Philippine Christian University

Graduate School of Business & Management

Master in Business Management

MMPA Quantitative Techniques in Decision Making

A Written Report on
Waiting Line/Queuing

by:
JO ANNE V. GUEVARRA
MIRIAM LUNINGNING C. PADILLA
DBP Head Office

November 2019

Professor: MYLENE A. ARCULLO, MBA

WAITING LINE/QUEUING

Waiting line is part of every day life. Whether it is waiting in line at a cafe to buy a coffee (by

taking a number) or checking out at the cash registers (finding the quickest line), waiting in line

at the bank for a teller, or waiting at an amusement park to go on the newest ride, we spend a lot

of time waiting. We wait in lines at the movies, office dining rooms, the registrar’s office for

class registration, at the Division of Motor Vehicles, and even at the end of the year mall sales.

Think about the lines you have waited in just during the past week. How long you wait in line

depends on several factors. Your wait is a result of the number of people served before you, the

number of servers working, and the amount of time it takes to serve each individual customer.

Whether we're staring at our watches in a checkout line or tapping our fingers on hold, the time

we spend waiting for service diminishes our customer experience. Long wait and hold times also

mean lost sales, bigger service issues when they do come up, and customers who are ready to

jump to another provider at a moment's notice.

In one industry study, 58 percent of respondents reported that being on hold made them feel

frustrated. This frustration often manifests itself in a lack of patience. A different study showed

that 15 percent of callers are likely to hang up at around 40 seconds.

Wait times not only affect customer frustration and abandonment but also how much customers

are willing to spend. A study by Gad Allon of the Kellogg School of Management discovered

that in the fast food industry, each extra second of wait time reduces the amount that customers

are prepared to pay. Each second.

Whatever the scenario, longer customer waits times translate into diminished customer

experiences.
One of the most important managerial applications of random processes is the prediction of

congestion is a system, as measured by delays caused by waiting in line for a service. Customers

arriving at a bank, a checkout counter in a clothing store, a movie ticket office, a fast food drive-

through, a supermarket checkout, etc. may perceive that they are wasting their time when they

have to wait in line for service. Repeated and excessive delays may ultimately influence the

customers' shopping preferences. The customers do not have to be physically present at a place

of business to be waiting in line. Waiting lines (also called queues) can form when people use the

telephone to call a business to place a catalogue order or request customer support for problems,

they are having with computer equipment or software. It doesn't matter to the customers whether

"elevator music" is played while waiting on hold, with frequent interruptions to assure them that

their call is important. If the wait is too long, the customer may "balk," which means leave the

system entirely, in this case by hanging up the telephone. Many people today do business online

over the Internet. Here, a long waiting line is apparent when the computer responds slowly

because the server is overloaded. The order may be lost during checkout, or the customer may

choose to log off and try again later. Sometimes the customer may decide that the shopping mall

is better, after all, and the results in lost business for the electronic retailer. Sometimes the

"customers" come from within the organization itself Computers or other equipment that break

down must wait for a technician to finish repairing other items that are already queued up for

service. Delivery vehicles may have to wait to use a loading dock. Products on an assembly line

may be delayed while they are in a bin waiting for the next manufacturing operation.

Organizations design their waiting line systems by considering the consequences of having a

customer wait in line or balk, versus the costs of providing more service capacity. Queues

happen when resources are limited. In fact, queues make economic sense; no queues would
equate to costly overcapacity. Queuing theory helps in the design of balanced systems that serve

customers quickly and efficiently but do not cost too much to be sustainable. All queuing

systems are broken down into the entities queuing for an activity.

At its most elementary level, queuing theory involves the analysis of arrivals at a facility, such as

a bank or fast food restaurant, then the service requirements of that facility, e.g., tellers or

attendants.

The origin of queuing theory can be traced back to the early 1900s, found in a study of the

Copenhagen telephone exchange by Agner Krarup Erlang, a Danish engineer, statistician and,

mathematician. His work led to the Erlang theory of efficient networks and the field of telephone

network analysis.

Queuing theory, which was first developed in the early part of the last century, provides a wide

variety of analytical models that can be implemented in a spreadsheet to facilitate decision

making in this context.

