CENTERS OF FORECASTING EXCELLENCE

Due to globalization and expanded product portfolios, many companies are considering creating centers of forecasting excellence within their corporate headquarters, particularly at larger global companies. Furthermore, they are staffing those centers of excellence with demand analysts, not demand planners. So, what is the difference? Demand analysts are responsible for creating the statistical baseline forecasts for all the regions/divisions. Then, pass those statistical baseline forecasts to the regional/divisional demand planners to refine (make adjustments) to the statistical baseline forecasts. Those adjustments are based on local sales and marketing activities, such as pricing actions, sales promotions, and others.

The skill sets of these newly created demand analyst positions are different from the demand planners. The demand analysts have advanced statistical skills and strong business acumen. They also have strong collaboration skills as they work closely with the regional demand planners. The regional demand planners do not necessarily have a strong statistical skill set, but work closely with the local commercial business teams to refine the statistical baseline forecasts reflecting regional sales/marketing activities (i.e., pricing actions. sales promotions, marketing events, and others). Another question that always follows is the ratio of demand analysts to demand planners. Based on my experience, it is recommended that there be one demand analyst for every three or four demand planners. This seems to be the optimal mix between demand analysts and demand planners. Once all the statistical models are generated demand analysts only need to tweak the statistical baseline forecasts on an exception basis requiring fewer resources. The demand analysts also provide ad hoc analysis in support of the global commercial teams to assess business strategic initiatives and tactics.

If you are implementing a demand-driven forecasting and planning process, those statistical baseline forecasts will include key performance indicators (KPI's) such as price and sales promotions, which the demand planners can utilize to make adjustments through what-if analysis, not gut-feeling judgment. Also, POS/syndicated scanner data (true demand) can be integrated into the demand planning process encouraging the commercial teams to engage with the demand planners. The goal is to have demand analysts building holistic all-inclusive statistical baseline forecasts that include KPIs, such as sales promotions, price, advertising, in-store merchandising and more. Then, demand planners work with the commercial teams (sales/marketing) running what-if simulations to adjust the forecasts based on data, analytics, and domain knowledge rather than gut feelings. The what-if analysis is done at the local divisional/country level for tactical planning, but can also be conducted at the corporate (global) level for strategic planning. This is done using large-scale hierarchical forecasting enabled by new technology.

The goal is to reduce judgment bias by using data, analytics, domain knowledge, and scalable technology, thus minimizing bias judgmental overrides that add error. This requires investment in people (skills/behavior changes), process (horizontal, not vertical) that includes the commercial side of the business, using analytics (not just descriptive, but also predictive analytics), and finally, supported by scalable enterprise technology. Most traditional demand planning processes focus only on the process and technology, and then rely on gut feeling judgment to enhance the accuracy of the demand forecast. That traditional process has failed. Furthermore, Excel spreadsheets are not scalable enough to handle thousands of SKUs across multiple market areas on a global basis. It requires an integrated scalable enterprise solution.

DEMAND MANAGEMENT CHAMPION

Companies are quickly realizing that an internal champion is needed to drive the change management required to gain adoption, because this new process design and added demand analyst role is a radical change for most companies. Also, even if you get adoption, you need it to be sustainable. Many companies gain adoption, but cannot sustain it to make it part of the corporate culture. In order to make it sustainable, companies need to incorporate predictive analytics into the process that are supported by a large-scale enterprise technology solution with an easy-to-use user interface (UI). Also, an internal ongoing champion involvement will be required to assure this new approach to demand planning becomes part of the corporate culture over time.

These interdependencies are also influenced by the strategic intent of a company's demand planning process. In other words, is the intent to create a more accurate demand response, a financial plan, marketing plan, supply plan, or a sales plan (target setting)? These different intentions are conflicting, and are not really forecasts, but rather, plans that are derivatives of the unconstrained demand forecast.

DEMAND-DRIVEN PLANNING

Demand-driven planning is the set of business processes, analytics, and technologies that enable companies to analyze, choose, and execute against the precise mix of customer, product, channel, and geographic segments that achieves their customer-facing business objectives. Based on recent observations and research, demand-driven planning on average is driven 60 percent by process, 30 percent by analytics, and 10 percent by enabling technology, depending on the industry, market, and channel dynamics that influence how companies orchestrate a demand response. Although enabling technology represents only 10 percent, the other 90 percent cannot be achieved without the enabling technology due to scalability and analytical requirements, not to mention data integration requirements that span across the global corporate enterprise. The need for an improved demand planning process focuses not only on process, analytics, and technology but also the importance of integrated collaboration across the global enterprise.

