Can Control Charts Be Used for Short Term Forecasts?

Six Sigma – iSixSigma Forums Old Forums General Can Control Charts Be Used for Short Term Forecasts?

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
  • #52326

    Adam Smith

    I’m working with a client who wants to provide short term financial forecasts. (60 to 90 days) 
    Is it feasible to use Shewhart Statistical Process Control Chart Theory and/or Confidence Intervals to provide short-term financial forecasting?
    Thanks for any help you can provide!



    I would have to say probably not. SPC charts can clear a pretty wide path for a statistically in-control process to follow. (Actually it’s the process itself that creates the path behind it, but you know what I mean.) What determines the width of the path is the inherent variation contained within that process.
    That “acceptable” common cause variation within a process may not be nearly tight enough once we start talking about tracking or predicting money flow. Remember an SPC chart is trying to help you determine what is normal for your process 95-98% of the time. That encompasses a lot of leeway. I don’t know of any financial planner that would be comfortable with as much leeway as an SPC chart builds in. And this is under the best circumstances where you only have common cause variation present. In instances where you have special cause variation, your predictability goes out the window.
    Of course, you could always build a control chart and see what you get, but my guess is it won’t be of much use in this circumstance.



    Yes it is true that the control chart will display the process as it is and yes your common cause variation can be very wide as stated by Outlier but it is a true reflection of your process. So maybe in the beginning your prediction will have a wide range but you can use your control chart to visualize the variaton in the process and start working to  understand the x’s of this wide common cause variation and when you reduce the impact of these x’s your prediction range will narrow down.



    There are many types of control charts and shewhart charts are only a few. These Shewhart charts do not require any specific distribution so sure go ahead and use them.
    EWMA may be your best choice since you can control the weighting on more recent data.
    Let us know how they work for your client.



    If your data is normally distributed, you can use a probability plot to make a prediction about future performance. Set percentile lines at y values of 10 % and 90% to get the 80% probability range.


    Jonathon Andell

    Also consider Winter’s method, which is available in Minitab and similar products.



    Would not a monte carlo-eque simulation be a better choice for such a thing?  I would be leary that the SPC would not be precise enough to adequately forecast from….but certainly time-series forecasting is possible…If using MTB, I believe you would find time-series options and explanations for potential financial modeling…..just my 2 cents….


    Andre Klener

    I would say that control charts are not the best tool.
    Depending on the type data you have, I would consider time series analysis, trend analysis, fitting data into a statistical distribution, extrapolation, or or just a plain regression analysis.
    Again, it all depends what types and formats of data you have.

Viewing 8 posts - 1 through 8 (of 8 total)

The forum ‘General’ is closed to new topics and replies.