FREE DM Review Site Registration!
Sign-up today and access DM Review on the Web!

Your FREE registration entitles you to:
FREE email newsletters

FREE access to all DM Review content

FREE access to web seminars, resource portals, our white paper library and more!

   

Business Process Improvement using Cause-and-Effect Analysis and Design of Experiments

Business process improvement is the holy grail of any company's operations. Improvements in business processes translate directly to better profits by cutting costs and increasing competitiveness at the same time. In many cases, business process improvements have accelerating cumulative effects on the company's profits. If an insurance company can underwrite policies or settle claims faster, they can provide better service; compete better with nimbler, smaller online competitors; and cut costs, which again help them compete better!

Unfortunately, any company has a limited amount of money to spend on business process improvement and that budget has to compete with many other priorities within the company. If the operations person had the choice, they would put in new hardware, software, applications; hire better qualified people; provide them more training; and have a better working environment. However, practical considerations always force companies to pick and choose the monies they can spend on the right priorities. Now the question is - what are the right priorities? How do you know that the training course on people skills is actually making a difference in customer satisfaction? How do you know that the expensive CRM system you are considering acquiring is going to make any difference? How do you know which one to do first?

This is where a combination of process modeling with cause-and-effect analysis combined with careful design of experiments (DOE) can help a company decide how and where they can allocate the monies they have to maximize their salutary effects on the company. Companies perform design of experiments informally anyway, without realizing it. For example, a smaller group of agents may try out a new customer relationship management (CRM) software application before deciding on acquiring it. Or you may decide to send a small group of your customer service agents to a new training course to test to see if it makes any difference to the quality of service provided. A combination of cause-and-effect analysis and design of experiments will help realize a systematic and scientific approach to doing the same things you already have been doing informally.

Cause-and-effect analysis is a systematic way of generating and sorting hypotheses about possible causes of a problem. Once the root causes of problems are identified, they can be addressed rather than just the symptoms.

A DOE is a structured, organized method for determining the relationship between factors affecting a process and the output of that process. The output of the process is the dependent variable that depends upon the independent variables that determine its outcome.

Let us consider for example a customer service business process. We are trying to track down the causes for poor customer service and fix them. A simple cause and effect analysis for this could look something like this:

Figure 1: Cause-and-Effect Analysis for Customer Service

Here the root causes that determine how good or how bad the end product of customer service might be are hypothesized and sorted in to a standard 3M & P model:

Methods - Methods are the processes and procedures used by customer service to deliver its services. These could be:

  • Call Workflow - Poor customer service, real or perceived by the customer, could be an artifact of how call workflow is implemented within the organization. How many times have we complained about waiting on hold or being passed on from person to person when we call for customer service?
  • Call Assignment - How are calls assigned? Does the customer interact with the right person within the company that can solve the problem, the very first time?
  • Call Escalation - If the first customer service person we talk to cannot solve our problem, whom do we talk to next? Did that help?

Materials - In the context of customer service, these are the policies, work environment, incentive and reward structures set up for the customer service agents within the company.

  • Work Environment - Customer service is bound to be poor if the work environment of the person delivering it is poor.
  • Incentive Structure - Metrics drive behavior. If the customer service agent is measured on how fast they close calls alone (average handling time), their incentive is to close calls whether or not they have solved your problem, the customer!

Machine - In the context of customer service, these are the tools available to the service agents to do their jobs.

  • CRM Application - Most customer service agents use a CRM system these days to keep track of all of their interactions. How good customer service is depends upon how well the CRM system is set up and fulfills the precise needs of the agents when providing service.
  • Problem Knowledgebase - Many organizations use a problem knowledgebase to see if the same problem has been solved for another customer.

People - For customer service to be good, there are some skills that the agents need to have:

  • Domain Skills - A customer service person trying to resolve a computer hardware problem needs to have the particular domain knowledge to be of help.
  • Problem Solving Skills - Customer service delivered over the phone requires a rather systematic approach to problem solving, eliminating obvious causes for a problem in narrowing it down to the root causes.
  • People Skills - Perception of the quality of customer service depends upon the people skills of the agent to a large extent.

