To outsource forecasting or not?
A growing number of consultancies are offering to take care of companies’ forecasting and planning processes. That makes sense, say advocates of this approach, because the specific knowledge and expertise needed – in particular for forecasting – is in increasingly short supply on the labour market. That’s a bad move, say opponents of the idea, because outsourcing leaves a significant dent in the quality of the forecasting process as a whole. To compound the dilemma even further, companies who outsource risk losing valuable knowledge which is not so easy to reacquire.
To manufacture its semi-conductors, ASML uses machinery in production facilities all over the world. And yet every day, the 2013 winner of the Dutch Logistics Award receives a data file from each and every machine which details not only every single error message but also other values indicating the state of that machine. Per day, this amounts to several thousand items of data such as temperature, voltage and pressure for all the various components.
The Dinalog ‘ProSeLo’ (Proactive Service Logistics) study explored whether this data made it possible to predict when a technical problem would occur. The answer was ‘Yes’. Using data-mining techniques, researchers succeeded – albeit for a very small dataset – in predicting technical problems better than using existing models based on the extensive experience of the ASML engineers who actually built the machines.
This example illustrates how ever-better data-analysis techniques are opening up opportunities for generating more accurate and more reliable forecasts – ones which do not even require specific understanding of the data; it suffices to identify patterns and correlations within the data. In the ASML case, for example, the ProSeLo researcher had no idea which machines the data referred to or what they looked like.
This technique for predicting technical errors can also be used to forecast sales. Based on historical data, perhaps complemented by details of seasonal fluctuations and current trends, it is possible to generate reliable forecasts for large sections of the assortment without any prior knowledge of the products or markets. That enables companies to outsource processes such as forecasting and planning.
One example is Bosal, manufacturer of automotive products, which has outsourced its forecasting to the consultancy EyeOn. Using the available data, EyeOn’s forecasting specialists prepare a forecast within a cloud solution so that it is accessible for all of the company’s sales and manufacturing facilities. They then have the opportunity to validate the forecast and, if necessary, adapt it in line with their own insights. “Making a professional forecast is a specialist skill. We could of course invest in learning to do it ourselves, but it would take us years to reach the same level as EyeOn,” says Jon Kuiper, group sourcing & supply director at Bosal.
In Bosal’s case the forecasts are mainly for mature products, but EyeOn says that it can also forecast promotions or product launches. One important constraint is that the forecasts are based on data alone. “This means that our expertise lies in data-analysis techniques and statistics rather than in the products and markets themselves. Companies such as Bosal do not regard statistical forecasting as part of their core business, which is why they outsource it. This allows them to focus on the parts of the forecasting process where they can add value, such as by enriching our forecasts with market intelligence,” explains Freek Aertsen, managing partner at EyeOn.