While visiting several global manufacturing companies, I could not stop thinking about a blog posting written by my co-author of Bricks Matter (Lora Cecere, AKA the Supply Chain Shaman). I realized that she has a really cool blog handle, and I need to get one, as well. With all kidding aside the blog posting was entitled: “Let’s Admit Seven Demand Management Mistakes of the Last Decade”. The reality is that those mistakes are still being made today.
It is concerning that after three decades of supply chain pioneers those mistakes are still inherent in the demand management process at many companies. In fact, it doesn’t matter the size of the company, or the location. I have sat across the table from several supply chain management teams over the past year at companies that span from less than $1B in annual sales to over $14B in annual sales. Talk about stirring emotions with the words demand planning, or demand-driven demand management, not to mention being market-driven. Although, the atmosphere in those rooms could be defined by despair, disillusionment, and most of all skepticism it was far from hopeless. It seemed like Déjà vu when I worked for a large CPG company back the 1980’s–no analytics, no technology other than Excel, and 100% “gut feeling” judgment injected into the process by multiple departments including sales, marketing, finance, and operations planning all in an attempt to create a one number consensus forecast. The story goes on with no real attention to accountability and little if no attention to the product mix as the focus is on a top down forecast. The supply chain leaders didn’t just say we want to improve our demand management process, but said that they had no choice due to the fact they were sitting on anywhere from $100M to over $600M in finished goods inventory, WHIP, and raw materials. Over $75M to $400M was in finished goods inventory, alone. Talk about being supply centric in their approach to demand management…wow!
As Lora mention in her blog that after two decades of process and technology refinement, excellence in demand management still eludes supply chain teams. This is an understatement. After two decades the demand forecasting and planning process is still the largest gap between satisfaction and performance. Given our research in writing Bricks Matter demand forecasting and planning is the key focus area for most companies over the next two years. For most it is the biggest challenge that they will face in the supply chain journey. Companies want to improve demand forecasting and planning, but have focused mainly on the process with little or no attention to data quality and analytics. As a result, their skepticism has become prevalent among their supply chain leaders that they can never be successful in improving demand forecasting. As Lora indicating in her blog demand forecasting and planning is important to supply chain leaders, but also an area with the largest gap in user satisfaction.
Based on my personal experiences visiting companies, we find that demand forecasting and planning is the most misunderstood supply chain planning process with little if any knowledge of how to apply analytics to downstream data. Also, well-intentioned consultants have given bad advice, particularly, that a one number forecast process is the key to success. In my experience, the one number forecast does not work. It only encourages well intended personal bias, and is used to set sales targets, financial plans, and other factors that are not directly related to an accurate demand response. What drives excellence in demand forecasting and planning is the ability to incorporate sophisticated data driven analytics into the process using large scale enabling technology solutions to create the most accurate unconstrained demand response. Once that unconstrained demand response is adjusted for sales, marketing, financial, and/or operational constraints it becomes a sales plan, marketing plan, financial plan, and/or a supply plan. In this article, I share insights on the current state and give actionable advice that supply chain teams can take to make improvements.
Why it Matters More than Ever to Embrace Demand-Driven Demand Management
Demand Management concepts are now twenty to twenty-five years old. The first use of the term demand management in the commercial sector was lay claim to in the late eighties or early nineties. Previously, the focus was on a more siloed approach to forecasting that was manual using very simple techniques like moving averaging and simple exponential smooth using Lotus Notes, and then, Excel, and a whole lot of “gut feeling” judgment. Sound familiar. In the mid-nineties demand planning and supply planning were lumped together, which gave birth to supply chain management concepts of demand planning and integrated supply chain planning.
As we have found is that most supply chain professionals are quickly realizing that their supply chain planning solutions have not driven down costs, and have not improved inventories or speed to market. What we have found companies globally across all industry verticals have actually moved backwards over the course of the last ten years when it comes to growth, operating margin and inventory turns. In some cases they have improved days payable, but this has pushed costs and working capital responsibility backwards in the supply chain, moving the costs to the suppliers.
As we mention through the Bricks Matter book in order to move forward, companies need to admit their mistakes of the past. They must be willing to fail in order to move forward. In their supply chain journey to sense demand signals and shape future demand the use of demand information will be critical in driving a more profitable demand response. Leaders must confront a number of mistakes made in the design of their demand management processes over the course of the last decade. The mistakes are many, but all can be corrected with changes to the process, use of downstream data, and most all, the inclusion of analytics. Let’s start with the biggest myth.