Demand-driven planning utilizes data from market and channel sources to sense, shape, and translate demand requirements into an actionable demand response that is supported by an efficient supply plan, or supply response. A true demand-driven forecast is an unconstrained view or best estimate of market demand, primarily based on corporate specific historical sales demand, preferably POS, sales orders, and shipment information. Demand shaping uses factors, such as price, new product launches, trade and sales promotions, advertising, and marketing programs, in addition to other related sales and marketing information, to influence what and how much consumers will buy.

WHAT IS DEMAND SENSING AND SHAPING?

Demand sensing and shaping are common terms that have been used loosely over the past several years with different definitions, depending on the industry and purpose. The most common definitions are associated with the consumer packaged goods (CPG) industry.

Demand sensing, especially in recent years, has come to embody using granular downstream sales data (mainly sales orders, preferably POS/syndicated scanner data) to refine short-term demand forecasts and inventory positioning in support of a one- to six-week supply plan. It is slowly being expanded to cover medium-term operational and inventory replenishment plans that require a 1- to 18-month demand forecast. Eventually, it will also include long-term strategic forecasting and planning (two years into the future and beyond). The term demand shaping often describes measuring the relationships of consumer (or customer) demand with respect to sales promotions and marketing events and/or price discounts, then using those consumer demand influence factors to shape future demand using what-if scenario analysis. These new, much broader definitions for demand sensing and shaping have been at the forefront of many conversations with senior executives across all industries globally.

True Demand Shaping

Demand shaping happens when companies use sales and marketing tactics like price, promotion, new product launches, sales incentives, and/or marketing programs to influence future consumer demand, to generate not only incremental unit volume but also profit. All too many times, companies believe that they are shaping demand but find that they are really just shifting demand (moving demand from one period to another). Moving demand from one period to another and selling at a lower margin without improving market share and revenue growth creates waste in the supply chain. The first step in the demand-driven forecasting and planning process is sensing market conditions based on demand signals and then shaping future demand using predictive analytics such as price optimization, trade promotion analysis, new product launch plan alignment, and social/ digital/mobile data convergence (see Figure 4.1). Demand sensing reduces the latency of the demand signal by 70 to 80 percent, allowing the company to better understand and see true channel demand. Demand shaping combines the tactics of pricing policies, sales promotions, sales and marketing incentives, and new product launches to increase demand.

Traditional demand forecasting and planning systems were not designed to sense demand patterns other than trend/cycle, seasonality, and level (unexplained). For that reason, it is impossible for traditional ERP/demand management systems to conduct demand-sensing and shaping activities associated with price, sales promotions, channel marketing programs, and other related factors. As the global marketplace has become increasingly volatile, fragmented, and dynamic, and as supply chain lead times have become overextended, companies are quickly coming to the realization that their demand management systems are no longer adequate to predict future demand. There are two primary factors that have contributed to this situation:

1. Limited statistical methods available in traditional demand management systems:
   1. Can only sense and predict stable demand that is highly seasonal with distinct trend patterns.
   2. Primarily use only one category of statistical models, called time series methods, with a focus on exponential smoothing models, such as simple exponential smoothing, Holt's two-parameter exponential smoothing, and the Holt-Winters three-parameter exponential smoothing.
2. Process requires domain knowledge versus judgment to:
   1. Define data availability, granularity, and sourcing.
   2. Assess the dynamics of the market and channel segments to identify factors that influence demand.
   3. Run what-if analyses to shape future demand based on sales and marketing tactics/strategies.

Research continues to show that there is a strong correlation between demand visibility and supply chain performance. As demand visibility yields higher accuracy in assessing demand, efficiencies continue to accumulate throughout the supply chain. Yet in most companies, there is still a wide gap between the commercial side of the business, with its understanding of the market and plans for demand sensing and shaping (e.g., sales/marketing tactics and strategies, new product commercialization, life cycle management, and social media), and the supply chain organization, with its ability to support those efforts.

