DM: Re: Mobile Churn

[Peter McBurney]

Dear Mr Rassy,

We are a specialist consultancy company providing marketing advice to
telecommunications operators. Telecoms network churn is a complex, and
for the most part, poorly understood phenomenon, and  there are other
factors which could explain your findings, namely:
D.  In most western countries, new mobile subscriptions show a marked
annual seasonal pattern (possibly different from the pattern you 
mention
in Hypothesis A.), with the largest portion of sales in any year 
usually
occurring between Thanksgiving and New Year, and the lowest portion in
the next 3 months (Jan through March).  Non-western countries tend to
exhibit similar patterns based around seasonal holidays (e.g. Lunar 
New
Year in Asia, after Ramadam in Islamic countries, etc).  As well as
being different in number, customers at different times of the year 
tend
to be different in type (e.g. there are more gift buyers before
Christmas), and hence may churn off the network for different reasons.
E.  Customers may churn more readily when their contracts (if they 
have
such) come up for renewal.  Contract renewal dates may differ by month
independently of any seasonal purchase patterns (Factor D) because, 
for
instance, of the dates of income tax years; of sales tax quarters;
payment periods for sales dealer commissions; etc.
F. More or fewer customers, and customers of different types, may 
churn
at a particular time as a reaction to marketplace events - e.g. new
competitors launching networks or services; new price plans being
introduced; the launch of new handsets; new services being offered 
(e.g.
pre-pay services); the opening of new channels to market; the 
withdrawal
of any of these; changes to the quality of service of after-sales
customer care.
G. Churn also arises because of underlying economic cyles - for many
customers, mobile phone service is discretionary, and hence is 
discarded
as a recession strikes. A new tax can result in a sharp, sudden rise 
in
churn.  Again, not all customers react similarly to these stimuli.
H. Customers may stop using the phone well before they officially
"churn" off the network.  Any churn analysis needs to deal coherently
and consistently with those customers whose usage levels were zero in 
a
month, but who paid any monthly access fee.  These customers may or 
may
not be motivated by the same factors as those customers who passively
stop using the service and stop paying the monthly fee, or those
customers who call in to cancel their service.
I.  There may be a usage-age component to churn. Most networks witness
some form of "first bill shock" - some customers are surprised at how
large their first bill is, and some of these cancel the service. These
customers are qualitatively different from those who churn after 
longer
periods with the network.
J.  Finally, churn generally differs according to the size of the 
total
customer base and its rate of growth.  Fast growing networks in their
early life typically have many logistical, process and operational
elements in flux, and this impacts the customer experience, 
potentially
resulting in greater churn than in the case of older, more stable
networks.  For example, many new networks experience churn from
customers who expect network coverage to be greater than it typically 
is
in the early years of network operation.  If the total customer base
grew (or shrank) markedly between the first 3-month period and the
second, this may result in different causes of churn, and hence
different types of customers churning in the two periods.

Analysis and prediction of mobile network churn is not generally a
matter of a simple statistical analysis!

Best regards,



Peter McBurney
Redwing Consulting Ltd


Hallberg Rassy wrote:
> 
> Dear Friend
> 
> I have now a problem and would like to share with you and to 
>receive, if
> possible your opinion.
> Hereinafter the picture of situation:
> 1.I'm presently involved in a customer profiling project for a large
> mobile operator.
> 2.The goal is to set up a system able to anticipate the likelyhood 
>of
> churn of  customers
> 3.As a pilot step I extracted call records for 10000 active 
>customers
> plus 4000  churned
> 4.Using SPSS neural connection I made up a neural network based on 
>a set
> of 4000  active+4000 churned
> 5.The data was: calling patterns of july, agoust and september the
> target was:  churn/no churn situation in december
> 6.The results was promising: 90% of real churn anticipated, with a
> cut-off  probability of 80%
> 7.The same network was used on october, nov, dec. data to anticipate
> march churn  the results dropped to a terrific 11% with the same 
>cut-off
> of 80%: totally  useless
> 
> I have formulated some hypotheses
> A.The low time span (three month) is affected by seasonality
> B.The data used are not sufficient to build a reliable network
> C.The tool (SPSS Neural Connection) is not reliable
> 
> Could you give your opinion?
> Many thanks in advance
> 
> ______________________________________________________
> Get Your Private, Free Email at http://www.hotmail.com