There is a raging debate among planning professionals about DDMRP vs. forecasting. Proponents of DDMRP claim forecasting is dead. The primary reason cites is – poor accuracy. Votaries of demand forecasting say firms cannot work in a vacuum. Steve Allanson, from Veloc Limited weighs in on the debate from the perspective of FMCG companies. And comes up with an interesting conclusion.
The FMCG sector in particular (but not exclusively) has characteristics which are not compatible with at least two of the key tenets of DDMRP. A different, more sophisticated, approach is required – Can DDMRP accommodate this?.
In this article I will discuss a major feature of FMCG which must be addressed by any approach to Supply Chain Management (SCM) and planning. I will then discuss why two central tenets of DDMRP appear, to me, to be incompatible with this and will go on to discuss an approach to strategic and operational planning and also S&OP/IBP which has worked well in driving major improvements in Supply Chain and Business Performance in this sector. Finally I will examine how the Veloc planning solution supports this approach.
FMCG is predominantly Make to Stock (MTS) and demand must be planned at SKU level
FMCG today has four characteristics which need to be considered.
- Customers (retailers) demand ever lower lead-times from order to delivery. In the UK and Europe Day 1 for Day 3 is about the longest, Day 1 for Day 2 increasingly common and Day 1 for Day 1 (i.e. Order today for delivery today) is on the rise.
- Sales are increasingly promotion driven so that not only does a promotion by one supplier impact it’s own sales but promotions by the retailer can have a positive effect and promotions/activity by other suppliers and other retailers can have significant negative effects. The end result is a highly volatile demand pattern and one which requires significant collaboration between retailer and supplier to manage.
- SKUs continue to proliferate and diversify driven not only by increasing promotional activity but also by retailer demand for uniqueness. Additionally there are increasing requirements for allergen segregation driving increasing changeover costs and increasing cycle times.
- Retailers are increasingly demanding more and more shelf life to be available on delivery so that the shelf life available to the manufacturer for stock holding is decreasing. In the UK and Europe 75% of shelf life on delivery is almost universal with some retailers demanding even higher percentages. If a product has a shelf life of 4 months retailers will not accept product without 3 months remaining and if a product has a shelf life of 4 weeks – retailers will demand 3 weeks remaining on delivery. I have recently seen that in Canada/US this is considerably more lenient – for now!
The net result of these 4 factors is that the vast majority of FMCG is MTS based with a need to plan what is to be made in the absence of actual orders. Every product being made only once per week is most common with once per month still not unusual (I still see some blue chip businesses operating with some SKUs only made once per quarter!). Conversely promotional and other activity can have significant impacts week to week and even day today.
The most common pattern is that decisions have to be made about what to make next week in order to supply demand for that week and the following week. And this has to be done often without this all of this weeks orders being received and with few orders for next week and usually none at all for the week after.
This pattern is reflected throughout the inbound supply chain – with order lead-times from suppliers, even with buffer stocks in place, being measured in weeks not days. Raw material lead-times can be even longer.
Strategic buffer stocks – I prefer the term Safety Stocks – are essential at key points of the supply chain. In this respect DDMRP is correct – how those safety stocks are calculated, however, needs to be based on deviation of actual demand from demand plan rather than on total variability of demand.
If a business is to avoid investment in safety stocks being a significant cash drain and avoid high levels of obsolescence then Efficient Demand Management at SKU level is essential. Simply holding sufficient buffer stocks to cope with total variability WILL drive high levels of stock holding and obsolescence.
Demand Plans can be precise at SKU level
While it is now clear that DDMRP, DDOM and DDS&OP do embrace forecasting as part of the overall strategy there are still many advocates of the methodology who will state that DDMRP replaces the need for forecasting. The true position of the DDI appears to be that forecasting is indeed required but only at high level and only for strategic capacity planning and buffer sizing.
First of all let me clarify and differentiate Forecasting from Demand Management. These are my definitions as used over many years with many clients.
The net result of these 4 factors is that the vast majority of FMCG is MTS based with a need to plan what is to be made in the absence of actual orders
Forecasting is the use of algorithms of some variety to attempt to extrapolate history into the future. I am in complete agreement with DDMRP advocates that this is highly inaccurate except at high level. There are solutions which attempt to overcome this by the use of increasingly complex and obscure statistical algorithms to try to project complex (and often random) historic patterns into the future. I believe with respect to forecasting sales, particularly in FMCG, this is doomed to failure. At best forecasting will be able to accurately project an overall pattern of steady state, growth or decline coupled with some broad repeating seasonality.
Demand Planning is the act of overlaying activity and market knowledge based adjustments on the forecast. If a sales manager has agreed a promotion with a retailer then he should be able to project when and by how much this activity will impact sales. The role of the demand planner is then to examine similar historic events and challenge the assumptions of the sales manager. Together they should arrive at a consensus Demand Plan – at the SKU/customer level!
The role of the Demand Planner here is absolutely key and will overcome the tendency of sales managers to be overoptimistic and/or target driven. The single point of accountability for the accuracy of the Demand Plan must lie with the Demand Planner. The sales manager should be accountable for delivering the planned activity on time and to planned penetration and duration.
The DDI criticism of S&OP compared with the DDS&OP model appears to suggest S&OP is based upon a Top Down approach
Demand Management is the final piece of the process to achieve the best possible Demand Plan. Demand management is the combination of operational collaboration with customers plus strategic feedback loop from the S&OP/IBP process. With the features of FMCG I have described it is clear that, if we are to avoid high levels of buffer stocks, the Demand Plan at SKU level must be effective and must be used to make operational production decisions.
