A review of short and long term data and the impacts that variation has over time.

1. Section & Lesson #:
Pre-Requisite Lessons:
Complex Tools + Clear Teaching = Powerful Results
Variation Over Time (Short/Long Term Data)
Six Sigma-Measure – Lesson 17
A review of short and long term data and the impacts that variation has
over time.
Six Sigma-Measure #16 – Testing for Special Cause Variation
Copyright © 2011-2019 by Matthew J. Hansen. All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted by any means
(electronic, mechanical, photographic, photocopying, recording or otherwise) without prior permission in writing by the author and/or publisher.

2. Short/Long Term Data Defined
o The timeframe for the collected data can significantly affect the analysis results.
• What would happen if your data only included samples…
 …for only one day? Or only weekdays in a 7 day/wk process? Or only one shift of a 24 hr/day process?
• The implications of these examples may be obvious, but what about when it’s not so obvious?
 How do you know what’s a reasonable timeframe to include in your data?
 Before we can answer this, we must first understand the two types of variation in a process.
o Remember, there are two types of process variation: Common vs. Special.
o These different forms of process variation influence what type of data is collected.
2
Copyright © 2011-2019 by Matthew J. Hansen. All Rights Reserved. No part of this publication may be
reproduced, stored in a retrieval system, or transmitted by any means (electronic, mechanical, photographic,
photocopying, recording or otherwise) without prior permission in writing by the author and/or publisher.
Short Term Data
• Collected from sub-groups in process
• Captures only common cause variation
• Data reflects a “snapshot” in time
Long Term Data
• Collected across all sub-groups
• Captures both common/special causes
• Data reflects a full range of time.
Characteristics Common Cause Variation Special Cause Variation
Variation Type Natural & Random Unnatural & Erratic
Distribution Type Normal Non-Normal
Process Impact Generally from within the process Generally from outside the process
Examples •Poor design
•Normal wear and tear
•Poor environment (moisture, temp, etc.)
•Poor maintenance
•Power surge
•Extreme weather conditions
•System/computer malfunction
•Poor batch of raw materials

3. Impact of Variation Over Time
o Processes tend to show more variation in the long term than in the short term.
• Long term variation is made up of both
short term variation and process drift.
• The shift between short and long term
can be measured by taking samples of
both short and long term data.
• This shift of short term processes over
a long term is about 1.5σ on average.
 This is used more often when measuring
process capability. We’ll expand this
discussion in the Analyze phase tools.
o What timeframe should you use?
• It depends on the data you’re
measuring and the amount of variation
you expect in the process.
 There isn’t necessarily a “right or wrong”
amount of time to reflect in your data.
 The key is to be aware of how short term
data is less likely to have special cause
variation, but more likely to shift over
time and affect your results.
3
Drift over time
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
short term
data 1
short term
data 2
short term
data 3
short term
data 4
short term
data 5
long
term
data
Copyright © 2011-2019 by Matthew J. Hansen. All Rights Reserved. No part of this publication may be
reproduced, stored in a retrieval system, or transmitted by any means (electronic, mechanical, photographic,
photocopying, recording or otherwise) without prior permission in writing by the author and/or publisher.
“true” mean
over time

4. Practical Application
o Identify at least 2 metrics used by your organization that are continuous values and do
the following:
• Determine what may represent short vs. long term data for each metric.
 For example, how frequently is the metric data reported? If it’s daily, then perhaps just a few weeks may
represent short term and a few months may represent long term data. Or if it’s reported monthly, then
perhaps just a few months may represent short term and one year may represent long term.
• Pull enough historical data for each metric to account for at least long term data.
• Calculate the mean and standard deviation for all of the data across the long term.
• Calculate the mean and standard deviation for only about 25% of the data across a short term.
 For example, if you have 24 weekly observations, then calculate them only using the first 6 observations,
then next 6 observations, and so on.
• Compare the results between each short term sets of values and the long term values.
 How do the short term values differ between each short term set?
 How do those short term values differ from the long term values?
 Which set of data appears to reflect the “true” mean and standard deviation for the process?
Copyright © 2011-2019 by Matthew J. Hansen. All Rights Reserved. No part of this publication may be
reproduced, stored in a retrieval system, or transmitted by any means (electronic, mechanical, photographic,
photocopying, recording or otherwise) without prior permission in writing by the author and/or publisher.
4