Dealing with Special Cause and Systematic Variation

We now have the data we need to quantify the magnitude of random (unpredictable timing) variation for safety stock:

 

  1. Recent (current) historical daily actual time-series demand
  2. Recent (current) historical daily actual time-series lead times (or an informed estimate of average, 5% minimum and 95% maximum lead-time ranges)

Often, this data does not contain significant non-random variation, and so may be used as-is. However, some time series may contain significant special-cause and/or systematic variation.

  • Special-cause variation may result from raw-materials shortages due to uncontrollable events, planned sales promotions, or other disruptions of an irregular and atypical nature that are external to the demand or lead-time process.
  • Systematic variation is integral to the demand or lead-time process. Systematic variation may include evidence of trends, seasonality, or other regular patterns that are part of the ongoing process.

To optimize our safety-stock levels when special-cause or systematic variation is present, we must:

  1. Remove the special-cause variation, as that variation is, by definition, not part of the ongoing process. This would involve, for example, identifying and removing special-cause outliers.
  2.  Isolate the random variation of the ongoing process by accounting for the systematic variation. This may involve de-trending and/or de-seasonalizing.

This is no small task. We must employ sophisticated statistical methodology to determine if there is evidence of special-cause variation, trend, seasonality, or other systematic variation. If there is, we must use state-of-the-art techniques to deal with that variation in an appropriate way, so as to obtain a reliable safety stock analysis.

Once we have “clean” historical time-series data, we must determine the structure and magnitude, or shape and scale, of each data set’s random variation. This is the topic for another blog, but as a sneak preview – do not expect to get good results by using normal-distribution formulas or functions!

And Now – A Totally Unabashed Sales Pitch!
TopDown Lean Systems has an on-demand, billable-hour consulting services to help ensure that your safety-stock analyses are based on correct historical models of your demand and lead-time. It is critical that you address this aspect of demand and lead-time data so that your safety stock – and your inventory level – is optimal. For more information, please visit our web site.

Posted in Factors Affecting Safety Stock, Safety Stock, Safety Stock Calculations, Safety Stock Formulas | Tagged , , , , , , | Leave a comment