It is my experience that a properly structured Demand Management Process along these lines WILL yield, at SKU level, a level of variance from Demand Plan which is very significantly lower than the overall demand variability. The SKU level Demand Management Process improves the demand signal rather than diminishing its precision so that the safety stock requirement at each of the Strategic Buffer points is considerably less than if total demand variability is used.
The key to this is that the Demand Plan (along with everything else) is driven from bottom up, SKU and activity specific actions
S&OP/IBP must be built bottom up.
The DDI criticism of S&OP compared with the DDS&OP model appears to suggest S&OP is based upon a Top Down approach. The top level number (Which it is asserted is the single – and wrong – forecast number) is cascaded down through MPS and MRP. It is also suggested that standard S&OP then drives a massive change to the plan each month driven by changes to the forecast from the top. (1)
This is entirely incorrect. In fact it is essential that both S&OP and IBP are bottom up processes. Both begin with a consensus, activity based, Demand Plan which is combined with the Supply Plan required to deliver the planned sales along with any safety stocks.
The resulting operational plan is then run through the finance process to yield projected capacity, inventory, cash flow, profit etc. projections which are pushed up for discussion/ approval. IBP incorporates more areas of the business than S&OP and a longer time horizon but the essence of the process remains this bottom up approach.
If the projected business plan is not what is desired then it is fundamental to both processes that change is not made at the top and cascaded down. It must being with different activities, changed costs etc. to change the plan. Without this disciplined approach it isn’t a business plan – it is a wish-list.
The Monthly Cycle.
In both S&OP and IBP the monthly cycle should always begin, at each level, with a review and understanding of variances.
Have we done what we said we would do in the last month and year to date?
1a. If not then why not – what are the root causes of variance?
This question must be asked not just of internal operational parameters but also of the demand plan. This is absolutely key to improving the planning operations – Demand, Supply, Financial,
Are we going to deliver during the rest of the plan what we said we would?
2a. If not why not and what corrective actions need to be taken?
In this way we can see that, far from driving a monthly top down revamp, well implemented and well run S&OP and/or IBP actually drive stability and continuity of delivery.
The key to this is that the Demand Plan (along with everything else) is driven from bottom up, SKU and activity specific actions. Without SKU/Customer specific actions and changes there should be no deviation from the long term projection of a base historic forecast.
The continued analysis of variance from plan drives continuous improvement and reduction of variance.
The Veloc Approach
The first step is to decide which SKUs are volatile and require Demand Planning and which, if any, are stable and therefore can indeed have safety stocks based solely on historic variability.
For those SKUs which are volatile we calculate the required level of safety stocks, at all strategic points of the Supply Chain, by examining variance from Demand Plan. We usually find that the variance data approximates well to a Gaussian distribution and for a given level of safety (e.g. 98.5%) we can calculate the number of days cover required as Safety or Buffer stock. This same calculation is used throughout the B.O.M. ensuring that we plan dynamic safety stocks wherever they are required.
The Planning algorithm will look at the next available replenishment point and will calculate what needs making at each point in order to have the correct safety stock (if any) in place immediately prior to replenishment.
Passing from Plan Horizon to Operational Horizon
Each SKU can be set to either MTS or Make to Order (MTO). For MTO skus the anticipated future demand is used to reserve capacity for anticipated production or purchase and could drive actual purchase orders to maintain anticipated material, packaging or component safety stocks if the leadtime is long and safety stock buffer looks like it will be insufficient.
When it comes to the short term horizon (within a defined firm plan horizon) – where commitment to actual production orders or purchase orders is required – the system allows for sophisticated consumption of the Demand Plan by Actual Demand. This can be set for each SKU and can be very specific.
For MTO order SKU’s no production or purchase orders will be placed unless actual orders are present. In this way production or purchase orders for these SKUs are indeed as DDMRP would define as pull based.
For MTS SKUs the Demand Plan consumption rules will refine demand based upon the rules set.
For example if I know that customer A always places orders Monday to Friday at 20% per day then if I am planning on Wednesday I can assume 2/5ths of orders are already in and add to these 60% of the Demand plan. This gives my current estimate of the demand for the week.
Now, because the Veloc algorithm is based upon replenishing safety stocks, if there is deviation from the demand plan the only action required is to replenish the safety stock when the opportunity arises. We avoid the reactivity which DDMRP advocates insist will arise from all forecast or demand plan based systems.
The Veloc system uses the same dynamic safety stocks and the same Finite Capacity algorithm for finished goods, components and materials. In this way we avoid completely the usual plethora of “Schedule In”, “Schedule out” MRP nonsense.
I think the Veloc system delivers the optimal mix of Pull where possible, Pull with Demand Planning support where appropriate. At the same time the system delivers all of the information required to perform the S&OP/ IBP cycle so supporting business change.
I believe there are some difficult challenges with FMCG (which may occur on other sectors too) which are not compatible with some aspects of DDMRP – notably the assumption that Demand Plans cannot be accurate at SKU level and also the idea of a top down approach to S&OP.
I have outlined why, in my view, these are likely to result in service failure or high stocks/ high obsolescence in the FMCG sector and have outlined an alternative approach which in my experience does provide increasingly accurate SKU based demand plans and is fully compatible with S&OP/ IBP.
The Veloc Planning system is built around and proven to deliver, with this alternative, more sophisticated approach. ♦♦♦♦♦