Queuing Theory

Queuing theory is the mathematical study of the congestion and delays of waiting in line.

Queuing theory (or "queueing theory") examines every component of waiting in line to be

served, including the arrival process, service process, number of servers, number of system

places, and the number of customers—which might be people, data packets, cars, etc.

As a branch of operations research, queuing theory can help users make informed business

decisions on how to build efficient and cost-effective workflow systems. Real-life applications of

queuing theory cover a wide range of applications, such as how to provide faster customer

service, improve traffic flow, efficiently ship orders from a warehouse, and design of

telecommunications systems, from data networks to call centers.

Queues happen when resources are limited. In fact, queues make economic sense; no queues

would equate to costly overcapacity. Queuing theory helps in the design of balanced systems that

serve customers quickly and efficiently but do not cost too much to be sustainable. All queuing

systems are broken down into the entities queuing for an activity.

At its most elementary level, queuing theory involves the analysis of arrivals at a facility, such as

a bank or fast food restaurant, then the service requirements of that facility, e.g., tellers or

attendants.

The origin of queuing theory can be traced back to the early 1900s, found in a study of the

Copenhagen telephone exchange by Agner Krarup Erlang, a Danish engineer, statistician and,

mathematician. His work led to the Erlang theory of efficient networks and the field of telephone

network analysis.

Foundations of Waiting Line Models

Even if a service system has the capacity to provide service at a faster rate than the rate at which

customers arrive, waiting lines can still form if the arrival and service processes are random.

Open the file Qsim.xls or QingText.xls in Excel to demonstrate this. As shown below, there are

cells where you can enter the average arrival rate, '"A, and the average service rate, ll· You can

also change the time that the clock will run in the simulation. Imagine that this is a small airport

with a single runway during a period of time when aircraft arrive at an average rate of 6 per hour

and can land at an average rate of 8 per hour. If we could depend on the aircraft arriving exactly

on schedule every I 0 minutes (1 /6 of an hour) and that each aircraft would land in exactly 7.5

minutes (1 /8 of an hour), there would be no waiting lines. Indeed, there would be a 2.5-minute

slack period between successive landings when the runway was not being used at all. However,

these aircraft do not arrive on a precise schedule of evenly spaced 10-minute intervals, even
though they arrive at an average rate of 6 per hour. Some aircraft take longer to land than others,

even though the average time for all of them is 7.5 minutes. This is a random process in which

events occur at random along a continuum of time or distance. A customer arriving

independently and randomly at a fast food drive through at an average rate of 20 per hour could

be described as a Poisson process with A= 20. Or, from the service point of view, repairs

completed on broken-down copy machines at an average rate of 2 per hour could be described as

a Poisson process with 11 = 2, assuming that there are always copy machines waiting to be

repaired.

A Poisson process for a time interval of I hour is simulated in the graph below. Each bar

represents an occurrence. You can view this demonstration in Excel using Poisson.xls or

QingText.xls. Initially, A is set at 5, but you may experiment by typing in different values.

A Poisson probability is the probability that, say, 4 customers will arrive in the next hour if the

average arrival rate is 5 (as shown in the graph above). It is easy and straightforward to calculate

such probabilities using Excel's =POISSON function, but it turns out that we never have to do

this when applying queuing models.

The time between events that follow a Poisson process with rate A is distributed according to the

exponential distribution, and the average time between events will be 1/A. If copying machines

can be repaired at the average rate of 2 per hour, then the average repair time is 12 hour, or 30

minutes. But some copy machines may just have a paper jam and can be repaired in 3 minutes.

Others will have broken mechanical components and may require 3 hours for the repair.
Exponentially distributed repair times with an average of 30 minutes 90 120 150 Time, minutes

Exponential probabilities can be calculated easily with Excel's =EXPONDIST function, although

it won't be necessary to do this when we are applying queuing models. However, it is important

to be familiar with the shape of an exponential distribution, because it is usually assumed by the

simplest waiting line models that time between arrivals and service time follow this distribution.

This is because we assume that both the arrival process and the service process are Poisson and,

if they are, the times between events will be exponential. Many service time distributions

encountered in practice are not distributed exponentially, however. Consider the time it takes to

withdraw cash. Tom an ATM. The average is probably about 2 minutes, which would give a

value of 1-1 of Y2 per minute. The corresponding exponential distribution would appear as

shown below.
This exponential distribution implies that shorter times are always more likely than longer times.