Demand sensing as a core capability isn't new; retailer POS data, syndicated scanner data, customer insights, and focus groups have guided marketing and sales promotional programming for over two decades. The challenge is how to translate these demand insights into actions that can drive an efficient supply response. The ability to sense, shape, and translate demand into an accurate demand forecast and a corresponding supply response requires more transparency and collaboration between the organization's commercial and operational functions.

The key to demand shaping is cross-functional collaboration between sales and marketing and among the other members of the supply chain (e.g., finance) by coordinating and agreeing on demand-shaping programs (see Figure 4.2). The core purpose of such programs is to drive unit volume and profitability among the company's brands and products.

At first, these activities typically are monitored and managed independently by each functional department, such as sales, strategic marketing, and product management, with little cross-functional integration. For example, a price change occurring simultaneously with a product sales promotion could erode the profitability of the product or create an unexpected out-of-stock situation on the retailers' shelves. Cross-functional collaboration among sales, marketing, and finance requires companies to shift to a cross-departmental market orientation that balances the trade-offs of each tactic and focuses on revenue generation and profit (see Figure 4.3), not just reducing inventory costs.

To better understand the dynamics of demand sensing and shaping, we need to break down the demand management process into a capability framework made up of five key components:

1. Large-scale hierarchical statistical engine. A set of more sophisticated statistical models is a key requirement to enable demand sensing and shaping, as well as scalability to forecast hundreds of thousands of products up/down the business hierarchy. Such models measure the effects of different sales and marketing events and enable a better understanding of the incremental volume that is associated with them. The ability to measure past events over time and clearly identify which ones are profitable helps companies avoid unexpected planning events that produce negative returns and exploit those identifiable events that are more profitable in driving incremental demand and profit.

   Companies can proactively influence the amount and timing of consumer demand by varying the future marketing mix elements that influence demand for a product through the use of what-if analysis. For example, varying future price, sales promotions, levels of merchandising, and advertising can influence consumers to purchase more of a company's products. More advanced methods, such as ARIMA, ARIMAX, and dynamic regression models as well as utilizing downstream POS/syndicated scanner data can help sales and marketing planners better understand consumer demand and uncover insights such as price elasticity. Combining these more advanced statistical techniques with decision-support tools, such as what-if analysis, enables sales and marketing planners (support by a demand analyst) to determine the right trade-offs within the marketing mix by market, channel, brand, and product that drive incremental unit volume and profit. Demand analysts and planners are moving toward the use of downstream data to help capture consumer insights to build on current trends and seasonality, utilizing marketing programs based on the combination of historical data and domain knowledge, not gut-feeling judgment.

2. Visualization analytics (VA). VA capabilities combine the power of descriptive analytics associated with monitoring, tracking, and reporting with the power of predictive analytics to uncover actionable insights with user-friendly interfaces. VA control towers/dashboards, along with predictive analytics, allow sales and marketing personnel to collect, integrate, and apply data from the statistical engine and the field to support business tactics and strategies, such as planning pricing changes, sales promotions, and measuring results against strategic and tactical business plans. Demand shaping can be used to reduce demand volatility, thereby reducing the need for supply network agility. For example, corporate leaders in various industries (e.g., food services, spare parts planning, and electronics) are looking to use Web channels to sense demand signals and shape future demand using distributor networks.
3. Post reconciliation of performance. It is important to measure demand-sensing and shaping programs after each completed demand forecasting cycle to determine the success or failure of the programs implemented to drive demand.

   Historically, it took weeks to review and assess the success or failure of a sales promotion after its completion. With new enabling technology, along with downstream data collection and synchronization processes, as well as market sensing and shaping capabilities, today it is much easier and faster to monitor, track, and report on the effects of demand-shaping programs. This allows companies to manage the demand-shaping process around real-time demand signals. Adjustments can be made to demand-shaping programs within a daily or weekly period to better manage the outcome.

4. Executive alignment to support change management. Establish clear decision criteria, empower senior managers and their staff, and develop an appropriate incentive program that includes rewards for accurate demand forecasts. Decentralize tactical knowledge-based decision making while balancing corporate strategic unit volume and profit objectives. Stress the importance of building a demand forecast based on sales and marketing programs that are profitable, not just volume generators. There will be a paradigm shift, moving from a view of unit volume in isolation of profitability (not considering profit, but only incremental volume for trial purposes) to a more focused view of how unit volume increases can affect profitability.
5. Continuous business process improvements. Short- and long-range business strategy and planning, operational tactical planning, and post-event analysis must be coordinated in the organization. Sophisticated analytics shared across the various departments within a company through well-designed decision support networks will provide more consistency and alignment of internal processes and work flow to drive profitability.

