Decent forecast Accuracy

[Former Member]

Hi,

When you are working with a forecasting model, in any business case, what is a decent forecast accuracy percentage that could be achieved? or is ti dependent on the business case?

Thanks.

[digambar_narkhede · Accepted solution]

Hi,

In any company, everyone has to recognize that the sales forecast is not exact but should be as accurate as you could reasonably expect for the business you are in and that, through measurement and correction of errors, the sales forecast accuracy is improving on a product by product basis. It is the job of the execution departments (planning, purchasing and production) to be able to meet the sales forecast given this expected level of accuracy. You do not need an accurate forecast to plan and control your business but you do need everyone to work to an agreed forecast that is updated regularly with any significant changes.

<b>The mathematical calculation is generally (forecast sales - actual sales) / forecast sales.</b> The reason forecast sales is taken as the divisor is that it is less likely to be zero and so give a divide by zero error.

The next important factor is the point in time to compare the forecast to the orders received to measure accuracy.

A make to stock type of company will need to measure the accuracy by part number at the effective procurement lead time. The procurement lead time will be supplier's lead time for a distribution company or cumulative lead time for a manufacturing company taking into consideration any safety stock of material, intermediate products or sub-assemblies.

A make to order company will need to measure either the forecast accuracy by part number at the cumulative lead time or by product family at the time needed to add resource, if this is longer than the cumulative lead time.

An engineer to order company will need to include design time to the cumulative lead time for the measurement point.

<b>In addition to the above, companies may choose to measure forecast accuracy by family at the time needed to create the business plan i.e. 12, 15 or 18 months.</b>

<i>If you are in a mature and stable business you should expect a reasonably accurate forecast. If the business is new and growing, the forecast will be less accurate. In either case the way to determine what you forecast accuracy should be is to start with the accuracy you have achieved in the past using the criterion above, this is the minimum starting point for forecast accuracy.</i>

Once you are measuring sales forecast accuracy, the next stage is to improve it. This is achieved by investigating any forecast errors outside the historical accuracy of the forecast

Regards ,

Digambar

[Former Member · Accepted solution]

Hi Digs,

Sorry for the late reply. Thanks for all that explanation and time. Is there a particular percentage range such as 70-80% or so for the accuracy?

Thanks.

[digambar_narkhede · Accepted solution]

Hi ,

You will not get it anywhere as documented as a range ,but yes by practice and experience one can say its range is between 60% to 75-80%.

Regards,

Digambar

[Former Member · Accepted solution]

If you are interested, I have modified a matrix from some material i was going through about supply chain and modified it a bit with forecast accuracy (the colored Cells are my inputs and are open to comments) and posted it <a href="http://spreadsheets.google.com/ccc?key=p8RZiP_6keoY286jl4cBCJQ">here</a>

[Former Member]

visu

my take on forecast accuracy -

There is nothing like a single decent forecast accuracy standard. Accuracy to start with is benchmarked against similar products in the market, accuracy over the precious periods and forecasting benchmark that a company's planning managere sets as part of the strategy. Again in many companies there are high accuracy products and low accuracy ones.

It would depend on a few things which is a combination of trends and market expectations such as

1. The kind of product. - If the product is a standard product which has been in the market for long it will have a high forecast accuracy and will follow historical trends.

2. Lifecycle of the product - forecast accuracy is not expected to be high in a product with short lifecycle such as fashion and seasonal clothing. In a product with a long lifecycle this will need to be high

3.Leadtime needed for procuring the RM and making the product- a high leadtime product should have high forecast accuracy while one which can be churned quickly can have lower accuracy since one can quickly manage to take care of unexpected fluctuations

It is possible to increase accuracy to a reasonable extent but this would mean time and effort and money for the technology and the resources. This may or may not justify the results of higher forecast accuracy (some industries run on full capacity and may not have subcontractors and when forecast overshoots max capacity, forecast accuracy does not matter)

The need for higher accuracy arises when the company knows

1. The implication of stock out on losing customers

2. The implication of high inventory holding

So it would be a good idea to qualify the products or product groups based on something similar and ask the businesses to focus on setting benchmarks for each group and work on improving forecast accuracy