Finance:Demand modeling
Demand modeling uses statistical methods and business intelligence inputs to generate accurate demand forecasts and effectively address demand variability. Demand modeling is becoming more important because forecasting and inventory management are being complicated by the increasing number of slow-moving items, the so-called “long-tail” of the product range, many of which have unpredictable demand patterns in which the typical “normal distribution” assumption used by traditional models is totally inadequate.[1] In these scenarios, successfully managing forecasts and inventories requires advanced demand and inventory modeling technologies in order to reliably support high service levels.[2]
Challenges of demand modeling
Product proliferation has resulted in an increasing number of products targeting niche markets with difficult-to-predict demand. These “long tail”[vague] products are becoming an increasingly important fraction of the product range, so companies are forced to deliver higher service levels even for these products to achieve overall service levels in the 98% to 99% range.[3]
Traditional demand modeling methods suggest safety stock levels which fail to achieve these service levels. These methods instead lead to manual corrections which often overcompensate, leading to uncontrolled overstocks.[4]
Traditional demand modeling methods
Many companies rely primarily on manual methods for tasks such as developing SKU, location and channel forecasts. For instance, a recent survey by Aberdeen Group showed that 78 percent of companies used Microsoft Excel spreadsheets as their primary demand modeling tool for trade promotions. Manual methods also involve time constraints that place limitations on the number of data streams that can be incorporated into the forecast, which therefore limits forecasting accuracy.[5]
Traditional demand modeling methods that assume most items follow a normal distribution are inaccurate for many companies that deal with intermittent, sporadic products.[6]
Advanced demand modeling methods
In contrast to traditional methods, advanced demand modeling methods sense the expected impact of different decisions, understand the distinction between sales channels, anticipate the impact of the decisions, and create a unified view of the demand. For example, if the demand for a certain product is expected to increase by 5% due to a line extension, decrease by 2% due to a reduction in selling space and increase by 1% due to pricing decisions, the demand modeling engine takes all these factors into account in calculating the demand.[7]
Big data analysis technologies make it possible to perform controlled experiments to test hypotheses and analyze results to discover complex correlations. For example, retailers can track the behavior of individual consumers from their Internet click streams, update their preferences and predict their behavior in real time.[8]
Traditionally, business departments have kept control of the data that is relevant to their own operations. Achieving the potential of big data-driven demand modeling requires integrating this data to find correlations across the organization and marketplace. The connection patterns that are achieved enable faster, more relevant decision-making. Pattern-based analytic solutions provide insights into customer-buying patterns to guide supply chain planning and execution.[9]
Demand modeling can also help find the right approach to designing print and web-based catalogs and advertising to have the greatest possible impact on sales. Social sentiment can be harnessed to yield better marketing strategies. Analytics can also help to focus location-based advertising and promotion.[10]
Benefits of demand modeling
Gartner research shows that retailers that effectively forecast consumer demand achieve a return on assets (ROA) twice as high as competitors, deliver three times the revenue growth rate and turn their inventory 4% faster than their competitors.[11]
Demand modeling, in general, makes it possible to reduce the amount of inventory that needs to be held to provide a given level of service. The result is that working capital requirements are reduced.[12]
Aberdeen Group research shows that organizations that are best-in-class, in the top 20% in terms of forecast accuracy, generated an average of 104.6% of forecast revenue. Aberdeen Group determined that these best-in-class organizations had established core business competencies including the ability to incorporate business drivers into the ongoing forecasting process, perform ‘what if’ scenarios and change analysis, re-forecast as market conditions change, perform multi-dimensional reporting with roll-ups, and trigger alerts to adjust forecasts based on internal events such as contract fluctuations or missed schedules and external events such as industry or financial indices.[13]
South African pharmaceutical manufacturer Cipla Medpro experienced difficulty in maintaining forecast accuracy as the company grew. The company was out of stock on 3 percent of its inventory at any given time, resulting in revenue losses that can never be recovered. The company implemented a demand modeling process that forecasts at both the product and the customer level. Today’s stock-outs are less than one percent of inventory and these are mostly caused by conditions that are out of the company’s control.[14]
Global electronics distributor Electrocomponents has automated the statistical forecast and inventory planning processes, enabling planners to focus on the use of market intelligence and their own knowledge to enrich and fine-tune demand models. Andrew Lewis, head of supply chain planning for the company has stated that this approach has generated a $3 million reduction in inventory.[15]
Demand modeling helped Groupe Danone predict the effects of trade promotion on a number of fresh products with short shelf life. The project improved forecast accuracy to 92 percent resulting in an improvement in service levels to 98.6 percent and a 30 percent reduction in lost sales.[16]
References
- ↑ Lora Cecere, “Of Long Tails and Supply Chains,” Supply Chain Brain, January 9, 2008.
- ↑ Kevin Perment, “Trade Promotion Management: An Excess of Excel! ,” AberdeenGroup blog, August 21, 2012.
- ↑ Lora Cecere, “Of Long Tails and Supply Chains,” Supply Chain Brain, January 9, 2008.
- ↑ Lora Cecere, “Of Long Tails and Supply Chains,” Supply Chain Brain, January 9, 2008.
- ↑ Kevin Perment, “Trade Promotion Management: An Excess of Excel! ,” AberdeenGroup blog, August 21, 2012.
- ↑ Noha Tohamy, “Attaining The Next Maturity Level In Demand Management,” Supply Chain Brain, February 19, 2009.
- ↑ Mike Griswold, Kevin Sterneckert, “Five Key Demand Management Lessons for Retailers," Gartner Group, March 24, 2011, ID Number: G00211678.
- ↑ Brad Brown, Michael Chu, James Manyika, “Are you ready for the era of ‘big data’?,” McKinsey Quarterly, October 2011.
- ↑ John Bruno, Radhika Subramanian, “The Big Value In Big Data: Seeing Customer Buying Patterns,” Forbes, September 12, 2012.
- ↑ Ann Grackin, “Shaking Up the Status Quo in Demand Management,” ChainLink Research blog, October 9, 2012.
- ↑ Mike Griswold, Allen Johnson, “Demand-Driven Retail Supply Chains Use Three Critical Skill Sets to Balance Operational and Innovation Excellence,” Gartner Industry Research, October 6, 2010. ID Number: G00207240.
- ↑ William Brandel, “Inventory Optimization Saves Working Capital in Touch Times,” Computerworld, August 24, 2009.
- ↑ Nick Castellina, “Improving S&OP with Planning and Forecasting Technology: An Integrated Look at Financial and Business Planning,” Aberdeen Group, October 2012.
- ↑ Joseph Ludorf, “Improve Your Supply Chain Forecasting,” Industry Week, October 26, 2012.
- ↑ Joe Shamir, “Next-Generation Supply Chain Planning Tools Should Serve People,” Supply Chain Brain, September/October 2012.
- ↑ Steve Steutermann, Noha Tohamy, “The Quest for Demand Management Excellence: Progress So Far ”, Gartner Supply Chain Executive Conference, London, UK, September 17–18, 2012.
 |
