When I was first getting started in this business a good friend and colleague who knows a thing or two about statistical modelling advised me; “you must understand your demand before you try to fit a statistical model to it”. This advice has served our team well over the years.
A number of statistical supply chain forecasting tools advocate that they will automatically forecast your demand for you. This is a very enticing sales pitch; it implies that the software will do all the work for you. But before you turn your back on the task of forecasting and leave the software to do its thing, a word of caution:
- Statistics are a great tool for summarising and projecting subtle trends in market demand when there is continual sales history;
- Statistical tools are poor at predicting demand when the demand is lumpy with periods of no sales. (Examples: Project work, promotions etc.); and
- Statistics will not predict abrupt changes to demand such as a customer changing their artwork, or a customer moving production of a particular range of products offshore. By the time your statistical model is responding, your warehouse could already be full of items that particular customer will no longer take!
One of our packaging clients had invested in a supply chain forecasting software solution that ‘automatically’ adjusted its forecast algorithms to seek the best fit. The sales team were delighted. They no longer had to spend their time creating forecasts. They no longer needed to talk to the customer about emerging trends or understanding the reasons for errors in previous forecasts. They now had more time to go out and sell more product.
Upon reviewing the plant performance, we found that there had been a significant increase in obsolete stock and key customer DIFOT was below expected levels.
When we attended the demand review the dynamic was interesting. Corrective actions that were assigned to resolve the stock outs, all focused on improving the statistics. Corrective actions to resolve slow moving and obsolete stock resulted in requests for the statistical algorithms to be tweaked. The business was allocating all responsibility for correct forecasts onto the systems statistical algorithms.
When we reviewed the new business, we found that sales had remained static. Some new customers had come on, but new sales to existing customers had declined. Perhaps lack of communication with existing customers was affecting repeat business.
We continued the use of statistics, but we passed the ownership back to the key account managers.
Specifically we provided a portal where the account managers could adopt the statistical forecasts, or they could override them where they knew the statics were not correct, either way they had to choose the forecast they wanted. The ownership for slow moving and obsolete stock (SLOB) was again pushed back onto the account managers.
We coached the sales staff in conducting Business Review and Development (BRAD) reviews with their key customers to understand sales trends and prepare for future sales opportunities. These meetings were scheduled regularly for key accounts.
Information about pending artwork changes and promotions and other business changes that were identified from these BRAD reviews were utilised by the key account managers to override or correct the statistical forecasts as required.
SLOB dramatically reduced by adjusting the forecasts for known changes in products and lost work.
With increased customer contact, new business from existing customers increased.
The statistical tools continue to give a source of information to the key account managers , but responsibility is now on the account managers themselves to determine if it is correct.
- Forecasting should be owned by those who face the customers;
- Statistics are of great assistance, if you understand their limitations; and
- Sales can use forecasts to periodically talk to their customers. This builds market intelligence and seeds customer loyalty.
Tim Gray is a supply chain industry commentator and advises several businesses across APAC on supply chain systems. He is the managing director at Prophit Systems.
When I was first getting started in this business a good friend and colleague who knows a thing or two about statistical modelling advised me; “you must understand your demand before you try to fit a statistical model to it”. This advice has served our team well over the years. A number […]Read More...