Executive Summary

  • Forecastability can be accessed by using Brightwork Explorer.
  • This article discusses this future option.

Introduction

Companies struggle with segmenting their database in a way that allows them to allocate their planning labor effectively. They also often don’t know or have a mathematical measurement for how forecastable their dataset is. The concept of forecastability is how inherently a dataset is with statistical methods. Many people think that more challenging to forecast datasets are more advanced forecasting methods when the opposite is the case. Furthermore, most companies’ forecastability is getting worse as marketing continually adds new products to the database that a company has to carry and therefore forecast.

We have developed a forecastability score. Each product location is declared either forecastable or not forecastable. The overall database is given a forecastability score base upon combining the individual product location combinations. This is something we address in the Brightwork Explorer. The main focus of Brightwork Explorer is segmenting the product location database to treat different segments differently in supply planning.

Our Approach

Our approach is to always start from a quantitative analysis and then to apply codes the product location database (which is kept externally to the system). Then the logic is understood for how each product location combination is assigned to being forecasted or not forecasted.
We would emulate a statistical forecast of the overall product location database and that provides us with quite a bit of information.

Rough Project Sequence

So a project like this goes in this sequence.
  1. Get to know the people in your company relevant to the project.
  2. Explain what we are going to do, and then obtain the data extract. (We have a data specification that we send IT, and we have sample files we can provide that give them an idea of what the data should look like, but it often takes several attempts to get the file with the correct data that we need. So we evaluate the data extract first before running any tests on it.)
  3. (After we have a correct file) — Perform forecastability analysis, and also a forecast error analysis.
  4. Create a draft assignment of each product location to either being push or pull.
  5. Present the results of the analysis, take in feedback and obtain buy in to make changes.
  6. Setup the external spreadsheet where the forecastability, and other values are maintained per for each product location combination. Its best if everyone uses something like Google Sheet, so that there is one master and everyone can see it. Later changes would be made to it as new products are added, products are removed from the system, etc..
  7. Apply the settings in the system. In ERP systems, there is a way of applying mass changes through importing a file. Normally we do this on the development or test system to make sure that things are working properly. Then after things are tested, we then roll out the changes to the production system.
  8. Provide explanation/training to the planners and IT to maintain this design.

Our Approach Versus Consulting Firms

We have subcontracted to consulting firms in the past, and we do things the opposite of consulting firms we have worked with. So if you have used consulting firms in the past, it is unlikely that our approach will be familiar. Consulting firms that we have worked with they spend a lot of time talking about methodology and they push the data analysis to the end of the project. In our view, this wastes a lot of the client’s money using inexperienced resources and having the projects managed by people who are more sales people than knowledgable in the domain of inventory management. Whenever we can work directly with a client, everything works far more smoothly, because we don’t have to debate the consulting company for how things will move forwards.
What We Normally Find on These Types of Projects
At every account we have worked on, we find inconsistencies in terms of the settings for the product locations in the system. This means that there is more inventory allocated than should be in some places and too little applied in others. Through this process we normalize the settings so that the inventory is more efficiently allocated. That way the same inventory dollars support a higher service level.

The Data Extract

  1. This overall process requires an extract of data from the systems — sales information and then inventory parameter information.
  2. We then run our analysis on that data — and provide the results — which is step 2, 3 and 4 above.
This means we work with the person who performs extracts in your IT department. As we do this, we then have meetings with planners to discuss the concepts, and take in their input and gauge their knowledge levels.

Forecast Error Measurement

Another very important part is forecast error measurement. Most companies do not effectively manage or measure forecast error. And they don’t know it. And the forecast error is an input to setting safety stock.
There is a lot of background to this, but we provide a much easier way to measure forecast error and also safety stock setting. That is its own discussion. But it is also part of what we do.