Using the Tools in Your Performance Excellence Toolbox: Part 7 Advanced Data Analysis

This is the seventh in a series of posts on using performance excellence tools.  In the two previous posts we looked at some basic data analysis tools including:

• Bar charts
• Histograms
• Line charts
• Pareto charts
• The 5 Whys

This post broadens the discussion of the basics of data analysis to look at an advanced tool the Cause and Effect Diagram also called the Fishbone Diagram.  One quick look at it and you see why it is called a Fishbone Diagram (Figure 1).

Figure 1 Cause and Effect Diagram

The C&E diagram allows you to delve deeper into the details of a potential cause of a problem.  The level of detail depends on how deep you want to go, how deep you need to go (at some point you may hit diminishing returns for your effort), how deep you can go (there may be limitations on data availability, etc. that constrains you), and resource constraints (time, people etc.).

There are a number of various ways to use this tool.  Most are geared towards manufacturing and assembly processes.  There is a method for transactional problems which is what will be described here.

For our demonstration the invoice team has discovered that nearly 15% of the invoices are paid after their due date incurring substantial late fees.

An easy way to think of the C&E diagram is as a detailed 5 Whys.  Just as in the 5 Whys you start by posing the problem.

Problem: 15% of invoices are paid late.

This is placed in the problem block of the diagram.

Figure 2 Problem statement

A formalized approach to labeling the causes for service based processes is the 4P’s which are policy, process, plant and people.  But you are not limited to these.  You can use them, some, all, more, or less.  For our problem we will use them.

Now you start brainstorming for the root causes.

• The basic use of brainstorming is identifying a wide range of ideas and solutions to existing or new problems

The basic rules for brainstorming are:

• Quantity, not quality is the goal so extreme ideas are welcome.  Think outside the box.  There are no bad ideas.
• No verbal or non-verbal criticism of ideas.
• No discussion of ideas.  An idea is offered and no one should be allowed to comment on it, either pro or con.
• Asking for clarification is OK, but should be focused and brief.  While there is no discussion of the idea, it is alright to ask for clarification of the idea.
• No belaboring ideas or telling of war stories (be succinct).
• Piggy backing on ideas is OK.  If someone’s comment triggers an idea related to it, offer the new idea.

The backbone of the fish points to the problem.  Off of the backbone attach each of the causes.  Now start adding in the causes for each.

Figure 3 The Causes

You continue adding causes and sub causes delving deeper until it is no longer adding value.

In our problem we can note two causes that are repeated.

• Lack of training
• Outdated policies and processes

A root cause could be management’s frugality towards spending on back office operations.

Using the Tools in Your Performance Excellence Toolbox: Part 6 Basics of Data Analysis – Continued

Using the Tools in Your Performance Excellence Toolbox: Part 6

Basics of Data Analysis – Continued

This is the sixth in a series of posts on using performance excellence tools.  It continues the discussion on the basics of data analysis.

Figure 1 shows the conversion of the data into a Histogram with a Pareto Chart output.  A short summary of the purpose of a Histogram is the data distribution shown by rectangular columns that represent the frequency of the data points collected into “bins”.  A Histograms displayed as Pareto Charts, provides a distribution in order of frequency from highest to lowest.  It also shows the percentage of each bin.

Figure 1 Histogram as Pareto Chart output

Figure 2 shows a standard Histogram where data is distributed in ascending order of the value of the “bins” vs. the percentage of the number of data points or occurrences.

 

Figure 2 Histogram

This starts to show some of the differences in more detail.  You notice that 60% of the clerks process an average of 16 or less invoices a day.  You need to look deeper into the data.

Try to answer why there is such a delta between the high and the low.  You will have to dig into the raw data more.  Look at each clerks’ data as a self-contained unit.  When you compare Bob to Liz (Figure 3) (Bob processes the most, Liz processes a little below the average) you see that Bob asks Liz for help every time he has an invoice from supplier D.  This interrupts Liz and tacks on additional time to her average processing time.  It does not add to Bob’s time since it is not counted as an interruption.

Figure 3 Comparison of April 5’s data

 But this doesn’t identify the root cause of the problem.  You now need to find out why Bob interrupts Liz when he has invoices from supplier D.  You can take to flights of fancy if you want (but it will be of little value to you in the end), e.g. maybe Bob has a thing for Liz.

