Duncan Alexander is Consulting Director for StrataBridge the Sales and Operations Planning consultants.
Second part of the S&OP blog series. Link to the first blog of this series: Sales & Operations Planning – the most important KPI’s for the S&OP process S&OP is a management business process that organizes the different departments and organisations in a certain way to get the best possible visibility and understanding. Providing one unconstrained demand plan is a major aspect of this process. The way how companies come to that one demand plan differs enormous and sometimes forecasting as part of the process is not considered at all. I talked to several companies and found out that there seems to be a lot of confusion about forecasting: Is planning the same as forecasting? Is it a process or a meeting? Is the solution just a good statistical forecasting tool? Is it about detail by SKU or aggregated data by product family? What is more important, the numbers or the story behind the numbers? Should we use constrained demand or unconstrained demand? Do Sales and Marketing or Supply Chain own the forecast? Where does customer collaboration fit (how important are POS data)? What is the difference between forecast accuracy and bias? Let me ask the expert Duncan Alexander from the S&OP consultants StrataBridge some questions from customers on how forecasting fits within typical S&OP processes. Marion: What comes typically first, the plan or the forecast? Duncan: In my opinion planning comes first BUT there is an interdependency since a forecast may inform the plan. Let me explain. Business planning or sales planning is a process where people develop strategies and tactics to generate revenue and profit. These strategies and tactics may be formalised in budgets or the Annual Operating Plan in effect they are the target. Forecasting is the process that estimates the results of implementing these plans. This is why the plan will never match the forecast. The plan is almost always a stretch target, is probably formulated weeks or months in advance and generally is created top-down. The forecast by contrast should be the most likely outcome, is regularly updated (incorporating updated plans and changed assumptions) and is often based on a bottom-up approach. The military have a phrase for this: No plan survives first contact with the enemy. The type of business planning favoured in the 1960s and 1970s was almost Stalinist in approach. Detailed 5 year plans meticulously put together by teams of corporate staffers in thick binders. In the 21st century, planning is much more about setting strategic direction and then using the S&OP process to help navigate in that direction, adapting flexibly to changes in market conditions and competitor activity. As the speed of change in business continues to accelerate with shorter product lifecycles, more innovation and technological advances, the forecast will become more and more important. So where does forecasting fit in S&OP? Is it just a meeting? There is commonly a meeting, but forecasting is really a process within S&OP. In a classical 5 step S&OP process, we call the second step Managing Demand, and this is based around the forecasting process. As usual in S&OP there is no one answer for anything the process needs to be tailored to suit your business, but in a consumer goods business the forecasting process within the Managing Demand step might look like this: Firstly new planning assumptions (perhaps market growth rates, the updated innovation plan, and a revised promotional plan) are issued. Next a Demand Planner or Forecasting Manager might use a tool like the Demand Planning module of APO to create a statistical forecast for the months ahead based on historical trends (probably over a minimum of an 18 month horizon). Sales and Marketing people would then use this forecast together with their own customer and consumer knowledge and plans to create a bottom-up forecast (in both volume and value terms). Then at the Demand Review meeting or Consensus Forecast meeting the statistical forecast would be compared to the bottom-up forecast, the top-down plan, last year sales and various other inputs such as the level of bias in previous forecasts, and the forecasting team will jointly agree on a consensus forecast. The consensus forecast may match one of the input forecasts, but is usually different so adjustments have to be made after the meeting in the forecasting system before the forecast is released to the third step of the S&OP process Managing Supply. Is the solution just a good statistical forecasting tool? Almost certainly not! While there are some great tools around both from SAP and other vendors, that contain lots of different statistical forecasting algorithms the system will not be a silver bullet. Critically all statistical forecasting tools rely on historic data from which predictions of the future are made. If we could guarantee that the future will follow the same pattern as the past then we could rely on statistical forecasting as an accurate guide, BUT . we know the future will be different, retailers are becoming more demanding, competition is becoming more intense, the rate of innovation is increasing etc., so we cant rely solely on a forecasting tool. Investing in a forecasting tool should be a high priority for any business looking to develop an S&OP process though. In additional to the valuable input of a statistical forecast to the process, the system will allow you to manage the data input in a much more structured way than spreadsheets, allowing multiple users to input forecasts simultaneously, allow forecasts to be created at different levels of aggregation, allow upsides and downsides to be forecast, provide forecast error and bias tracking, and provide financial as well as volume views. But however good the systems tool is, it is only one input to a good forecasting process. Is it about detail by SKU or aggregated data by product family? Again it depends. In the immediate