1) One-number Forecast. It is a Myth. Well-intentioned academics and consultants tout the concept of one-number forecasting. Enthusiastic supply chain executives have drunk the Kool-Aid, as they say. But, the reality is it does not reduce latency and it is too simplistic. The sole concept of a one number demand forecast is that “if everyone is focused on one number the probability of achieving the number is great”. As a result, the concept adds unintentional and in many cases intentional bias, or forecast error to the demand forecast. The reason is it is too simplistic, but the reality is that all the participants have different purposes, or intentions. I ask supply chain managers, “what is the purpose of your forecasting process?’ They say, “To create a demand forecast”. I respond, “What is the true purpose of their demand forecasting and planning process?” Is it to set sales targets, create a financial plan, or create a true unconstrained demand forecast? They pause, and say all the above. I ask, all the above are plans, not an unconstrained demand forecast. There is only one unconstrained demand forecast, or as close as possible “unconstrained” with some inherent constraints whether self-inflicted or customer specific. The people who push this concept really do not understand demand forecasting and planning.
A demand forecast is hierarchical around products, time, geographies, channels, and attributes. It is a complex set of role-based time-phased data. As a result, a one-number thought process is naïve. An effective demand forecast has MANY numbers that are tied together in an effective data model for role-based planning and what-if analysis. Even the eventual demand plan is sometimes not reflective of the original demand forecast due to capacity restraints, which results in demand shifting to accommodate supply constraints. In fact, most companies who described demand shaping during our 75 interviews with supply chain managers were actually describing demand shifting, not demand shaping.
A one-number plan is too constraining for the organization. A forecast is a series of time-phased plans carefully architected in a data model of products, calendars, channels and regions. The numbers within the plans have different purposes to different individuals within the organization. So, instead of a one number forecast, the focus needs to be a common set of plans for marketing, sales, finance and operations planning with different plan views based on the agreement on market assumptions and one unconstrained demand response. This requires the use of an advanced forecasting technology solutions and the design of the system to visualize role-based views that can only be found in the more advanced demand forecasting and planning systems.
2) Consensus Forecasting and Planning: Can be a Reality, but in many companies it is a myth. The entire basis for the concept of consensus forecasting and planning is based on the belief that each organization within the company can add insight (value) to improve the accuracy of the demand forecast. In concept, if designed properly this is correct. Like many concepts the reality is a result of the how it is implemented. In this case, the implementation has been flawed. The challenge is that most companies did not hold the groups within the organization accountable for their bias and error. Each group within the company has a natural bias (purpose), and corresponding error based on incentives. The old adage, “be careful what you ask for because to may get it”. Unless the process has structure regarding error reporting, the process of consensus forecasting and planning will distort the demand forecast adding error despite well-intended efforts to improve the forecasting and planning process.
We have worked with many companies that have redesigned their collaborative demand planning processes many times. Each time it was to improve the user interface to make data collection easier by sales. What I refer to as “automate what I do, but don’t change what I do”. In each redesign not once did they ever question the value and appropriate uses of the sales input or apply structure around the input that was driving a 40%-60% forecast over/under bias. We struggle with why more companies do not apply the principles of Lean to the consensus forecasting and planning process through Forecast-Value Add (FVA) Analysis. This is best described by Mike Gilliland in his bookThe Business Forecasting Deal: Exposing Myths, Eliminating Bad Practices, Providing Practical Solutions. In its simplest form FVA measures the impact of each touch point in the consensus forecasting and planning process before and after the statistical baseline forecast is adjusted by one of the participating organizations (i.e., Sales, Marketing, Finance, and Operations Planning). If that particular touch point isn’t adding value then you need to either eliminate it or weight the bias up/down. This requires that all the forecasts be captured each cycle and compared to determine any bias.
3) Forecast by Exception. It is a Reality. Given all the acquisitions and consolidation that has taken place of the past 20 years, SKU proliferation, as well as companies selling their products across geographic regions, markets, channels, and key accounts (customers) has made it difficult to touch every product every cycle. It is not uncommon for a company to have anywhere from 2,000 to 18,000 products (SKU’s) thatspan across multiple channels (e.g., grocery, mass merchandisers, drug, gas and convenience, and others), across multiple regions and countries, not to mention customers and demand points. This could lead to millions of forecasts each cycle. It is virtually impossible to touch every product every cycle. Companies forecast at some aggregate level in their product hierarchy with little attention to the lower levels (product mix). Then, imagine doing this in an Excel spreadsheet. Well that is reality. The biggest attributor to forecast error is the product mix due to the sheer number of products and locations (SKU/Ship to location). This requires a large scale automatic forecasting system that can do all the heavy lifting using analytics, and can filter on an exception basis those products and locations that need the most attention based on a set of business rules, and error statistics (e.g., MAPE, WAPE and others). Excel is simply not scalable, nor does it have the depth and breadth of analytics.
