Dear Dr. Jain,
I work for a large company, a group which provides services for Oil & Gas Companies. It has several activities including :
– Air transport
– Logistics (rig-moves, location of heavy transport)
– Civil engineering,
– Catering services
How can I implement a Demand Planning process to improve forecast accuracy, service level, cash flow and reduce costs of operations whilst insuring involvement of the sales department in our service activities? Is there any best-in class process you implement in a service company?
Thank you,
Mehdi Mostefaoiu
RedMed Group
Answer
Fundamentally, there is no difference in implementing a Demand Planning process for physical products or service related products. The key to all – whether improving service level and cash flow or reducing operational costs – is forecast accuracy. When forecasts improve, everything else will improve. The best practices in preparing forecasts is the use of a consensus process, where forecasts are first generated statistically, and then in a monthly meeting all the functions including Finance, Sales, Operations and Marketing get together and review the numbers.
Where necessary, they overlay judgement over the statistically generated forecasts. Judgemental overlay is needed because there are certain elements that have a bearing on a forecast but cannot be quantified, or certain information was not available at the time forecasts were generated. Also, at times, you look at the forecast numbers for certain products and see they don’t make any sense. Based on your experience, they would do much better or worse than what were forecasted. Here again, judgemental adjustment is needed. In so doing, make sure adjustment is not politically motivated. After actuals are in, do a postmortem of forecasts to see what worked and what didn’t, and why. This will help to improve your next forecasts.
Make sure forecasts are transparent so everyone knows what happened. The problem in forecasting generally arises from the use of wrong data, wrong assumptions and the wrong models. Sometimes, they are biased. Check each one thoroughly and see if there is an opportunity to improve them further.
I hope this helps.
Dr. Chaman Jain
St. John’s University