Three Most Important Metrics in a Continuous Improvement Culture

How many things do you measure in order to tell you how you are doing in business?  I have seen organizations that measure everything and anything.  Many times, what is being measured is not something the business can do anything about.  In some cases, the measures collected, reported, and reviewed, mean little to nothing to the operation and quality of the business.

However, I can tell you that in order to run a quality organization, all you really need is three things.  That is right, three very simple measures that many organizations fail to collect, or if they do, they collect at the wrong level.  Too often, managers and leaders are focused on high-level aggregate measures with little understanding of what generates the data and what the impact is if the numbers go up or go down.  The key to measuring is to do it where the work is done, on every process, and every day.

So, what are these three magical measures I speak of?

The three things that you need to measure are: 1) How much work you are doing, 2) How long it takes to do it every time, and 3) How well you do it every time.

If you measure each and every one of your processes on a daily basis using these three measures, then you can build a quality organization and a continuous improvement culture.  At all levels, employees, managers, and leaders will not only know what is happening, but they will know why and where to focus their energy to fix and improve things.

Workload Volume – how much work you are doing.

“Aggregating work volume to the monthly, quarterly, and annual levels is fine for management and leadership but does nothing for managing the work where the work is done.”

As I have said, you need to collect data based on every time you perform a process.  This means that you need to have a way to measure every instance of a process.  More than likely, you have some type of system that measures work coming in or going out of your organization.  Take a hard look at the systems associated with your processes today and see where the counts occur.  There probably is some sort of system of record that is quantifying the volume and type of work occurring within your organization.  If you do not have a way to track the workload on a daily basis, then you need to build that system first.

The importance of collecting work volume on a daily basis is so that you can look at your process by individual days.  Typically, especially in a service environment, work flows at different volumes between Monday and Friday, manufacturing normally includes seven days a week and is more steady day-to-day.  It is important in any organization to look at work from a workday perspective.  Aggregating work volume to the monthly, quarterly, and annual levels is fine for management and leadership but does nothing for managing the work where the work is done.

Once you have the data, put it into one system of record—a simple Excel spreadsheet will work, but other more advanced systems can be built or bought.  Working with Excel, start with a data sheet showing the daily volume of all your processes for a specific team or group.  From a reporting perspective, you can create three views (tabs): Annual, Monthly (Jan – Dec), and Workday (Mon – Fri and Sat/Sun if you work those days).  For each of these views, which can be tabs on your spreadsheet, you want to look at your data in three ways:

• Pareto Chart—the first view is a Pareto Chart of the total volume of each of your processes presented in that view (year, month, day). This will show you what processes are taking up the most of your time, and which are taking up the least.
• Box Plot—the second view is a Box Plot of the total volumes from each of your processes presented in that view. This will show you the variance of volume of each process and the outliers of the volume.
• Control Chart—the last view is an individual Control Chart of the total volume of each of the processes presented in that view. If the view is by year, then chart the data monthly.  If the view is month, then chart the data daily.  If the view is by day (e.g., Tuesday), then chart the data for every instance of that day.  This will show the actual variance of the data between charted points.

Having the volume of this data collected and reported in this manner provides a very powerful view into what is going on in the processes of the organization.  However, by itself, volume is a statistic that really has little value other than providing a way to forecast the potential volume of your work.

It surprises me when organizations do not have a view into their workload at this level.  Build this type of view into your daily work and teach your employees how to read the three types of charts and how to predict their workload.  If you see that Wednesdays and Thursdays are always your lowest volume days, this can make for a powerful planning tool for when to send people to training, when you can cross-train employees, when to have that monthly team-building event, etc.

Process Cycle Time – How long it takes to do your work every time.

“There are two basic ways of measuring time: 1) good operator timing and 2) physical timings.”

As said above, workload alone will tell you very little about your process.  Also, normally it is not something you normally can control, especially if your processes are the result of someone else’s process in your organization, which is often true of big companies with large end-to-end processes and value chains.

However, when you add in the second crucial item — time — you now have a very valuable combination.  Getting to process cycle time; however, can be a bit more challenging than getting to workload volume.