4) Fitting Demand to Supply versus Fitting Supply to Demand: This is Reality. Traditionally, companies focused on forecasting what manufacturing should make, rather than what the market and channel were demanding. This is a supply centric approach to demand forecasting and planning that compensates for the lack of a strong demand management process. The process needs to focus on identifying market opportunities and leveraging internal sales and marketing programs to influence customers to purchase the company’s products and services, also known as shaping future demand. This situation has changed the process focusing on modeling what is being sold in the channel to determine the best demand response. This difference, while it may sound insignificant, is a major change. It requires an additional step after demand sensing and shaping to translate demand into a more accurate demand response. Forecasting channel demand reduces demand latency and gives the organization a more current demand signal. It also allows the augmentation of the forecast with demand insights (signals) to improve the quality of the forecast. For most companies, this requires a re-implementation of demand planning methodologies and new enabling technologies.
5) Lack of Statistical skills: This is Reality. Recently, while meeting with the supply chain management team of a large appliances manufacturer, I was asked to provide them with a detailed description of the skills required to hire demand planners. This is not uncommon as most demand planners have minimal statistical skills. Their primary role in the demand management process is focused on taking aggregate level forecasts and disaggregating them into ship to location by SKU forecasts. This requires minimal statistical skills. This is done using Excel spreadsheets, and then, manually entered into a legacy ERP system. Those companies who invested in demand planners with advanced analytical skills combined with new demand forecasting and planning enabling technology based on demand sensing and shaping have significantly improved their forecasting processes.
Most traditional demand planning organizations are positioned in the operations planning departments too far upstream to understand how to apply analytics to downstream channel data. When meeting with supply chain managers, I ask “Who is responsible for demand generation?” They always respond, “Sales and marketing”. Then I ask, “Why then are the demand planner’s positioned in the operations planning organization?” When in fact, they should be positioned in the marketing organization where the domain knowledge exists. In other words, demand forecasting and planning requires analytics and domain knowledge. The new demand management organization of the future needs to be positioned in marketing for two key factors; 1) to provide statistical support, and 2) gain domain knowledge. As marketing products managers move every 2-3 years the demand planners will remain as the product domain knowledge experts, as well as the analytics expert over time. As a result, the companies will begin to models sell through versus sell into the channels of distributions, as demand planners begin to analyze using statistics to measure the effects of those factors that influence customers to buy their products. As a result, inventories will be managed more efficiently in the channels avoiding discounting, sales promotions, and other vehicles required to push products through the channels. This will have a positive impact on profit margins resulting in higher revenues.
6) Who should ultimately be Accountable for Demand Forecast Improvements? Sales and marketing are responsible for demand generation, and ultimately for creating the most accurate demand response. The primary role is to identify market opportunities, translate those opportunities into demand signals, measure the signals, and use them to shape (influence) future demand. The consensus should be between sales and marketing with financing assessing the programs to determine if they are profitable. If not, then it is finances role to push back on sales and marketing. This is a truly a market-driven demand management process. Operations planning should not provide another input into the consensus forecasting process other than to assess the implications from a supply perspective. If there is a capacity issue it should be raised at the S&OP meeting to determine a strategy to resolve the constraints (i.e., add another manufacturing shift, OEM the capacity to a third party manufacturer, or shift demand by moving a marketing program to accommodate the capacity constraint).
There is a large CPG manufacturer who does this best by following a structured demand management process that is supported by new demand-driven technology that allows them to measure sales promotions and marketing events mathematically calculating the lift, and then, assesses the lift to determine if it generates profit. If not, the sales promotion is not implemented. This combination of data, analytics, domain knowledge and financial assessment has significantly improve forecast accuracy as well as performance resulting in higher profit margins and lower finished goods inventory safety stock.
Looking to the future?
So while companies want to move forward, and the desire is to re-implement demand planning, in our opinion, they cannot be successful unless they admit to the myths and realities of the past.