Perfecting the art of demand forecasting in volatile market

Improved demand forecast accuracy with next generation analytic solutions are ensuring on shelf availability of product with reduced inventory, says Alagu Balaraman and Sujit Sahu of CGN & Associates

The essence of demand forecasting is aptly captured by these words by anonymous, “A good forecaster is not smarter than everyone else, he merely has his ignorance better organised”. The one thing that has definitely changed now are intelligent algorithms helping in managing the ignorance or in more technical terms reducing the forecast error. 

A study by Teraa Technology suggests that new methods of demand forecasting have reduced the forecasting error by 37 percent at SKU level. Demand forecasting tools are widely appreciated and adopted over the past several years across the industries. The companies which have well mastered this tool are enjoying better returns over the competitors in terms of improved sales, low inventory, and higher profit margin. 

Aggressive competition and demanding customers are pushing companies for adding new products on regular intervals. Distinct products available for sale have nearly tripled in last five years. With shrinking product life cycle, increasing numbers of SKU and wide geographic operations, there is high stress on the value chain to ensure on-shelf availability. A report by Efficient Consumer Response group says that stock out in Asia Pacific region is as high as 18 percent. The same survey also highlights that forecasting error and poor ordering are the two major reason of stock out. This by a rule of thumb would mean a loss of sales of around 9 percent. 

A stock out also implies further hidden costs to the value chain such as obsolescence and returning excess inventory to suppliers let alone the margin loss due to 9 percent volume loss. A recent survey done by Supply Chain Digest shows that 61 percent companies see improving forecast accuracy as a significant area of improvement. Survey also points out that an average 5 percent increase in forecast accuracy increases the profit margin by 2 percent. 

But there are several challenges in implementing demand forecasting tools. The foremost challenge is to justify the cost involved in infrastructure investment. Off-the-shelf demand forecasting tools are very expensive and require significant consulting effort and time to implement. Skill availability within the organisation to model the business process, market variables, together is another challenge. Besides, one has to deal with other challenges like data availability, data accuracies, and data bias. These challenges make demand forecasting a tough problem to solve. It is no wonder that most organisations have continued with their traditional methods of bottom-up demand forecasting where a certain percent is added to actual sales for the same period last year as forecasts. 

Solutions to the challenge have broadly fallen into two components, either implementation of software with best of breed forecasting algorithms or a consulting effort with excel based tools as one-time investments with the hope that organisation will be able to improvise it further once consultants have left. Either of these approaches has failed. Software implementations without policy level changes have failed to give results. And consulting effort focussed on policy changes with the less sophisticated computing power of excel has users settle for sub-optimal solutions. But the biggest challenge has been in finding the right talent within the business to solve the problem of scientific forecasting.

Alagu Balaraman (left) and Sujit Sahu of CGN & Associates

If one could solve the problem of right talent at affordable price points without worrying about constant up gradation of software tools and methods, one would have found the Holy Grail. Feasible answers would require different thinking models. Cloud computing has lowered the cost of computing significantly with world-class data security. Open source platforms and open source software adoption has also opened the avenues for low cost solutions. And to add to it, the current push by the tech giants like Oracle, SAP, Salesforce have made ON-Cloud platforms an accepted common lexicon among the business community where having your business data secured in cloud based infrastructure is increasingly gaining acceptance. Once data security is taken off the discussion table, businesses can then attempt to solve the issue of software, algorithms, and processes to overcome the twin challenge of tools and talent. 

Talking about tools and talent, traditionally sales force has been responsible for providing forecasts in most Indian businesses. Everyone acknowledges that they are not best equipped to run mathematical programs to create scientific forecasts. But one has to just look around multinationals that use India as a base to develop world class algorithms to conduct sales analytics, fraud detection and other such sophisticated applications. So India has the talent pool to solve this problem, but these resources are expensive and difficult to retain.

This is where businesses should see opportunities for developing new services model to solve twin challenges of talent and tools. This is not about cost reduction, but delivering tremendous business value with high category of skills in demand forecasting at affordable cost. There have been new innovative services companies that have experimented with newer models to solve the problem of demand forecasting. Results are very encouraging. A handful of Indian businesses that are already using such services models have got their ROI in less than 6 months. All results show that statistically generated forecast accuracies have significantly outperformed forecasts generated by sales based on empirical rules and intuition.   

Advantages for customers who have experimented with newer services model have been several. To begin with, investing in demand forecasting tools is no longer a laboured task of justifying returns. Sales team continues to be accountable for forecasts but they now get additional forecast input to decide the future course of action. They retain the right to override. Freedom to do so has allowed sales team to provide better inputs from markets to improve the statistical forecasts and take ownership on data quality. 

The other big advantage has been driving the rigor of the demand planning process in term of timely data inputs, consolidation of forecasts from several regions and conducting the S&OP in a time bound manner so that other departments can take their actions from the forecast in a timely manner. Having a third party manage the process has created less resistance to sticking to predefined calendar.

Going forward the advantages of services approach to demand forecasting will be even more pronounced. As businesses become more and more complex, using data from external sources will become more and more critical. Businesses will always find it difficult to capture, understand and model the impact of these external factors into the forecasting process. Demand sensing is a global phenomenon and there are many companies like ZARA, P&G, Shell, Unilever, Mondelez, Kellogg, and General Mills etc which have implemented the demand sensing processes and tools. But using data captured from demand-sensing technologies to better your forecasts is going to even more critical. The world is moving ahead and adopting these cutting edge technologies to resolve their problems. Do not stay behind, start considering newer ways to use services to solve the demand forecasting challenge.
_______________________________________________________________________________________________

Alagu Balaraman is a partner in CGN & Associates and the managing director - India Operations. CGN is a global business performance consulting firm that specialises in end-to-end supply chain redesign and implementation.

Sujit Sahu is principal at CGN & Associates India Pvt Ltd