You then do the same type of analysis with the other clerks to see if a pattern arises.  The results show that Bob asks for help far more than anyone else and he asks for the most help from the senior clerks who also happen to process the lower numbers of invoices.

To get to the root cause(s) use the 5 Whys.

The 5 Whys is an extremely simple tool to use that is also very effective. You state the problem and then proceed to ask why is this a problem?  Answer that and ask why is this a problem, and so on.   It is like peeling an onion.

Pose problem: Bob requires help with Supplier D invoices thus interrupting other clerks.

1. Why? Bob needs to learn the intricacy of dealing with all the suppliers.
2. Why? Bob has been targeted as a potential supervisor and is receiving OTJ training.
3. Why? Management wants potential supervisors to be taught by the SMEs.
4. Why? They think it saves them training expenses.
5. Why? No direct cash outlay for training.

Another result of your data analysis identified that Supplier D’s invoices required more time to process than the other vendors almost uniformly regardless of which clerk processed them (Figure 4).

Figure 4 Average time to process an invoice by supplier

Pose problem: Supplier D’s invoices require more time to process.

1. Why? Supplier D invoices are lengthy and more complicated than the other suppliers.
2. Why? Supplier D submits multiple separate invoices but requires payment by one check.
3. Why? It is Supplier D’s standard policy and Bob’s company’s procurement doesn’t push back on this.
4. Why? Supplier D is the sole source of this material for the company.
5. Why? Procurement doesn’t want to risk losing their sole supplier.

You have found two potential root causes.

Using the Tools in Your Performance Excellence Toolbox: Part 5 Basics of Data Analysis

This is the fifth in a series of posts on using performance excellence tools.  It covers the basics of data analysis.

You have gathered your data and you are saying, I have all this stuff, now what do I do with it? It doesn’t make any sense to me.  Your goal is to turn the data into information that will lead you to actionable opportunities for improving your process’ performance.  You are looking for the root cause of the problem.

Unlike an attorney who should never ask a question in court without already knowing the answer, if you think you know the answer, you might not be asking the right questions.  But if you want to draw the curve then plot the data to meet it, you are wasting time.  The goal of the analysis process is to start to understand what is happening within your process.

Note: It is perfectly fine to gather data to verify an assumption that you have or to make certain that your process is working within specs.

There are many analytical tools available to you.  The important thing to note is that the tools only get you part way through your analysis.  You need to interact with the data, discuss it with others.  Don’t get upset if you go down a few dead-end streets.

Using the accounts payable invoice model we started with in the earlier post we begin to analyze the results of the data gathering.  There are 10 accounts payable clerks each tracked their time for 5 weeks.  This gives us 50 days of data to work with.

What do we want to learn from the data?  Let’s say that one of your assumptions is that some clerks markedly process more invoices than others.  To confirm this, some basic questions you may want answered are how many invoices does each clerk process in a day?  How long does it take to process an invoice?  How do clerks compare to each other?

The first thing you have to do is make the data manageable.  You have 50 data sheets. Each sheet has roughly between 100 and 300 data points.  That gives you a potential of 5,000 to 15,000 data points.  You need to get them into a usable format.

Figure 1 is an aggregate of the data needed to determine each clerk’s daily average of processed invoices.

             Figure 1 Aggregate Daily Average of Processed Invoices

If you tried to plot this you would wind up with a graph that looks like Figure 2.

Figure 2 A Confusing Way to look at the Data

A better way is to get at the data is to look at each clerk’s daily numbers as an average and compare them to each other.  Figure 3 shows Bob’s average daily output and includes the median and mean times which in this case happen to be the same at 18 per day.

Figure 3 Individual Clerk’s Average Number of Invoices Processed

You calculate the daily average for each clerk.  Divide the lowest number by the highest number and put it in percentages.  This gives you 57%.  Invert that and you find that there is a 43% difference in productivity between the clerk who processes the most and the one who processes the least.   But when you calculate the Mean (16), and the Median (15) you start to see that there is more to this than a simple delta between high and low.

 

Figure 4 shows the data plot on a bar chart.

Figure 4 Bar Chart Comparing Clerks’ Output

If you convert the data into a histogram with a Pareto Chart output (Figure 5), you see there is more to the story.  So you need to dig deeper which we will do in the next post.

Figure 5 Histogram