I would like to share some experiences related with Sales & Operations Planning (S&OP) with you and to open a discussion forum. The majority of people think "Demand Planning" is equal to S&OP. Demand Planning is just one of five steps of the whole S&OP process.
The Demand Planning step uses the statistic sales forecast and the experience of other areas in order to estimate future demand. S&OP is a communication and decision making process whose main goal is to balance offer, demand, mix and volume. The buffer to achieve this balance is inventory and orders backlog. It integrates financial, supply, sales and development plan.
The S&OP process consists of five steps:
1)Forecast
Deliverable: Statistic forecast
2)Demand Plan
Deliverable: Demand plan agreed upon with Sales, Marketing and, in some cases, with end-customers.
3)Supply Plan
Deliverables: Production high-level plan, raw materials purchasing plan, resale products purchasing plan and report of supply restrictions such as plant capacity and raw materials availability.
4) Pre-S&OP Meeting
During this meeting the demand and supply plans are reconciled. Issues are communicated. Human Resources and Financial first recommendations are drafted.
5) S&OP Executive Meeting
A summary of the Demand and Supply plans are presented to the CEO accompanied by a summary of recommendations.
Important things to take into account
For the forecast:
-Quality of information (sales): Garbage in, Garbage out
-Aggregation level: Each demand must be treated in a different way. One must choose the best aggregation level in order to get the best output, for example, by business unit, family, subfamily, SKU, customer, family-SKU.
-Planning horizon: S&OP is used for long-term decisions. Ideally 36-month history must be taken in order to plan 18+ months ahead.
-Clean history: Keep a logbook in order to clean outliers and non-repeatable events.
-Adequate forecasting model: there is an appropriate model for each type of demand (seasonal, intermitent, etc.).
-Software: Choose the right software for your needs (Excel, Forecast Pro, Demantra).
For the Demand Plan:
-Revision of sales history versus past forecasts.
-Variation Coefficient (VC) – I recommend treating differently those products with a VC < 0.75 (stable products) than those with high variability (VC > 1.3). Do not waste time reviewing the forecast of stable products; focus on figuring out why high variations are given.
-Promotions
-Discounts, price elasticity
-Analyze sell-out and inventory level at the point of sale
-Life-cycle stage of products
-Product cannibalization
-Macroeconomic variables
-Special projects
It is recommended to visualize information in different columns, the base forecast (BAU or Business-as-Usual) in one column and building blocks in others to visualize the reason of deviations from the base. Examples of building blocks are: Promotions & campaigns, launches, price elasticity, Marketing initiatives, projects, new stores and new SKU's.
Demand Plan accuracy is directly impacted by the forecasting / collaborating technology used, the Demand Planner's understanding of the forecast model (linear regression, exponential, Brown, Holt, Winters, ARIMA, Box-Jenkins, etc.), trend and patterns (seasonal, cyclical, stationary, random), how close the planner is to the demand source, aggregation level for intermitent demand, strategy used for each variability segment, sell-out observance and overall, the level of collaboration with internal and external customers.
In the case of home centers and supermarket chains, it is recommended to have a very close relationship with the purchaser in order to achieve a Collaborative Forecast.
For the Supply Plan:
Current inventory level, batch and lot size, inventory objective to start the following month calculated in days of inventory with next month's forecast.
It is very important to provide feedback to Sales of potential supply restrictions.
Final Recommendations
Measure WMAPE (Weighted Mean Absolute Percent Error). The majority of books and experts recommend using MAPE (Non-weighted Mean Absolute Percent Error); however, MAPE gives the same weight to a 200% deviation over a product whose forecast is 10 units than a product with a 200% deviation over a product whose forecast is 1,000,000 units. The second one will have a greater impact on the manufacturing plant; therefore, we need to understand the source of deviation first.
Include a postmortem analysis of deviations to forecast as part of the routine.
Include portfolio rationalization and batch-size revision meetings as part of the S&OP schedule.
Measure production plan accuracy (equivalent to WMAPE).
Keep collaborating, discussing, validating and challenging the plans.
What is your experience?