But Exponentially distributed ATM times with an average of 2 minutes 5 6 1 8 9 Time, minutes

10 think back to all of the times that you have used an ATM. How often did it take 10 seconds, I

second, I nanosecond, etc. All of those times would be more likely than two minutes if the

service time is exponentially distributed.

The distribution shown below (a gamma distribution) is probably closer to the truth. Note that

there is a lot less variability in this distribution than there is with the exponential, and therefore

Jess of a tendency for waiting lines to form. The amount of this variability can be adjusted by

changing a second parameter.

An important characteristic of the exponential distribution is that the mean is equal to the

standard deviation.
Kendall Notation

Gamma distributed ATM times with an average of 2 minutes 4 5 6 1 8 9 10 Time, m1nutes

Queuing models are classified using a system called Kendall notation. The general format is

*/*/s, where the first character denotes the assumptions made about the arrival process. M means

Poisson, D means deterministic (no randomness), and G means general-no assumptions are

necessary about the arrival process. The second character denotes assumptions made about the

service process. The last character, s, is the number of channels, or servers in the queuing

system. Note that if a queuing system has several channels, it is still assumed that there is only

one waiting line, similar to a post office. An M/G/2 queuing model for example, would have

Poisson arrivals, no assumptions about the service process, and 2 channels. The Simple M/M/1

Model Consider a single server that can process customers at the rate of 1 0 per hour (11 = 1 0),

facing an arrival rate of6 customers per hour (A= 6). Arrivals and services are both Poisson. This

is an example of the simplest queuing model, M/M/1. Most of the calculations are intuitive.
First, the utilization of the server is the proportion of time that the server will be busy. Arrivals at

the rate of 6 per hour to a server that can handle 10 per hour results in a utilization of6/l 0 or 0.6

or 60%.

The average dwell time (W) in the system is l/ (11- A). This formula is not intuitive. Dwell time

includes time spent waiting in line and time spent being served. For this example, it would be 1/

(11- A) = 1/(10- 6) = 1 14 hours.

This is often called waiting time, even though it includes time being served prefer to call it dwell

time.

Since the average service time is 1/11(1/10 hour in this example), the average time in the queue

(Wq) would be W- 1/11 = .25-. I = .15 hours

Once W and Wq are known, it is easy to calculate the average number of customers in the

system (L) and the average number of customers in the queue

(Lq). L = A W = 6 * Y4 = 1.5

Lq=AWq = 6*.15 = 0.9

Thus, on the average there are 1.5 customers in the system and .9 customers in the waiting line.

One might expect that if there are 1.5 customers in the system on the average, then one of them

is being served which would leave an average of 0.5 customers in the queue, instead of the 0.9

that was just calculated. However, the average number of customers being served is not I; it is

only 0.6 since the server is busy only 60% of the time.

This relationship L =A W is referred to as Little's flow equation to see why it has to be true,

consider a customer who has just entered the system. The average amount of time that it takes for

this customer to finish his business is W, which is 1 14 hours in this example. When that

customer leaves, the number of customers in the system will be the number that arrived during

that quarter hour. Since customers are arriving at the average rate, A, of 6 per hour, an average of

6 * V4 = 1.5 customers will have arrived.

These formulas only apply when the arrival rate is less than the service rate (A< J), because

otherwise the waiting line would increase without limit.

Other Waiting Line Models

When there is more than one server, the computations of the waiting line characteristics become

tedious and are no longer intuitive. You have two options for performing these calculations. One

is the workbook called Q.xls, which is available at

http://www.sunllowcr.com/~dashlev/XLAP.htm Also, an Excel add-in called Queuing ToolPak

is available at http://www.bLL<>.ualbeJ1a.ca/aingoltsson/QTP/. Queuing ToolPak provides new

Excel functions to compute waiting line characteristics.

Queuing models available in Q.xls include M/M/s and three others:

  M/M/s with finite queue. This model should be used if there is a limit to the number

of customers that can be waiting, or if customers will balk if the waiting line is too

long.

