I agree with Mike. The data is too small to forecast for December. Typically in the call volume forecasting situation, the day of the week and the month of the year are important factors. Although November daily data will give you some clues about the day of the week effect, it will miss the calendar seasonality posed by De…[Read more]
Very interesting observation.
Why do you need safety stock if you always over-forecast? By over-forecasting you are indirectly adding safety stock to the forecast. In such a case, you are better off using the Mean of the Demand over the past few months and adding a factor for the volatility into your safety stock fo…[Read more]
The weighted MAPE is a useful cross-sectional metric to understand forecast performance across SKUs, divisions etc. This is volume weighted so that the scale differences in SKU do not impact the error calculation. This is also called the PMAD (percent Mean Absolute Deviation).
When it comes to setting safety stock strategy, each SK…[Read more]
The formulas you gave measure forecast bias, not exactly forecast error and actually you did mention this is to measure the tendency to over-forecast or under-forecast.
Forecast Bias = (Actual – Forecsat)/Actual.
The numerator can be expressed either way, only the interpretation of the bias will change. In the above, if you have negative…[Read more]
If you are evaluating the model that derived the forecast, the best is to calculate the Mean Absolute Percent Error or the Mean Absolute Deviation. Since MAD weights the errors by magnitude, this can be a better diagnostic.
Calculate the Absolute deviation for each forecast first as forecast a – actual and forecast b – actual. Average all th…[Read more]
I am familiar with a number of projects on Sales and Operations Planning and the related S&OP forecast accuracy. I can provide you some improvement benchmarks that we have experienced.
For an initial understanding, review the following resource:
There are different measures for forecast accuracy depending on the purpose. Generally it is the difference between the actual observed quantity and the forecast.
The most commonly used measure is the MAPE or the mean absolute percent error. Alternate measures include Root Mean Squared Error, Mean Absolute Deviation, Tracking Signal et…[Read more]
Sorry I have not been able to get back to you on this earlier. Can you email me at [email protected] so we can set up a discussion to understand what exactly are you looking to do. We do have methodologies to separate the value of forecast accuracy on the inventory versus other effects. Although other initiatives do have an…[Read more]
Confidence Interval refers to the band covering the estimate that is the difference between the upper limit and the lower limit.
Confidence Level is the degree of confidence one has in a particular estimate and the confidence levels.
Here is a concise example:
Using 5% confidence, the interval [25,35] contains the population mean. This ac…[Read more]
Colloquially there is no difference between the arithmetic mean and the average. When people say these words, they mean the same concepts.
However, you should note that the average can be any one of the mean, median or mode and further the mean can be any one of arithmetic, geometric or harmonic.
Now your question about sample or population is…[Read more]
There is a significant correlation between forecast accuracy and inventory levels and customer service. There is quite a bit of literature available in the discipline of Demand Planning.
Although forecasting is not thought of at the tactical day-to-day planning levels, weekly and monthly forecasts at the SKU level are considered critical for a…[Read more]
Although used pretty interchangeably, technically there is a difference between the word Mean and the word Average. The word Average is a generic connotation to a representative observation in a sample or population. To make it a little easier, an Average can be either the Mean, the Median or the Mode. So it is just semantics and th…[Read more]
I am not sure what this tool is. I have heard on this discussion forum about a forecasting wizard but does not look like anybody was able to obtain one. There are several good off-the-shelf packages available for forecasting and some that come as an excel add-on as well. If you want some recommendations email me at ma…[Read more]
Thanks for copying and pasting from the source. However, I submit humbly, that it is in appropriate to quote the content without acknowledging the source. I agree it reads easier in one page but still it is important to insert the name of the source.
Now as to your point on the error, I don’t see it. May be it is semantics. The CLT is…[Read more]
There are several web references on the Central Limit theorem and sample size of 30 that are interesting. When the sample size approaches 30, we don’t have to worry about the distribution of the population since the it can be safely assumed to be normal for inference purposes.
Remember for interval estimation, the standard error is c…[Read more]
What you are talking about is the discipling of demand planning. A lot of work has been done in the CPG and high tech manufacturing sector. You may want to review the material on my website http://www.demandplanning.net.
There are number of conferences being organized by The INstitute of Business Forecasting as well as the Fo…[Read more]
- Load More