It's just on noon. At the local food hall, fast-food outlets are gearing up for the lunchtime rush, the bank of registers waiting to open quickly
as the queues form. Unless you're at a too-cool-for-school nightclub, no
one likes a queue - except, maybe, a rival business.
"We have a saying: two-deep is too deep," says Jeff Fisher, CEO of
Portuguese chicken restaurant operator Oporto.
"You get your rushes where everybody comes in within a two-hour window.
Once you get more than two people in a queue, the queue is too long and
they will basically walk past you and go to the next outlet."
Queuing is an inevitable part of modern life. Whether it's the chicken
shop, the bank or the call centre, it is one of the most vexing
customer-management issues. So it is no surpise that it's become the
object of a great deal of intellectual energy and management attention.
The problem is that serving customers costs money. If you had all the
money in the world, you could serve them all at once and no one would
have to wait, but that wouldn't be very profitable. The nub of the
queuing dilemma is: how long can you afford to make your customers wait
before you lose too many and your revenue walks out the door?
How long is too long?
The first man to make a formal stab at answering this question was Agner
Krarup Erlang, a Danish mathematician and engineer who worked for the
Copenhagen Telephone Company. Erlang gave birth to the modern science of
queuing theory in 1909 when he published a paper that addressed the
question of how many circuit boards and operators were needed to provide
an acceptable telephone service.
At the beginning of the 20th century, telephony presented a classic
queuing problem. Calls came in, but in unpredictable volumes. How long
callers had to wait and how many calls were successful depended on a
number of variables: how many calls were coming in at any one time, how
long they went on for, how many circuits were available to host those
calls, and how many operators there were to switch calls between
Erlang's solution provided the foundation for the science
of queuing that is now applied to problems across the economy, and with
which we are unknowingly manipulated each day as we go about our
The fundamentals of the problem are always the same, however. Whether it
is outpatients waiting to be seen at a hospital or instructions waiting
to be processed by a computer chip, customers arrive at varying rates
throughout the day, and they are served by "parallel processors" which
take a varying amount of time to serve each.
The length of a queue is determined by the speed of arrival of the
customers and the amount of time it takes to serve them. Because both
factors - arrivals and service times - are probabilistic, rarely will
the arrival of customers exactly match the ability of the operation to
service them. And figuring out how to minimise costs while maximising
the number of customers served is a complex simultaneous equation
problem that can be approached in a number of ways
Andrew Hume, chief operating officer at Australia's largest call centre
outsourcing operation, Salesforce, says cost is at the root of it all.
"You could answer every call that came in immediately if you had enough
operators. But that is often simply not commercially viable."
Salesforce's 5500 employees around the country deal with hundreds of
thousands of calls each day, many of them involving irate customers who
have had to wait, as far as they are concerned, too long to get on to
someone who can help them.
The trick is to get customers routed as efficiently as possible to
someone with the skills to help them - which is why call centres so
often rely on Interactive Voice Response (IVR) systems ("press 1 for
mortgages, press 2 for credit cards, press 3 to be driven completely
The key to good customer management, says Hume, is to ensure that when
the call is finally put through, it is to a sales agent who can be
helpful and who is trained in handling irate customers. IVR system
design is crucial for companies that offer multiple products. Who hasn't
been lost in a maze of digits, forgotten which number relates to which
option, and screamed at the handset to deliver them a human being?
Bad IVR design is a problem that is spawning its own technological
solutions. Mitel, a Canadian company, has recently been granted a patent
on technology that puts angry callers straight through to operators with
conflict resolution skills. Anyone who pushes too hard at the phone
keys, speaks too loudly or in a stressed tone, or uses swear words
indicating anger, will be immediately connected to an operator with a
Salesforce uses a workforce management system that forecasts call volume
arrivals and patterns, and matches that against the number of call
centre operators available, the number of hours they are available to
work, and their individual preferences.
"It layers all that and comes up with a brick wall of rostering times
and numbers," says Hume. "At the root of it all is Erlang."