 M/M/s with finite (calling) population Use this model if there are a relatively small

number of potential customers who will ever try to use the system. This produces

different results than the standard M/M/s model because if several customers are

already tied up in the system, the overall rate at which customers arrive for service

will be smaller.

 M/G/1. This model provides results for a one-channel system with no assumptions

necessary about the service process. However, it will be necessary to supply the

average service time and the standard deviation of the service time.

To use Q.xls, select the model and then type in the parameters that are needed. The first sheet

in the workbook provides instructions. The Queuing ToolPak functions are used like any

Excel function. If you install version 2, there are some very informative help files as well.

Queuing ToolPak can be used only for the M/M/s and M/M/s with finite queue models.

M/M/s only works when the arrival rate is less than the overall service rate (A< SJ), where s

is the number of servers.

Example 1: The G. Clue Company The G. Clue Company sells top-of-the-line men's and

women's clothing through a printed catalogue and at their website. Recently, there has been

some concern that the company is losing market share to its major competitor, BBLean. Most

customers place their orders by calling a 1 800 number, which is staffed on a 24/7 basis.

During the midnight to 4 am shift, calls arrive randomly at an average rate of 14 per hour,
and the single sales representative on duty at that time requires an average of four minutes to

process a call. Thus, he can process calls at the average rate of 15 per hour. Calls that occur

while the rep is busy are automatically placed on hold, and the caller hears a message to the

effect that he or she will be served shortly. Find the utilization of the sales rep, the average

time that a customer waits before talking to the rep, the average number of customers waiting

on hold, and the probability that a customer must wait. Assume that service time is

distributed exponentially. Solution A-= 14 11 = 15 s=l Using the M/M/1 queuing model,

Utilization= lc I 11 = 14 I 15 = .933 W =I I ( 11- A)= 1/(15- 14) = I hour Wq = W- l/11 = I -

1/15 = I - .0666 = 0.933 hours 56 minutes Lq = A Wq = 14 * .933 = 13.067 customers The

customer must wait if the server is busy, so P(wait) = utilization = .933.

Example 2: Economic Analysis of the G. Clue Company Thomas J. Hilfiger, a newly hired

MBA, is the graveyard shift manager in Order and Fulfillment Operations at G. Clue. He has

reviewed the results of Example I for the midnight to 4 a.m. shift and is preparing to

recommend to his boss that two more sales representatives be added to this time slot. He

realizes that his boss will want some economic justification for his proposal, so he checks

with the Human Resources department and finds that the average hourly cost to employ a

sales rep is $11 per hour. Although there is no direct cost to G. Clue when a customer's time

is wasted while on hold, Tommy believes that if customers aren't treated as if their time is

worth at least $40 per hour, they may elect to do their shopping elsewhere. He then modifies

the MMI sheet ofQ.xls as shown below.

The formula to calculate hourly total cost of all servers in celll6 is =ServerCost*s, the cost of

a server times the number of servers. The formula to calculate hourly total cost of all waiting

customers in cell 18 is =CustCost*F8, value of a customer's time times the average number

of customers waiting in the queue. Total cost per hour in cell I I 0 is then the sum of 16 + 18.

The total cost per hour could also be calculated using Queuing ToolPak with the formula

=ServerCost*s+CustCost*Lq(lambda,mu,s) where Lq() is the Queuing ToolPak function to

calculate expected queue length. It is then straightforward to construct the following data

table and graph, showing that two sales reps would minimize total cost. This will be

Tommy's recommendation.
Example 3: Economic Analysis with the Finite Queue Model About two hours before

Tommy was to present his recommendation to his boss, he thought back to the display in

Q.xls showing the probability distribution of the number in the system when there is I server,

and wondered whether there really would ever be 30 or more customers on hold waiting to

place an order. After a couple of phone calls, he found out that intact the telephone system

could only accommodate 4 customers on hold-any others that called would receive a busy

signal.
Example 4: A Finite Calling Population Model the Package Delivery Service operates a

depot in a small city. There are nine delivery trucks that are loaded at one of two loading

docks at the depot. On the average, each truck requires I~ hours to make the deliveries,

whereupon it returns to be loaded again. Thus, the arrival rate tor each truck is I /1.5 = 2/3

arrivals per hour. All nine trucks together have an overall arrival rate of9 * 2/3 = 6 per hour.

