of business generates tonnes of it. And across industries there
is a farily common denominator in terms of CRM-happy customers.
Therefore, be it services, manufacturing, or retail, all industries
can deploy CRM tools like churn management. Agrees Hosangady: "The
common theme for companies irrespective of their industries is to
identify trends and patterns about their customers and serve them
Churn in Banking
|StanChart: relationship banking
If you had 2.5 million customers,
competitors who routinely poach your customers, and products
more or less generic, how do you keep your customers from
straying? As the bank in question, Standard Chartered Grindlays,
discovered, by getting inside the mind of your customer. That
means understanding the customer segments, assessing and maximising
lifetime value of each customer, modelling "what-if"
scenarios, calculating customer risks, and designing effective
marketing campaigns. What it boils down to is turning a mountain-load
of data into intelligence. For instance, by analysing the
mix of products a customer purchases, StanChart gets to know
what other products to sell to her and when. In fact, the
Diva credit card for women was the product of one such exercises.
Says Sedjwick Joseph, Head of Business Intelligence Unit,
StanChart: "By deepening our relationship with the customer
and adding value to the products and services we offer, we
make sure she does not migrate to competition." Makes
sense, since losing a customer hurts two ways: One, the bank
has to spend on acquiring a new customer and, two, that's
bad for the brand.
Churn in Retail
How do you manage churn when there
is no contract to start with? That's the question the Mumbai-based
upmarket retailer Ravissant faced when it wanted to not just
keep its high-value customers but also sell more to them. Ravissant-it
retails exclusive fashion wear, home furnishing, sterling silver
and jewellery-decided to rely on IT. Using a customer relationship
management software from SAS India, the retailer began building
intelligence on its customers' buying habits. The data warehouse
builder provides information on a range of attributes, including
highly profitable customers, their buying history, fast moving
SKUs, inventory levels at each department, and profitability
and qualitative sales analyses. Thanks to superior customer
profiling, Ravissant is able to cross sell and "up sell"
to its most precious customers. Says Ravi Chawla, Managing Director,
Ravissant: "The outlet is now able to attend to the individual
needs of each customer and offer products as per their liking
and spending range." The result: A happy customer is a
|Ravissant: smart sales
Take, for instance, banking. With varying customer
demands and ridiculously low switching costs, banks are focusing
on building customer loyalty. One way to do that is to sell her
more and more of the bank's products. That's exactly what StanChart
is doing with its churn management and datamining tools. For example,
when an auto loan customer nears the end of her tenure, her value
to the bank starts diminishing. Enter analytic capability in a CRM
solution, which enables the bank to better anticipate customer behaviour
and thereby identify new opportunities for continued value. Hypothetically
speaking, if a customer is in the age group of 30 to 45 years, then
for the same EMI is she likely to trade her old car (say, Maruti
800) for a more expensive car such as the Ford Ikon or Opel Corsa?
The CRM solution will most likely have the answer.
Selling new products to existing customers
pre-empts additional costs of advertising, marketing, administration
and all the other elements of customer acquisition. This in turn
implies high margin and a pricing advantage over competitors who
must bear these costs. Says Sedjwick George Joseph, Business Intelligence
Head, StanChart Grindlays: "For any bank that is scaling up
or adding customer to its base, it becomes imperative to use analytics
to add value to the customer during (her) lifecycle with the bank
and thereby maintain a competitive edge."
Sometimes, CRM tools can also help relaunch
products. Just ask Goodlass Nerolac Paints. In December last year,
the company revived an acrylic emulsion paint brand (Allscapes)
given up for dead soon after it was first launched in 1994. Using
sales data to analyse buying patterns, the company realised that
if it manufactured the base and did shade matching at the dealer
counter, there could be a new efficient way of selling paints. Allscapes,
therefore, was relaunched in 38 ready-to-use shades.
Similarly, last year when the demand for paint
slumped, the company analysed sales of paints in bulk, retail and
small packs for the corresponding period of June-August, 2000. The
analysis revealed that while overall paint sales were lower, bulk
packs were doing brisk business. Explains Anuj Jain, General Manager
(Marketing), GNPL: "This meant new construction work was on
while repainting requirements had fallen." Predictably, the
company increased production of bulk packs.
Be warned, though. CRM tools like churn management
are not a panacea to marketing ills. For one, the investment in
such solutions ranges from Rs 80 lakh to Rs 2.5 crore, typically
spread over three years. Besides, data is data. Unless you as the
user can innovate and make the data sweat, you are unlikely to get
the breakthroughs you would expect this kind of investment to yield.
Points out Nayar of Bharti: "Sometimes the success rate is
down to 30 per cent, but being mathematical models, we can tweak
the model a bit to increase the success rate of the tool."
That said, the goodness of CRM tools lies in the fact that they
force you to consciously look for customer risks and opportunities
not visible to other marketing mortals.