You can generally split organizations into two categories; those who don’t forecast and those who do it poorly. And those who don’t forecast look at those who do it poorly and think that good forecasting is merely impossible. The main reason for poor forecasting is the naive belief that you can just take out-of-the-box software, press start, and get good forecasts.
The explanation is simple. Generally, an organisation’s product mix includes their main products with stable sales, products with seasonal sales, new products, products with high volume, others with less volume etc. Different products will always require different strategies, i.e. different forecasting models and tailored algorithmic methods.
Out-of-the-box software has, nevertheless, gained some momentum and one can only imagine that the reason is that organizations were lacking better alternatives. Simple time-series modeling can certainly be better than guesswork in Excel.
In recent years the evolution of sophisticated statistical modeling and AI has certainly created the possibility of forecasting better, and organizations are taking notice. Companies are outsourcing their modeling to analytical agencies and are improving their forecasting more than they could imagine some years ago.
Demand forecasting impacts almost every element of the business, including financial planning, inventory management, production planning, distribution, marketing, and customer management and should, therefore, have major financial consequences. The uncertainty organizations are facing under the current economic conditions are forcing business leaders to take further measures in cost control and profitability and reduce unnecessary risks by all means. On top of the list is better forecasting, obviously, which, with better accuracy, will have a direct impact on the bottom line. But how much money can organizations save (make) by forecasting better?
A real world example of a major beverage production company in Spain, tended to produce 9% more than it sold, i.e. safety stock that was supposed to compensate for the risk of running out of stock. Obviously, adding up, this sort of practice turned out to be very expensive guesswork in the long run. After carefully analysing the problem they identified three areas for improvement:
Their demand forecasting, which was somewhat correct for high-volume products, was considerably off for the largest part of their products sold.
Different factors within sales management and planning were not very helpful, which created a tendency to round up the size of incoming orders.
When interviewed, sales representatives admitted that the current forecasting software was unreliable which made them use their intuition to plan sales.
Those challenges had been easily overlooked for years, even though they were having considerable impact on the bottom line. However, the solution was rather simple: Better sales forecasts.
By addressing the problem directly with tailored demand forecast modeling for their products, they improved forecast accuracy by more than 35%. The forecasts were fed directly into their current ERP systems, and a new interface allowed the sales reps to visualize their forecasted products in more detail than before.
As the company improved their forecasting considerably, they strived to estimate the connection between their forecast accuracy and the bottom line; i.e. how much impact does forecasting error have on operational cost and sales & marketing cost. They identified operational cost in the following situations:
Expediting raw material creates higher inbound raw material cost
Lost opportunities to procure raw materials at a favourable time
Long-term production planning impossible and short-term quick-fixes a usual practice
Changing production schedules increase cost due to equipment alignment and personnel planning
Producing the wrong product leads to increase in inventory cost and levels
Shipping to the wrong location incurs higher logistics costs for storing the product in the wrong location, and to ship the product back to the right location
Discounting price to get products sold
And sales & marketing cost in these situations:
Inefficient use of company sales and marketing resources
Improper allocation of company resources across products
Reduced or even lost revenue
Lost sales opportunities
When it comes to forecasting error, the impact is slightly different whether the company is over-forecasting or under-forecasting. The general assumption is that under-forecasting is more expensive than over-forecasting, hence the 9% over production.
As the figure above shows, the basic implications of over-forecasting is tying up of financial resources in excessive inventory, while the consequences of under-forecasting is the potential loss of profit resulting from an increase in cost and loss in sales.
Keeping in mind the two basic implications of over-forecasting and under-forecasting, a method for monetizing forecast error can be developed. Primarily, the total stock keeping unit volume per month needs to be identified, and then this volume is multiplied by 1% to suggest the unit volume associated with 1% forecast error. Now, the unit volume associated with a 1% forecast error is multiplied by the average product cost per unit. This basic calculation gives you the cost of inventory associated with 1% forecast error due to over-forecasting, with the formula being as follows:
A similar approach can be taken for under-forecasting situations where the unit volume associated with 1% forecast error is multiplied by the profit margin per unit. This suggests a potential lost sales figure associated with 1% forecast error with the formula being as follows:
Each number estimate for over-forecasting and under-forecasting represents a potential cost range for 1% forecast error. Multiplying this number with the respective monthly mean absolute percent error (MAPE) can provide the potential total cost of forecast error incurred by the organization per month.
A valid criticism of this methodology is that the above assumes excessive inventory becomes obsolete. This may or may not be the case, because excessive inventory would have some value, although a reduced value. This reduction can be incorporated into the forecast error cost calculations. It’s also a common pitfall of thinking solely in terms of MAPE, which is a common evaluation of forecast accuracy, whereas the standard deviation of MAPE over time would tell you how much your forecast accuracy fluctuates.
Using the above methodology for the Spanish beverage company, which produces around 7,5 million cans of their beverage per month, are losing close to EUR 4,0 million with ONLY 1% forecast error. The forecast improvement of over 35% can therefore be translated into a serious amount that has a significant impact on the company’s profitability.
Obviously, there can be more factors to include and the calculations can certainly be made more complicated. Therefore, this approach offers a straightforward method to simplify calculation efforts enormously and provide a ballpark estimate. What this method shows is that the cost of forecast error can indeed be quantified and in relatively straightforward fashion. Results using this method further exemplify that the cost of forecast error across organizations is not trivial.
While clearly not a panacea and arguably an oversimplification of estimating the linkage between forecast accuracy and the bottom line, the method can quickly identify ballpark estimates on which to discuss what the “true” impact might be.
The Spanish beverage company employed this method to stress the importance of the sales forecasting endeavour because of the financial impact. The attention of management across departments in the organization was certainly stimulated when the problem had been monetized. Additionally, this exercise justified the cost of hiring an analytical agency to create advanced forecasting models customized to different products in order to attain the highest forecast accuracy possible.
The linkage between forecasting and the bottom line is an emerging topic and a top-of-mind issue for senior management. To use an out-of-the-box approach to forecast for the sole purpose of forecasting, despite high forecast error, is not a small issue. Forecasting poorly is very expensive for organizations, and forecasting better will save millions of euros.