Thus we see that the quality of customer service provided could depend upon many factors under the 3M& P cause-and-effect analysis. How does this relate to design of experiments?

Good customer service depends to a large extent on the above factors, but how do we decide that spending money on training is a more prudent investment than investing in a new CRM system? This is where design of experiments gives us a way of measuring the relative efficacies of addressing one cause over another.

The Pareto Principle (80/20 rule) applies to customer service as much as any other process within a company. 80 percent of the improvement in customer service is likely to come from the 20 percent of the above-mentioned causes. The question is which 20 percent?

To address this problem, let's consider you do the following as part of your DOE exercise. The idea here is you consider the quality of customer service provided as the dependent variable and the factors identified in the cause-and-effect analysis as the independent variables. The experiments you do could be the following:

  • Selective training of a sub-group of agents: A sub-group of agents becomes the experimental group while the rest of the agents become the control group. Now if you can measure the quality of customer service in some objective way (say a comprehensive customer satisfaction survey), you could compare the results of the experimental group with that of the control group to see the extent to which a training course improves agent performance.
  • Sub-group of agents using a new CRM system: If you are implementing a pilot project of a new CRM system, have a sub-group of agents use it initially. Their quality of service compared with the rest of the agents will give you some idea of the real effects of the CRM system on agent performance.
  • Trying out a new proposed incentive structure on a sub-group of agents: Before rolling out a new proposed incentive structure to all agents, try it out on a sub-group of agents and see how it affects agent performance.
  • Trying out a new workflow or escalation process: New workflow processes or escalation processes are constantly experimented with, in companies. Performing it as a DOE exercise helps you measure the results in as scientific a way as possible.
  • Setting up of a proof of concept of a new knowledgebase system: Try out a new knowledgebase system on a few agents first and measure their performance as a DOE exercise.

The above list is a gross simplification of the kinds of information that DOE can provide. The above correlations of single factors as a determinant of quality of customer service can be analyzed using analysis of variance (ANOVA) to see how related quality of customer service is to any of the above factors. One-way ANOVA relates one of the independent variables to the dependent variable here which is quality of customer service. Sometimes in practice, it is argued that the combination of two factors is really worth more than just the two factors added up together. For example, experienced customer service managers know that good problem solving skills, combined with a powerful knowledge base, can improve quality of service dramatically. Two-way ANOVA can help consider two factors together and analyze their effects on the quality of service on the whole, with proof obtained from data collected when processes are executed.

The key in these analyses as you may have noticed is collection of data. When collecting process execution data, it is just a simple additional step to collect data that has details such as the agent, years of experience, skill levels, which CRM system was used (older or the newer one being considered), etc. When process execution data is collected this way, doing design of experiments (DOE) becomes just extracting subsets of data already collected and analyzing them.

Cause-and-effect analysis when combined with design of experiments provides a very powerful combination for business process improvement. In practice, many factors seem to be qualitatively related to the business process performance you are trying to improve. However, when it comes to measuring the precise correlation between any one factor and the process performance, you need cause-and-effect analysis to catalog the independent variables that you are trying to tweak. And you need DOE to guide you in identifying which factors are worth more than others! Cause-and-effect analysis and design of experiments (DOE) provide a sane, data directed approach to business process improvement.


Nari Kannan is the CEO of Ajira, a company that designs and develops service process management tools. Kannan has 19 years of experience in information technology and started out as a senior software engineer at Digital Equipment Corporation. He has since served variously as vice president of engineering or chief technology officer of five Silicon Valley startup companies dealing with a variety of problems in IT consulting, automotive claims processing, human resources and logistics applications. He can be reached at nkannan@ajira.com or at (925) 487-1768.

 

For more information on related topics, visit the following channels:



Industry Vendors