Demand shaping focuses on creating an unconstrained demand forecast that reflects the sales and marketing activities that shape demand rather than using supply to manage demand. Demand shaping is a process that requires predictive analytics supported by enabling technology. The system should be flexible and easy to use, with quick response time and closed-loop feedback to measure and report the value of those adjustments made to an initial statistical forecast. Without access to the intuitive system, the sales organization has legitimate reason to resist participating in the forecasting process.

A NEW PARADIGM SHIFT

Educating sales, creating a well-structured demand forecasting process, clearly defined organizational roles and responsibilities, and access to information are all important to help incentivize the sales team to participate. Among these, there are two axioms that should also be understood: (1) what gets measured gets done; and (2) incentives drive behavior.

Sales team performance metrics often fall into a revenue generation category. Typical measurements are based against a sales quota (gross or net value), and may also be measured based on a margin or profitability target. As a result, the focus becomes how much revenue can be generated by the sales organization and how profitable is that revenue for the company. At its basic level, the quota is a forecast. Since compensation is determined by the results compared to a quota, or forecast, there is a tendency to understate the forecast in order to improve the chances of exceeding the target.

Another common measure for the sales team may be based on customer service. This is normally an order fill rate, such as delivered in-full, on-time (DIFOT), or some other similar measure, such as perfect order fill rates. When customer service measures are used to determine compensation, the tendency is to overforecast demand in an attempt to make sure adequate inventory is available to maximize order fill rates.

Tying sales team performance to forecast accuracy can balance demand and service with the least amount of inventory possible. In order to more closely align sales team performance with forecast accuracy, it is important to establish the correlation between forecast accuracy and the amount of inventory required to support desired customer service levels. Drawing that correlation between forecast accuracy and the costs of inventory can be somewhat of a challenge, but there are experts and processes capable of pinpointing the relationship.

Inventory optimization provides visibility to the multiple drivers of inventory investment, including forecast accuracy and supply lead-time variability to name a few. If an inventory planning function exists within the supply chain team, they can help drive an understanding of current forecast accuracy, as well as the trade-off between customer service levels and inventory investment. With this understanding, performance measures can be developed that provide a much better incentive for meaningful sales team involvement in the forecasting process.

Finally, the forecast isn't a number pulled off the top of someone's head, although in many cases that's not too far from the truth. The demand forecast is the sum of many parts working together toward a common goal. The sales team can add significant value to the forecasting process. However, the process must harness the intelligence the sales organization provides to align consumer demand at strategic and tactical levels with the company's marketing capabilities, resulting in improved revenue and profitability. At the strategic level, the emphasis is on aligning long-term marketing investment strategies with long-term consumer demand patterns while maximizing marketing investment effectiveness. At the tactical level, the focus is on understanding customer demand patterns, and proactively influencing demand to meet available supply, using the marketing mix to sense and shape price, sales promotions, marketing events, and other related factors to influence demand generation and profitability.

LARGE-SCALE AUTOMATIC HIERARCHICAL FORECASTING

Most companies review their forecasts in a product hierarchy that mirrors the way they manage their supply chain or product portfolio. In the past, product hierarchies in most companies were simple, reflecting the business at the national, brand, product group, product line, and SKU levels. These product hierarchies ranged from hundreds to a few thousand SKUs, spanning a small number of countries or sales regions and a handful of distribution points, making them fairly easy to manage.

During the past two decades, however, many industries have gone through major consolidations. Larger companies found it easier to swallow up smaller companies to increase their economies of scale from a sales, marketing, and operations perspective rather than growing their business organically. They realized additional benefits as they flushed out inefficiencies in their supply chains while increasing their revenue and global reach. Unfortunately, with all this expansion came complexities in the way they needed to view their businesses.