Some organizations might have systems that automate and track the workload and processing time of what is done.  These systems can take many forms, from home-grown to off-the-shelf.  A couple typical systems that are available to any organization are Microsoft’s InfoPath (being replaced by other Microsoft tools) or Salesforce.  If you have your processes on some type of automation like this, then harvesting the time might be relatively easy.  That is, of course, if the systems were set up to track the process step-by-step and end-to-end.  The most basic measures from the system are the start and stop time of the process, but if the steps are broken down in detail, you can get additional data.  For a start, you just need the start and stop.

In the absence of these kinds of systems, then you need to measure the process cycle time in its current form.  Every time that you make process changes, you need to re-measure the process’s cycle time.  Doing it is a bit time-consuming, but is crucial to your operation’s quality.

There are two basic ways of measuring time: 1) good operator timing and 2) physical timings.  From an industrial engineering perspective, there are several ways of timing a process, but these seem to work best for many organizations.

Good Operator Timing

This type of timing is what I often refer to as Weighted Per Accomplishment Time (WPAT).  It is a pretty simple approach to get to an 80% answer for your timings.  It also is very easy to collect, which reduces the time and man-hours required — hence, it costs less to collect.  Your timings are based on the personal estimates of people that work on the process and are good at it.  You do not want to interview people who are new or not very good and you do not want to interview your star employees, otherwise, your estimates will be skewed.  I try to interview or survey up to five employees that perform the process and then average the results.  Here are the steps on how you do this:

1. Select at least one to up to five employees that work in the process, focusing on those that are “good” at the work.
2. Ask each of them three things: 1) How long does the process take on average (Average Time), 2) How long does the process take in the worst situations (Pessimistic Time), and 3) What is the fastest time that the process can be completed (Optimistic Time)?  Always translate these times into seconds, if possible.
3. Now, ask each of them two more things: 1) Of the times that it takes the longest, out of 100% of the time, how often does this happen, and 2) Of the times the process is fastest, out of 100% of the time, how often does this happen?  The remaining percentage is the average occurrence.
4. Multiply each time (Average, Pessimistic, and Optimistic) against each of their corresponding percentages.  This will give you three fractional times.  Then, add these three times together.  This will give you a WPAT.
5. Average all the WPATs that you collected from the employees to come up with a final Average WPAT.  Over time, this is the estimated average that the process takes per accomplishment.  If you perform the process 100 times a day and multiply 100 against the WPAT, then you should have a good estimate of how much time is required to do this every day.

Some additional information can be gleaned from these timing estimates.  If you look at the Pessimistic and Optimistic times, there is a range (Range = Pessimistic – Optimistic).  This range reflects variance in the process.  If this variance is fairly wide, this should cause you to ask some questions.  If the Pessimistic or Optimistic is way off from the Average time, then this also should cause you to ask questions.

Why is there so much variance and what is happening?  Perhaps you are really looking at more than one process or there are some serious exceptions in the process that could be ripe for process improvement.

Additionally, if the estimated times per employee are significantly different, then this could lead you to ask why they are so different.  Clearly, one person is doing something different than the other and this can help you identify why.

Physical Timing

Physical timing obviously takes more time and effort and essentially costs more to accomplish, but it definitely provides a more accurate assessment of the process cycle time.  One person performing the timing can be rather time-consuming; however, you can establish a timing methodology and turn this methodology over to the process owner and make them own the timing.  This way, every time the process changes, they can retime the process following the documented steps and they feel more comfortable with the end result.  All you have to do is analyze the data when they are done.  These are the steps to conduct this style of timing:

