sigma favSix Sigma

Span – GE’s Variation Weapon

By Ron Updated on March 5th, 2008

GE is arguably one of the best examples of Six Sigma excellence today. An often heard phrase is, “Motorola invented Six Sigma and GE perfected it.”A slick “variation weapon” GE has developed is called Span. I have never worked for GE but have worked with many former GE employees who have been kind enough to fill me in on this variation busting tool.

What is Span?

Span is another measure of dispersion much like the range. When calculating the range we simply take the difference between the largest data point and smallest data point. When dealing with non normal data using the range is likely safer than using the standard deviation. Read here for an explanation on why this is.

With Span we have another option for dealing with this non normal data, and let’s face it, much of the data out there is non normal. The secret to Span is that it ignores extreme outliers which can sometimes mislead us and instead looks at the difference between the 95th percentile and 5th percentile of the data set.

Specifically, Span is calculated as follows: 95th percentile – 5th percentile

Where can we use Span?

GE uses Span for things like focusing in on how well they are doing with on time delivery performance. There is an excellent article on iSixSigma about Span. In it you will hear from people like the former 80/20 man himself, Jack Welch, comment on how GE uses Span to focus in on their customers. You can access the article here.

If you have any real life experience using Span please do share.

  1. The Avery Point Group

    April 12, 2007 - 8:25 pm

    As a former GE executive I can personally attest to the power of such variation focused tools as Span. But I must tell you that this tool has been around as part of GE’s broader toolkit of Stable Operations for almost a decade now. I was part of a core leadership team at GE Lighting back in the late 90’s that implemented both Span and other variation focused tools like the Stability Factor (worth another entry in your blog). Good to see that these great tools still have legs after almost 10 years!

    Tim Noble
    Managing Principal
    The Avery Point Group

  2. robert

    April 13, 2007 - 1:37 am

    Hi Ron

    I have to admit I’ve never actually heard of this tool, despite working on numerous lean/six sigma projects. It simply wasn’t in my toolbox. Is this not a metric which could be included in say a value stream mapping exercise?


  3. Ron Pereira

    April 13, 2007 - 8:12 am

    Thanks for the comments Tim.

    Robert – Interesting idea the more I think about it. For example, since we often state 2 numbers – Production Lead Time and Cycle Time which are normally non normal data (as most ‘time’ data is) we could easily state (assuming we had the data) that our PLT is 23 days with a Span of 5 days and our C/T is a 4 hours with a Span of 30 minutes. Hmmm…. maybe I will try this out. Thanks for the idea!

  4. Jon Miller

    April 13, 2007 - 9:20 am

    To add to the confusion, span time is sometimes used to refer to lead time.

    Span, wonderful span.

  5. Ron Pereira

    April 13, 2007 - 10:04 am

    From Wikipedia:

    In the United States, the residents of the state of Hawaii and the territories of Guam and the Commonwealth of the Northern Mariana Islands (CNMI) consume the most Spam per capita. On average, each person on Guam consumes 16 tins of Spam each year and the numbers at least equal this in the CNMI. Guam, Hawaii, and Saipan, the CNMI’s principal island, have the only McDonald’s restaurants that feature Spam on the menu. One popular Spam dish in Hawaii is Spam musubi, in which cooked Spam is combined with rice and nori seaweed and classified as onigiri.

    Oh wait… that is Spam not Span! Sorry to add to the confusion.

  6. JWDT

    April 13, 2007 - 10:39 pm

    I have shown SPAN (P5/P95) on VSM’s for operations that have a lot of variation within the processing time. Especially in the Transactional settings. Primarily used it to show where the point kaizens or string kaizens should be performed for the max impact. May find that you need to show P1/P99 for tighter processes or where standard work has eliminated a lot of the variation in some ops.

  7. Ron Pereira

    April 14, 2007 - 8:26 am

    Excellent JWDT! Thanks for sharing… I am going to give it a shot with my next VSM. Cheers!

  8. Anonymous

    April 25, 2007 - 10:26 am

    Hi Ron,
    Your method of using the normality plot is incorrect. The reference lines at 5 and 95 are drawn based on the fitted line and not the actual data. So, for data that is non-normal, the fit is bad and your results are going to be wrong. To convince yourself of this repeat the extraction using reference lines at 25 and 75 percentiles and compare to the 1st and 3rd quartile levels from a boxplot.


  9. Ron Pereira

    April 25, 2007 - 7:16 pm

    Thanks for the comment Sushil. The issue is the data I used for this graph is normal. I used Minitab to generate randon normal data. I simply wanted to show how you can calculate span with some data. In hindsight I should have used non normal data in my example and then showed how taking the difference between 95 and 5 percentile gives us span. Thanks for the feedback I hope to see you check back in to keep me honest!

  10. Dave Kershaw

    October 18, 2007 - 5:42 pm

    With respect, I think you’re both missing the point. If a data set exhibits normal distribution then stats are fine for making predictions. The beauty of span is it doesn’t try to make predictions but acts as an indicator of how good / poor performance is. Then act to reduce the span. I’d appreciate feedback though, always willing to learn so if I’m talking nonsense tell me.

  11. Ron Pereira

    October 19, 2007 - 5:50 pm

    Thanks for the comment Dave. As with all stats the key, in my opinion, is to use them to solve business problems.

    Span is great when dealing with non normal data since more traditional descriptive statistics such as standard deviation begin to lose their power when dealing with non parametric data.

    If data are normal I would not personally use Span… I would use standard deviation as my measure of dispersion.

    Would you agree?

  12. Ryan Meyers

    November 12, 2008 - 5:01 pm


  13. Sion Weaver

    January 11, 2010 - 6:15 pm

    Span is used simply to measure the fit. If you improve on an average, you can still have the product arriving too early or too late. Like Dave mentioned reduce the Span is the target. As Span is reduced the variation goes down and performance is improved.

    As GE pushed Six Sigma into management and service units, Span was introduced as a way to look at variation. Much of the data in service and management can be non normal.

    If the customer needs the delivery in one week then the Span between 2 and 17 would include deliveries too early and too late.
    2days——————–Span——————17 days

    Bring deliveries where they are always 7days or 8 days and the customer is getting what they need.

Have something to say?

Leave your comment and let's talk!

Start your Lean & Six Sigma training today.