forecast horizon, the forecast will need to be detailed at SKU level because the supply chain needs to be replenishing at item code level. In the medium to long term however the need for SKU level data is less evident. The Managing Supply step of the S&OP process is trying to get the right resources organised at aggregate level extra capacity, recruiting and training new crews, warehousing space, raw material supplies etc. They usually dont need to know by SKU 15 months out what the exact variant split will be, but they do need to know the shape of the demand when is it influenced by a major promotion or new listing opportunity for example. Creating a forecast at detailed level for 18 months out is time consuming and very inaccurate so why bother? Its easier and more accurate to create the medium to long term forecast at product family or brand level and then use the forecasting system to disaggregate back to detailed level if required. We call this roughly right, not precisely wrong forecasting. What is more important, the numbers or the story behind the numbers? One of the other key reasons for using the highest level of aggregation possible for forecasting within S&OP is that this gives you the ability to understand the forecast better by tying words to it. By this I mean the assumptions that the forecast has been based on. Consider this simple forecasting example. The Marketing departments forecast of 3,000 units assumes that a new advertising campaign will have a dramatic uplift on previous sales figures. The Sales department, however, has a forecast of 2,400, because a key competitor is about to launch a new competing product. Supply Chains forecast is 2,500 because sales were 2,500 this time last year. Meanwhile Finance has a forecast of 2,750 – because thats what is in the budget! The statistical forecasting model in the system says 2,673.425, based on an algorithm no-one outside Demand Planning understands. Sound familiar? Unless assumptions are recorded and then shared and reconciled at the Consensus Forecast meeting, there will be constant disagreement about the numbers. Without assumptions the story behind the numbers, the budget number always wins. So the forecast will fail in its purpose to support effective decision making that ensures overall business optimisation. Should we use constrained demand or unconstrained demand? In the forecasting process we should almost always be using unconstrained demand. By this I mean that the forecast is what Sales and Marketing want and expect to be able to sell at the appropriate margins. It can be considered a formal request to Supply Chain to make the inventory available. Constrained demand is where Sales could sell more but because of capacity constraints for example there is limited stock availability. The danger in this sort of situation of forecasting constrained demand is that there is no pressure on Supply Chain to remove the supply constraints, and so the business will enter a self-fulfilling prophecy of limited growth. Constraints to the forecast should only be applied later in the S&OP process and visibility of what the constraints are must be maintained right up to the highest levels of the business. Do Sales and Marketing or Supply Chain own the forecast? I think this is an easy one. Sales and Marketing should own the forecasting process they are closest to the customer and consumer and can use this insight to enhance the statistical forecast. From an S&OP process point of view this is also useful because the responsibility for making each step of the process work can then be shared out amongst functional directors e.g. the Marketing Director is responsible for Step 1 Managing New Activities, the Sales Director is responsible for Step 2 Managing Demand, the Supply Chain Director is responsible for Step 3 Managing Supply etc. You can make an argument that a Demand Planning role should sit in Supply Chain since that gives them more independence to robustly challenge the bottom-up forecast from Sales, but even in this case, or where Supply Chain manage the forecasting system, Sales and Marketing should own the final forecast. Where does customer collaboration fit and how important do you see the use of POS data? As anyone who has been at the upstream end of a supply chain knows, the demand signal will probably show more amplification here than at the downstream (consumer) end of the supply chain the Bullwhip Effect. So improving visibility of the ultimate demand signal (consumer shelf purchases) will help reduce these fluctuations and allow quicker response to demand changes. So collaboration with customers on forecasting initiatives, by sharing EPOS data for example, will definitely help improve the forecasting process. What is the difference between forecast accuracy and bias? Forecast accuracy is the difference between the forecast submitted say 3 months in advance of an accounting period and the actual sales for that period. Since over-forecasts are just as bad as under-forecasts, it is commonly measured by taking the absolute % deviations of the forecasts from the actual and averaging them (MAPE Mean Absolute Percentage Error). Taking the absolute deviation adjusts for the minus signs that would otherwise cancel out some of the pluses. Forecast accuracy then shows you the variability of forecasts, forecast bias on the other hand will show you whether particular sales people or business units are persistently over or under-forecasting. This is commonly caused either by pressure to make the forecast equal the target resulting in constant underachievement against target, or by a wish to always overachieve the target by putting in soft targets. In this case we need just Percentage Error (see the pluses and minuses) so we can see whether there are several periods of under or over-forecasting. Forecast bias should be considered at the Consensus Forecasting meeting and an adjustment made to the forecast before it is released to the Managing Supply step of the S&OP process.