It takes an average of 15 minutes to load each truck, so each loading dock has a service rate

of 4 per hour. The hourly cost of an idle truck is $20, and the capitalized hourly cost of a

loading dock is $15. That is, the fixed cost of constructing a loading dock is financially

equivalent to an hourly cost of $15.

The manager of the depot is studying the capacity of the depot, especially in view of

anticipated increases in traffic due to online shopping. He used the MMs sheet in Q.xls to

produce the following analysis:

This analysis would seem to justify the construction of a third loading dock. But the manager

is troubled by a couple of things. First, the probability graph shows that there could be more

than nine trucks in the depot, which is impossible. Second, on the average more than 1/3 of

the trucks (3.4) will be in the depot at any time. But if this is true, then the overall arrival rate

would be closer to 4 than 6, which would change all the other numbers in the worksheet.

The manager begins to suspect that the model he is using does not reflect the essential

relationships in this application, and he is correct. The correct model is M/M/s with finite

calling population, which is found on the finite population sheet. Copying the economic

analysis cells used on the MMs sheet to this one produces the following results:
Example 5: Service Time Does Not Follow an Exponential Distribution An executive is

recruiting a new secretary, and the Human Resources Department has provided two

candidates. On the average, it takes the first candidate (Bob) 15 minutes to complete a task,

but there is quite a bit of variation. The time follows an exponential distribution, and the

standard deviation of the task time is also 15 minutes (the exponential distribution has the

property that the standard deviation is equal to the mean). Midge, the second candidate, is not

as fast. Her average task time is 16 minutes, but she is very consistent. Her standard

deviation is only 2 minutes. The executive generates tasks at the average rate of3 per hour, or

.05 per minute.

If we enter Bob's performance data into the MMs worksheet, and then click over to the MGI

sheet, we see that both produce the same results. Since Bob's standard deviation is equal to

his average service time, this is equivalent to assuming that Bob's service time is distributed

exponentially. The formula typed in for service rate on the M/M/s sheet is =1115, so that the

rate is expressed per minute.

Midge is slower, and she will be busy a higher percentage of the time than Bob with the same

workload, but her turnaround time will be faster (48.5 minutes compared to 60 minutes for Bob).

Her in-basket, with an average of 1.625 items, will also not be as full as Bob's, with 2.25. The

real point of this example is that the queuing models that assume exponentially distributed

service times will usually overstate the degree of congestion in the system, because in most

applications the standard deviation of the service time distribution is smaller than the mean of

that distribution.

Benefits of Queuing Theory

By applying queuing theory, a business can develop more efficient queuing systems, processes,

pricing mechanisms, staffing solutions, and arrival management strategies to reduce customer

wait times and increase the number of customers that can be served.

Queuing theory as an operations management technique is commonly used to determine and

streamline staffing needs, scheduling, and inventory, which helps improve overall customer

service. It is often used by Six Sigma practitioners to improve processes.

Ways/Tips to Decreasing Customer Wait Time

The wait time that your customers deal with, however, isn’t the only thing to consider. Given

that there’s a psychological aspect to queues, the perceived wait time can differ significantly

from the actual wait time.

By making queues enjoyable, you can drastically alter the perceived wait time. What’s more, you

are showing your customers that “putting our clients first” is not just empty words.

a. Reduce wait time by going transparent. Put yourself in your customer’s shoes. In a

situation where you already have little control, it can be difficult to deal with lack of

information. How long is each customer going to take? Why has the queue not moved in
 so long? Am I stuck here? If you don’t equip your staff with the knowledge to answer

these questions for your customers, you’re in for a big eye-opener — in the form of

hundreds of customers walking out through your door. The problem, however, is

that businesses often don’t have a way of accurately answering these questions.

