Open Source ERP Guru

Our last post introduced an A.T.P. (Available To Promise) flash gadget under development, running on top of an Openbravo system. This post will demonstrate a new feature of the ATP viewer – predictive ATP.

Predictive ATP  is based on a prediction of future demand for a product, rather than on actual demand.

There are many ways to predict future demand. Actually, there are quite a few applications who’s sole purpose is to predict demand . One of the most simple, yet useful techniques is to use time series. By using time series, you examine all of your past demand transactions (sales orders, internal orders etc), and try to predict future demand for a specific product based on these past transactions.

For our purposes, we want to calculate past demand for a certain period, let’s say a month, or  week. So we need to come up with some kind of ‘average’ for that specific time period.

Some of the techniques used to calculate that average are:

1. Simple Average – you could just calculate (by summing all period quantities and divide that number by the number of periods) the average demand quantity for a given item at a given period, let’s say month, and predict that future sales will behave in a similar fashion. This is the most basic and simple method, but also highly unreliable.
2. Weighted Average- in most scenarios, you want to give higher weight to demand transactions that are more recent. for example, when you calculate your new  monthly average, you  could give a weight of 0.6 to the month that just ended and 0.4 to the last calculated average. That way, the impact of the last month on the average is much higher than previous months.
3. Weighted Exponential Average – a more sophisticated version of weighted average, exponential smoothing is a technique in which each month demand figures are multiplied by a factor that decreases exponentially.

The open source ATP viewer now supports all 3 methods of calculating demand. It also features a new ATP vector, that takes into account only the predicted future demand (instead of actual demand).

The way the predicted ATP works is by first calculating average monthly demand for a specific product, based on historical data, and then using that figure to reduce the amount of stock that will be available in the future for a specific time period. The predictive ATP is calculated weekly, so for any future week, the amount of available to promise stock decrease by the calculated monthly average demand, divided by 4.3 , the number of weeks in a month. Similar to the actual ATP, production or purchasing orders increase the amount of stock available to promise.

I highly recommend that you watch the first ATP viewer demo along with this one to understand the business value of the ATP viewer.

Open source ATP for Openrbavo – Recorded demo (click image below to view demo)

We plan to release the ATP viewer as open source, and also provide content that will make it work with several open source ERP systems. If you are interested in using the ATP viewer, please let us know.

You could also register to our RSS feed to stay up to date with more open source ERP news.