Business runs on data. In fact most companies have too much data. They collect a lot of data that’s not needed, and the data that is used contains errors. This data is usually used  to calculate averages and long term trends i.e: average output of a shift, average value of goods purchased, Sales.

One thing we do know is that all data has errors in it. Many decisions are made using data, however all data is wrong, yes that’s right all data is not accurate. It is a question of how wrong that data is.

The risk from using faulty data to make decisions is potentially huge. Measurement error occurs in decisions that effect Customers, Suppliers, and Producers.

Think this is not important? Have you noticed that Petrol Bowsers at the petrol stations have (or should have) a calibration sticker on them to indicate that the measurement error is within acceptable limits? If it doesn’t then are you getting what you paid for?

What about at the supermarket when fruit is purchased by the kg. How accurate are the scales? Have much error do they have?

How accurate are the power consumption metres that measure the volume of electricity consumed? The measurement from the metres determine your power bill each month.

Data Integrity

All is not lost as there are tools and methods that can be applied to your data to find out how much of the variability in your data is due to the measurement system and how much is due to the actual process performance itself.

We use tools that can check the integrity of your data, we call it a Measurement System Analysis. Whenever we are using data, that data is made up of two components:

  • The actual data variation.
  • The variation in the measurement system.

It is our experience that approximately 50% of the data that companies are using has unacceptable levels of error. Many companies will ignore that error  and proceed as if nothing is wrong, however, smart companies will reduce the measurement error to an acceptable risk level.

One company we were involved with was conducting a pass/fail test on the end of a Printed Circuit Board production line. Goods were either sent to the customer if the test was passed or rejected if it failed. After conducting a Measurement System Analysis we were able to determine that the gauge could not tell a good part from a bad part. in other words, the measurement system was a waste of time. The company subsequently declined to fix the machine, as the customer insisted on keeping it.  I have often wondered how many bad units were sent to the customer or how many good products had been rejected; think of the cost of all that waste.

There are three types of Measurement System Analysis (MSA). The determining factor in choosing the type of MSA to be used is the type of data that you have.

Types of MSA:

There are three types of Measurement System Analysis:

  • Continuous data MSA -data that is measured on a continuous scale for instance, weight, speed, cycle times, spend etc.
  • Attribute data MSA -Data that is count in nature and is based upon an attribute i.e.; pass or fail, for instance checking goods coming off a production line and determining if they are a pass or fail, checking a purchase order has been filled out correctly.
  • Data Audit: Data comparison between two lists of data, often between manual records and  computer system records and checking how many records match. This is the simplest MSA to conduct and is usually very quick to run.

When looking at the error in continuous and attribute type data we are looking at two areas of potential error:

  • Gauge calibration and selection.
  • Gauge Variability.

Gauge calibration and selection is ensuring that the gauges that we have are ‘fit for purpose’, or they can measure what they need to measure with an acceptable error.

We are checking for:

  • Measurement stability; is the measurement device stable over time
  • Measurement bias; is the measurement device reading high or low consistently
  • Measurement linearity; Does the measurement device have the same measurement error across its full operating range.

Gauge variability on the other hand looks at two area:

  • Repeatability; the ability of the operator taking the measurement to get the same measurement reading when taking the measurement more than once.
  • Reproducibility; the ability of more than one operator to use the same measurement device and get the same measurement.


We have tried to highlight the importance of making decisions based on data, however probably half of the data you are using is faulty.

In summary, 2 reasons you maybe making decisions based on faulty data is:

  1. You do not understand that all data has error, it is just a question of how much error there is and if that level of risk is okay. 
  2. You have never measured how much error there is in the data. Error types can be determined through Measurement System Analysis, and appropriate action can be taken.

Look to our You tube for an upcoming screen cast on Measurement System Analysis where we will go into more detail about how to determine the error in our data and the appropriate action that needs to be taken to reduce the risk to appropriate levels. We also have our upcoming course ‘Jump Start your Process Productivity‘ which will go into the theory and more importantly practical application of MSA techniques.