As companies consolidate through acquisition and expand globally, their product portfolios have grown exponentially from hundreds of SKUs to excess of hundreds of thousands of SKUs across multiple intersections (e.g., geography, region, market, division, channel, brand, product group, product, SKU, customer, demand point, and others). This situation has required companies to generate a large number of forecasts (millions in some cases) based on time-stamped data stored in their transactional or time series databases. Social media, point-of-sale (POS) data, syndicated scanner data, shipments, sales orders, pricing and promotion data, inventory data and others are examples of data that are stored in transactional databases. A skilled analyst can forecast a single time series by applying good judgment based on his or her knowledge and experience, by using various time series analysis techniques, and by utilizing good software based on proven statistical theory. Generating large numbers of forecasts and/or frequently generating forecasts requires some degree of automation. Common problems that a business faces are:

• No skilled analyst is available (little or no statistical skills).
• Many forecasts must be generated.
• Frequent forecast updates are required (weekly and/or monthly).
• Time-stamped data must be converted to time series data (weekly and/or monthly intervals).
• Difficult to run various statistical forecasting models for each time series.

TRANSACTIONAL DATA

Transactional data are time stamped data collected over time at no particular frequency. Some examples of transactional data are:

• Internet data
• Point of sales (POS) data
• Syndicated scanner data
• Shipment data
• Sales order data
• Inventory data
• Key account data

Companies often want to analyze transactional data for trends and seasonal variation for demand forecasting and planning. To analyze transactional data for trends and seasonality, statistics must be computed for each time period and season of concern. The frequency and the season may vary with the business problem. For example, various statistics can be computed on each time period and season. For example,

• Web visits by hour and by hour of day
• Sales per week by month and year
• Inventory depletions per week and by week of month
• Sales promotion volume per week by month and year
• Price per product by week and year

TIME SERIES DATA

Time series data are time-stamped data collected over time at a particular frequency. Some examples of time series data are:

• POS data per week or month
• Inventory depletions per week or month
• Shipments per week or month
• Sales orders per week or month

The frequency associated with the time series varies with the challenge at hand. The frequency or time interval may be daily, weekly, monthly, quarterly, yearly, or many other variants of the basic time intervals. The choice of frequency is an important modeling decision. This decision is especially true for automatic forecasting. For example, if you want to forecast the next four weeks, it is best to use weekly data rather than daily data. The forecast horizon in the former case is 4, in the latter case is 28.

Associated with each time series is a seasonal cycle or seasonality. For example, the length of seasonality for a monthly time series is usually assumed to be 12 because there are 12 months in a year. Likewise, the seasonality of a daily time series is usually assumed to be 7. The usual seasonality assumption may not always hold. For example, if a particular business's seasonal cycle is 14 days long, the seasonality is 14, not 7. Time series that consist of mostly zero values (or a single value) are called interrupted or intermittent (sparse) time series. These time series are mainly constant valued except for relatively few occasions. Intermittent time series must be forecast differently from non-intermittent time series.

FORECASTING MODELS

There are numerous types of forecasting models that a skilled analyst can use. For automatic forecasting of large numbers of time series, only the most robust models should be used. The goal is not to use the very best model for forecasting each time series. The goal is to provide a list of candidate models that will forecast the large majority of the time series well. Overall, when an analyst has a large number of time series to forecast, the analyst should use automatic forecasting for the low-valued forecasts. Then, the analyst can spend a larger portion of his/her time dealing with high-valued forecasts or low-valued forecasts that are problematic.

The candidate models need to be more robust including all categories of statistical forecasting methods (e.g., time series—moving averaging, exponential smoothing, ARIMA; intermittent demand methods; causal methods—ARIMAX, regression, multiple linear regression, dynamic regression; and weighted combined methods). These models have proven their effectiveness over time. They not only consider the local level, trend, and seasonal components of the time series, but also causal factors like price, sales promotions, marketing events, in-store merchandising, and others. The term local is used to describe the fact that these components evolve with time. For example, the local trend component may not be a straight line but a trend line that changes with time. In each of these models, there is an error or random component that models the uncertainty.

The components associated with these models are not only useful for demand forecasting but also for describing how the time series evolves over time. The forecasting model decomposes the series into its various components. For example, the local trend component describes the trend (up or down) at each point in time and the final trend component describes the expected future trend. These forecasting models can also indicate departures from previous behavior or can be used to segment time series.

The parameter estimates (weights) describe how fast the component is changing with time. Weights near zero indicate a relatively constant component, and weights near one indicate a relatively variable component. For example, a seasonal weight near zero represents a stable seasonal component, and a seasonal weight near one represents an unstable seasonal component. Weights should be optimized based on the data for best results using statistics, not human judgment.