Telephone queues are one thing, but for many organisations the dilemma
is how to manage large numbers of customers physically queuing for
service. In addition to the variables related to customer volumes,
service times, addition and attrition, these organisations need to
consider the physical attributes of the customer space, which provides
limitations or opportunities for customer service that can hinder or
help queuing issues.
Australian Customs, for example, processes tens of thousands of
passenger arrivals and departures each day. But how they do that is
determined by complex models that take into account the volume of
passenger arrivals, where they are coming from, and the shape of the
passenger arrivals and departures lobby.
"Together with the airport company, we engage external consultants to
model how we can ensure our part of the total processing is done as
efficiently as possible," says Gayle Brown, national manager, airport
operations north, for the Australian Customs Service.
As in a call centre, the most efficient way to process passengers is to
separate them as much as possible into different categories and have
them processed by specialist customer service people who focus on their
particular issues. That is why airports split customers into locals and
foreigners. Locals are processed more quickly, and those customer
service agents who are freed up after all the locals are dealt with can
then turn their attention to foreigners.
Rostering remains the most crucial piece of the puzzle. Customs changes
its rosters twice yearly to adjust for the northern summer and winter
travelling seasons. It then adjusts this roster daily to take into
account the forecast passenger numbers. The biggest challenge is the
large number of upstream variables, such as flight delays and
redirections, weather, international incidents, and fluctuating
passenger numbers that can impact on standard rostering plans.
Take a number
Nearly a decade ago, the banking sector was hit with a community
backlash against branch closures and front-line staff cuts. Growing
customer fury over queues and fees for over-the-counter transactions
eventually sparked a concerted effort behind the scenes for the banks to
improve their systems.
At the banks, long lunchtime queues to deposit cheques and withdraw cash
are being dealt with by directing customers into self-service channels -
internet banking has been a big hit, for example. But as the backlash
against branch closures attests, there are still numerous transactions
that demand the human touch. For these, banks have increasingly been
relying on technology to keep costs down.
At the National Australia Bank, managers use a software tool that
crunches centrally held data about queue lengths and waiting times -
essentially it is a rostering and management system for understanding
branch profitability. The bank's branches are managed on a
franchise-type arrangement, with local managers adjusting variables such
as rostering to manage their own productivity. But NAB's general
manager, deposits and retail transactions, Lisa Gray, says head office
lays down service standards and keeps tabs on customer volumes in each
branch by the half hour, checking that branches have the right number of
staff with the right kind of skills in place.
ANZ began installing take-a-number machines in 1999 and has since
expanded the system to 40 per cent of its branches. It also allows
managers to instantly access queuing performance at any of their
"To a certain extent you can make decisions on staffing levels
intuitively, but from an executive level the system gives me a record of
each branch and how it is tracking, and what needs improving and
changing stands out," says Louis Hawke, ANZ's managing director, retail
banking. The bank is looking to take the system into other areas of
branch performance, into teller management, enquiry counters, staff
scheduling, and to help give greater insight into what kind of
transactions happen when, as part of its branch management strategies
into the future. Just remember to take a number.
Maths or models
There are two fundamental ways to deal with a queuing problem.
Maths: You make assumptions about the underlying distribution that
governs the arrival of customers, their behaviour while they are in the
queue (will they give up and go away after they have waited for a
certain amount of time, on average, or not?), and the average amount of
time it will take to serve them. Then you solve the equations to find
out on average how long each customer will have to wait in the queue,
how much serving each customer in a timely manner is worth to you, and
therefore how many "parallel processors" you need to employ.
Simulation: In effect you build a model of your operation inside the
computer and see what happens. You don't need the complex mathematics,
but you do still need to make assumptions about the behaviour of your
customers and service agents, and you can use the model to see what
happens when you change particular variables, and play with it until you
get a feel for what the ideal service environment would be.
© Fairfax Business Media
Join the CIO New Zealand group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.