1. Establish a specific person that will conduct all the timings for the process.  They must understand and follow the process otherwise the timings could be collected incorrectly.  The same person should collect the timings for one process.  Having more than one process timer can cause confusion and you could end up with two totally different sets of timings.
2. Identify the process start and stop time by observing the process.  It is important to start and stop the process at the same time with every timing to ensure complete accuracy.
3. Determine the number of processors that you will time.  If the process is done by one person, then obviously, you can only time one person.  If the process is done by less than 10 people, then time 2 processors.  If the process is done by 11 – 25 people, time 3 processors.  If the process is done by more than 25 people, select 4 people to time.  Try to ensure you select good operators (not new or star employees).  You, like the WPAT above, are trying to get a good operator average.
4. Outline a timing schedule based on the process.  In a normal process, which is performed constantly throughout the work day, here is an example approach below.  Regardless, you need to determine the appropriate timing approach based on when and how the process occurs.  This example uses a normal process approach.  At 9 a.m., 10 a.m., and 11 a.m., take three timings each.  After lunch, at 1 p.m., 2 p.m., and 3 p.m., take three timings each.  Try to take each timing back-to-back, but it is not imperative.  This equates to 18 timings one day.  Follow this approach for 1 week, which would result in 90 timings per processor for the week.  This is more than enough data to understand the process.  You can double the number of weeks for 180 timings per processor if you want.  If the process occurs less frequently, or only on certain days, adjust accordingly, but try to obtain at least 90 timings per processor with your approach.  Caution:  Avoid limiting collection to certain days or to certain times, unless the process warrants this approach.  For analysis, it is important to get a good picture of the entire process through all the timing.
5. Conduct the timing following the schedule developed above.  Use this schedule if the process changes to retime the process.  Time everything — do not throw out timings or stop because of exceptions (defects) in the process.  If something happens during the process that causes it to take abnormally long (e.g., process exceptions, taking a phone call, having to get supplies, answering a question, etc.) write down what happened during that timing that was unusual.  This is important data at this point; however, these situations are currently part of the process.
6. If not already done so, turn all timings into seconds.  Seconds are the easiest to work with when analyzing and provide the cleanest view of the process.
7. Now comes the time to analyze the data.  The first thing you want is the average time.  Do not throw out any of the data at this point — simply compute the average process cycle time by adding all the timings together and dividing by the number of timings.  This is not the cleanest data at this point, but it is true data.  This is the best process cycle time you can collect.  However, this is just the start of the analysis…you can look at the quickest and longest times and obtain a variance, you can examine the individual variances, you can examine multiple processors, etc.  With this basic data, you can gather a great deal of info about each and every process you perform.

When you multiply the volume of work and the time it takes, you now know the total time it takes, by day, to perform the process.  If you consider that there are traditionally 2,080 man-hours in a typical work year that means there are 173 man-hours available per month per person.  Comparing available man-hours and required process time (volume x cycle time), you now know how many full-time employees it takes to perform the work if they were to work 100% of their available time on the process.  Knowing this can lend to manpower forecasting by the day of the week and can prepare your operation for known surges in work.  Suddenly, you have very powerful data for your operation.

Additionally, like volume, you can look at this data using the same tools—Pareto Chart, Box Plot, and Control Chart.

Process Defects – How well you do it every time.

“If you find ‘exceptions’ in the process, these are probably defects in the current process or caused by an upstream process.”

Knowing the work volume and process cycle time are very powerful tools, but getting to a level of defect analysis really starts to help you improve the process on a continual basis.  Based on what you know now — volume and time — identify the processes that you want to examine first (i.e., those that have the highest volumes and highest processing times (use the Pareto analysis)).  Do the following things to better understand the process:

• Map the process.
• Observe the process.
• Determine what exceptions (aka defects) exist in the process that causes variances.
• Identify what provides value and what does not in the process.
• Look for signs of 5S needs or the deadly wastes.

If you find “exceptions” in the process, these are probably defects in the current process or caused by an upstream process.  Start collecting how many of each occur per day and if possible when they occur during the day.  Use a check sheet to collect this data.  Analyze the results with a Pareto analysis to determine which exception you want to examine first.  Look for ways to eliminate the defect and then move to the next one.  Never stop measuring to ensure the exception/defect does not appear again.

Look at what is considered waste in your process — non-value work.  Look for ways to remove this waste or transition it to other less costly areas.  Try to get your process to 100% customer value-added work.

With these three measures (volume, time, and defects) being dutifully captured in your processes, you have a great foundation for continuous improvement in your organization.  If every process owner focuses on these three things, you will be able to drill down from the highest KPIs to the lowest processes (see last month’s article) and know exactly what is working and not working in your business.

---