When managers are manually handling queues, there’s only a limited amount of

predictability to the flow of your queues. That is a very different situation compared to

an automated retail queue management system. With a QMS, it’s not possible to be

ambiguous about things like the length of the queue. The interface of Qminder is

specifically designed to provide clear and detailed information. Setting up Qminder at

your location is extremely easy. A wireless system connects the check-in counter (an iPad

or a tablet), the store backend, and a big TV display which updates information in real

time. All the information your customers need, right where they can see it.

b. Give customers something to do. The biggest source of frustration when standing in a

queue is inactivity. From the perspective of a typical queue-stander, they’re simply

wasting their time. And the longer the wait time, the stronger this feeling. For some

insight on how to deal with queues, let’s take a look at Disney. Disney is certainly no

stranger to long queues, as Disney amusement parks deal with tens of millions of visitors

every year. People are there to have a good time, and waiting isn’t what “good time”

entails. That’s why Disney has designed its queues to be not the tedious pre-fun part of

the amusement park, but an integral aspect of entertainment. A typical queue in the

Disney park has cameras and large interactive screens that allow visitors to see

themselves and play games. This simple trick makes us feel as though we’re already

being entertained — and, in a way, we are. Another clever trick used by Disney is for
 people waiting to see Monster’s Inc. Laugh Floor, a light-hearted comedy show. When

waiting to enter the theater, audience members are asked to text their favorite jokes.

Some of these jokes end up featured in the actual show, and the submitter even gets a

mention in the credits. This level of interaction makes for an unforgettable experience

that easily trumps long waiting. Obviously, not every business has the luxury of

providing amusement park-level of entertainment. But despite all that, the lesson is

simple — boredom is the biggest mood-killer. Again, put yourself in your customer’s

shoes. Would you want to wait for minutes, if not hours, without anything to occupy your

addled, bored-out-your-skull mind? For better or for worse, we’re seeing the birth of the

ADHD generation. When all people care about is distraction, anything goes. Customer is

always right? Try “Customer is always entertained!”

c. A fair queuing system. One of the most important characteristics of any queue

management method is the queuing discipline used. Simply put, the queuing discipline is

the rule used to decide who goes next in a queue. Two of the most commonly used rules

are: First in, First out. Last in, First out. Bottom line, people expect queues to be fair.

It’s not like they’re happy to be stuck waiting in line, to begin with. But when everyone

abides by the same rules, we can’t help but follow them too. Don’t believe me? Let’s

look at Exhibit A. A group of sociologists helmed by Stanley Milgram took a ticketing

counter as their base of experiments, planting several stooges to cut in line. Despite

visible irritation from the other people, the protests were mild to non-existent. However,

with every additional line-cutter, the dissent grew ever louder. This means that people

are willing to ignore outliers as long as they remain the exception. Once people start

routinely bypass the queue, all hell breaks loose. The problem of line-cutting and queue-
 jumping is a major one for unmanaged queues, but not so much for queues controlled by

automated QMS. There, the possibility of someone jumping in ahead simply doesn’t

exist. Once you adopt a queue management system for your retail, the queuing discipline

is out there for everyone to see. Social fairness is a major component of peaceful queuing

experience. All queue-standers were created equal, and it’s your job that it stays the

same.

d. Unless your customers are, for whatever reason, inmates or cattle, you need to reconsider the

use of token numbers. Paper-based queue systems were all the rage back when we’ve just

figured out a more civilized way to queue up. But now? It’s nothing short of CX

embarrassment. If you want the personality of your business to stand out, you have to make

it personal. People respond well to seeing strangers remember their name — even better when

they refer to them using said names! It is a simple gesture that goes a long way in making your

business appear people-friendly. This simple scientific fact is the backbone of Qminder’s

name-based queuing solution. All you need is a tablet and a customer eager enough to join a

queue. Once they put down their name, the system remembers them and continues to refer to

them by this name until the end of the service session. Best thing about this system? Since you

let your customers check themselves into the queue, you give them a sense of agency and

control. Also, because there is clear information available for everybody, physical queues

become obsolete. Customers are given an accurate estimate of their wait time, so they’re free

to wander about and busy themselves with other things. Thus, we take the worst part of

queues out of the equation — that nagging feeling that you’re wasting your precious time.

Improving customer service with a QMS is a surefire way to success as a business. Once you
 transform the worst part of the customer journey — queues, waiting, idleness — your first-

time visitors will quickly turn into long-time customers.