Today, with global reach across multiple countries, markets, channels, brands, and products, the degree of granularity has escalated tenfold or more (see Figure 4.4). Company product portfolios have increased exponentially in size due to consolidation through acquisition and global expansion. Subsequently, the proliferation of the SKU base for many companies has expanded into the thousands and in some cases hundreds of thousands. It is not unusual to see companies with more than 10,000 SKUs that span across 100 or more countries (see Figure 4.5).

Further escalation occurs as marketing departments segment their consumer base by ethnicity, channels of distribution, and purchase behavior. The resulting increased granularity has further complicated company product hierarchies. All this proliferation in business complexity has made it difficult not only to manage the data but also to process the data in a timely manner. As such, companies need more robust technology that can not only forecast up/down their complex business hierarchies but also apply more advanced statistical methods to capture and measure the effects of all the marketing activities used to generate demand. This new technology capability provides not only top-down and bottom-up overrides and reconciliation but also middle-out at any middle level in the business hierarchy. In fact, it has been found that on average middle-out overrides and reconciliation tend to be more accurate.

SKILL REQUIREMENTS

Business leaders need to deploy large-scale automatic forecasting systems to aid in their decision-making processes. Deploying such solutions will require different skills to effectively utilize this new technology. Demand analysts (statisticians, econometricians, mathematicians, data scientists and others) speak in terms of statistical techniques (model building, parameter estimations, predictions, statistical testing, etc.). Demand planners who are responsible for facilitating collaboration across the organization with sales, marketing, finance, and operations planning to create a final consensus demand plan speak in terms of demand management regarding forecast accuracy, inventory costs, rough cut capacity planning, and other supply chain terms. Finally, business users (marketers, brand managers, category managers, and others) speak in terms of business domain (pricing, sales promotions, marketing events, market channels, decision making and more).

Demand analysts who need a rich set of statistical techniques are more skilled or concerned with the information stored in the statistical model repository (model specification, model selection lists, etc.), and may be less skilled or concerned with data management, but require business domain knowledge. If a business does not have statistical analysts, it must rely solely on automation or must contract these skills to outside statistical consultants.

Demand planners need a rich set of extraction, translation, loading, and data quality techniques. They are more concerned with the information stored in the time series data, event repository, and the overall process flow and may be less skilled or concerned with statistical analysis and business domain issues. Business users need a rich set of domain-specific reporting and easy to use point-and-click interfaces. They are more concerned with the information stored in the forecast results repository and may be less skilled or concerned with statistical analysis and data management issues.

A large-scale automatic forecasting system allows the separation of these various skills and permits the necessary intercommunication between skill sets while at the same time applying sound demand forecasting principles to be used by the business.

SUMMARY

Corporate supply chain networks have been evolving from their traditional supply-driven architectures to becoming demand-driven due to global expansion and demand volatility. This has made their supply chain networks increasingly complex, requiring more effort and resources to orchestrate demand. As a result, companies can no longer use buffer stock (inventory) to protect against demand volatility and long lead times. Transitioning from a supply-driven process to a demand-driven requires investment in people, process, analytics, and technology. Process and technology alone is not enough. At the mature stage of the demand-driven transformation process, companies must focus on balancing profitable growth and marketing investment efficiency with inventory costs and customer service while reducing working capital. When demand-driven maturity is achieved, there is not only better balance but also greater agility across the supply chain.

The implementation challenges associated with the transition to demand-driven networks require change management, which can be enormous from a people, process, analytics, and technology standpoint. Companies that attempt to navigate a demand-driven transformation process must tackle corporate cultural changes head-on. The most important changes are:

• Incentives: The role of the commercial teams. As long as sales are incentivized only for volume sold into the channel and marketing only for market share, companies will never become demand-driven. To make the transition to demand-driven, companies must focus on profitable sales growth through the channel.
• Traditional view of supply chain excellence. For demand-driven initiatives to succeed, they must extend from the customer's customer to the supplier's supplier. Most company supply chain models encompassed only deliver and make. Customer and supplier initiatives usually are managed in separate initiatives largely driven by cost.
• Leadership. The concepts of demand latency, demand sensing, demand shaping, demand translation, and demand orchestration are not widely understood. As a result, they are not included in the definition of corporate strategy.
• Focus: Inside out, not outside in. Process focus is from the inside of the organization out, as opposed from the outside (demand-driven) back. In demand-driven processes, the design of the processes is from the market back, based on sensing and shaping demand.
• Vertical rewards versus horizontal processes. In supply-based organizations, the supply chain is incentivized based on cost reduction, procurement is incentivize based on the lowest purchased cost, distribution/logistics is rewarded for on-time shipments with the lowest costs, sales is rewarded for sell-in of volume into the channel, and marketing is rewarded for market share. These incentives cannot be aligned to maximize true value.
• Focus on transactions, not relationships. Today, the connecting processes of the enterprise—selling and purchasing—are focused on transactional efficiency. As a result, the greater value that can happen through relationships—acceleration of time to market through innovation, breakthrough thinking in sustainability, and sharing of demand data—never materializes.

The demand-driven value network implementations are not a traditional approach of adding ERP + Advanced planning and scheduling/Customer relationship management + Supplier relationship management, and shake until well blended, as Lora Cecere (Founder and CEO of Supply Chain Insights) mentioned recently in her Supply Chain Shaman blog. In fact, some of the most demand-driven companies have legacy systems that have not supported the process, but actually have hindered it. In order for the transition to be adopted and sustainable, focus on:

• Process. The implementation requires a focus on the processes: revenue management, new product launch, downstream channel data management, and use of demand insights.
• Network design. The design of the network is an essential element to actualizing this strategy. Demand-driven companies have made deep investments in supply chain modeling software—optimization and simulation—and actively model scenarios for the network reflecting changes in both demand and supply.
• Demand sensing and shaping. These companies also have a control tower to actively sense network changes and adapt the network for changes in market demand, constraints, and opportunities. This overarching group crosses source, make, deliver, and sell to work hand-in-hand with customer service to maximize the use of resources while minimizing costs and maximizing profitability.

So, does this mean that we give up on demand-driven concepts? The answer is emphatically no. It is the right concept, but it will take more time and investment in people, process, analytics, and technology.

KEY LEARNINGS

• Due to globalization and expanded product portfolios, many companies are considering creating centers of forecasting excellence within their corporate headquarters.
• Companies are now hiring demand analysts who are responsible for creating holistic statistical baseline forecasts for all the regions/divisions.
  • They pass those all-inclusive statistical baseline forecasts to the regional/divisional demand planners to refine (make adjustments to) the statistical baseline forecasts using what-if scenario analysis, not gut feeling judgment.
• Companies are now investing in downstream data (POS/syndicated scanner) to integrate into the demand planning process.
  • The goal is to have demand analysts build statistical baseline forecasts (demand sensing) that include KPIs, such as sales promotions, price, advertising, in-store merchandising, and more.
  • Then, pass those statistical baseline forecasts to demand planners to run what-if scenarios with sales/marketing to shape future demand.
• An internal champion is needed to drive the change management required to gain adoption, because this new demand-driven process design and added demand analyst role is a radical change for most companies.
• Demand-driven planning is the set of business processes, analytics, and technologies that enable companies to analyze, choose, and execute against the precise mix of customer, product, channel, and geographic segments that achieves their business objectives.
• Demand sensing is the translation of downstream data with minimal latency to understand what is being sold, who is buying the product (attributes), and how the product is impacting demand.
• Demand shaping enables companies to influence the future volume and profits by orchestrating a series of marketing, sales, and product tactics and strategies in the marketplace.
  • Companies believe that they are shaping demand but find that they are really just shifting demand (moving demand from one period to another).
• The next generation demand management process capability framework is made up of five key components.
  1. Large-scale hierarchical statistical engine for forecasting hundreds of thousands of products up/down the business hierarchy.
  2. Visualization analytics (VA) for monitoring, tracking, reporting, and exploration.
  3. Post reconciliation of performance to determine why the statistical forecasts did not meet expectations based on the analytics and domain knowledge.
  4. Executive alignment to support change management to gain sustainable adoption.
  5. Continuous business process improvements.

FURTHER READING

1. Leonard, Michael, “Large-Scale Automatic Forecasting Using Inputs and Calendar Events,” SAS Institute Inc. white paper, 2005, pp. 1–28.