Summary

Wait time is affected by the design of the waiting line system. A waiting line system (or queuing

system) is defined by two elements: the population source of its customers and the process or

service system itself. In this supplement we examine the elements of waiting line systems and

appropriate performance measures. Performance characteristics are calculated for different

waiting line systems. We conclude with descriptions of managerial decisions related to waiting

line system design and performance.

Whatever the scenario, longer customer wait times translate into diminished customer

experiences.

Here are five tips that will work in almost any industry:

1. Find the chokepoints

Wait times often result from bottlenecks. Even the most efficient processes have chokepoints

where the process slows down.

Most business school graduates are familiar with Eliyahu Goldratt’s seminal work The Goal. In

The Goal, Goldratt puts forth his theory of constraints, which proposes that the limitations of a

system are usually created by a small number of constraints. Goldratt uses the metaphor of

Herbie, an out-of-shape child who was slowing down a group of hikers. Once the constraint – the

Herbie – was identified, its impact could be mitigated.

Look for the Herbies in your organization. Where do your processes tend to congest and how

does this affect customer wait times?

At what point in your customers’ journey (or path-to-purchase) are they experiencing the longest

wait times?

2. Focus on the front end

Why do many customers end up waiting? Because they are in the wrong place. Take a look at the

beginning of your customer’s experience.

 Does your website have clear detailed information that helps customers know the correct

channels to use or processes to follow?

 Does your team have the information they need to correctly route customers the first

time?

 Do your systems ensure that customers who are routed will be connected with a service

rep promptly?

Wait times often occur because your customer is not in front of the person who can help them.

Make sure your front-end processes are as accurate and effective as possible.

3. Create self-service opportunities

In customer experience, we often speak about the importance of the human touch. The reality,

however, is that many customers like to be self-sufficient. They often do not want to talk to us at

all or want to do so via the channels and methods they prefer.

Research by Steven Van Belleghem and SSI found that 40 percent of respondents prefer self-

service to human contact for their future contact with companies. In the digital realm, a study by

one live chat provider found that 88 percent of respondents said having live chat made the digital

experience better. If we assume that a successful live chat session prevents a phone call, then

both the customer and the organization win.

Technology is often key to creating self-service opportunities. Whether your organization uses a

better web interface, has a self-serve checkout lane, or arms its floor reps with tablet CRM

systems, technology that takes into account customer preferences creates the ability to decrease

wait times and create more enjoyable customer experiences.

In the end, the equation is fairly simple: the less customers have to wait on someone from your

organization to assist them, the less they have to wait overall.

4. Focus on customer expectations

The impact of customer wait times is inevitably dependent on context. A 15-minute wait time at

your doctor’s office will hardly register; a 15-minute wait time at a fast food restaurant will send

many customers off the rails.

 What are the expectations in your industry? If you are in the quick oil change business,

your wait times should not be two hours.

 What are the expectations created by your marketing? If your marketing campaign

promises a one-day delivery, a two-day delivery will not cut it – even if it is still faster

than your competitors.

 What are the expectations created by your past performance? We can become victims of

our own success. It doesn’t matter if your competition makes customers wait a half hour

for an appointment. It does not matter if your marketing guarantees 20 minutes or less. If

you have served the customer in less than ten minutes on the last five visits, that will be

the expectation going forward.

Do your best to make sure both you and your organization understand your customers’

expectations at every stop along the customer journey.

5. Optimize your staffing

Wait times are often the result of too few people to help. Once you have focused on optimizing

your systems and processes, you need to make sure you have the right teams in the right places at

the right time.

We’ve all had to checkout at a busy grocery store that had only one register open – it is not a

great customer experience. Here are a few quick tips for approaching the relationship between

wait times and staffing:

 Analyze your customer data to understand your peak times and to make sure you are

adequately staffed.

 Create contingency plans for when employees call out.

 Look for ways to operate as seamlessly as possible when you enter a period of being

understaffed. (Hey, in the real world, it happens!) Can you sacrifice some internal

procedures like low-value reports or maintenance in the short term while you staff up?

Look for ways to make sure your customers feel the shortage as little as possible and you

keep your wait times to a minimum.

These tips should provide a strong starting point for analyzing your organization’s service

delivery and beginning the process of decreasing customer wait times. Make sure to focus first

on the few changes that will yield the greatest results; you can refine the smaller details once you

have made progress on the key drivers.