- To improve accuracy
- To improve demand management process efficiency and productivity
- To support on-going improvement at demand management expertise
- To support best practices in Inventory Planning and Replenishment
- To support best practices in Sales and Operations Planning
- To significantly improve all matters having to do with IT that are involved with doing best practices demand management
FD7 Demand Management can be initially set up in a very simple way, and sophistication can be added later, to the level of the very best practices in demand management, factoring in all available information. An example of a simple setup would be automatic best-fit statistical forecasting followed by user overrides driven by factors outside of statistical forecasting, like the sales pipeline. Customers can easily advance their forecasting and demand planning process from the areas covered in the following paragraphs of this section.
The FD7 statistical function is an objective, statistical, extrapolative forecasting system that provides multiple methods, best-fit automatic expert selection, and interactive statistical modeling and management.
FD7 Supports Three Approaches to Statistical, Best-Fit Model Selection
- Tournament-based formula selection with user adjustment by exception
- Automatic, best-fit statistical analysis and forecasting with user adjustment by exception
- Interactive statistical analysis and forecasting
Automatic Statistical Forecasting Using THE World-Class Statistical Engine
FD7 incorporates THE world-class statistical best-fit engine. McConnell Chase licensed this engine and integrated it within FD7 because it won by a significant margin the most extensive, authoritative and recognized international statistical forecasting accuracy competition to date - the M3 competition. In the primary results of this competition this engine performed most accurately of all submissions for 16 out of 18 time horizons and beat the average submission by 11% for all 18 time horizons.
The FD76 automatic statistical forecasting engine runs in batch as well as interactively. It chooses from among a wide range of forecasting models and selects optimal parameter settings. It automatically determines the statistical applicability of seasonality, trends, and events such as promotions.
Interactive Statistical Modeling and Forecasting
Users can easily override automatic model selection, experiment with different models and settings, adding judgment, applying insight, and building complex models using independent, explanatory data. For example, users can specify statistical models such as:
- Box-Jenkins: ARIMA
- Exponential Smoothing: Twelve different Holt-Winters models
- Event models
- Curve fitting: straight line, quadratic, exponential and growth
- Low volume and intermittent demand models
- Models with non-normal error distributions, such as Poisson
- Build dynamic regression models interactively involving independent variables, lagged, transformed variables, Cochrane-Orcutt terms, promotional data, and event data
- Build event models to statistically adjust for events such as promotions
- Apply one of 5 different seasonality methods and/or apply group seasonality
- Adjust demand data and/or use various outlier conditioning methods
- Limit data by date or number of periods
- Optimize accuracy at a target time horizon, commonly lead time
To facilitate applying various factors to the forecast, FD7 features extraordinary hierarchy and view maintenance so that different personnel involved in your forecasting and planning process can apply their judgment, thinking, and business knowledge to forecasted expectations in the terms and at the levels of detail or aggregation that the business makes sense to them. FD7 provides excellent bottom up, top-down, and middle-out change capability.
Operating over the Web or over the company network, sales force individuals, dealing with customers, perhaps working collaboratively, can use FD76 to keep the rest of the company informed of current field intelligence. FD76 can also import sales rep forecasts directly from a company's CRM system and from customers' demand planning systems.
To streamline sales force forecasting FD76 supports updating on an exception basis.
To allow sales reps to work offline, FD76 supports an approach where reps connect via the web, download their territory data, work offline, and post updates when they are done.
FD7 provides a separate marketing forecast area so that marketing and sales managers can influence sales expectations by factoring in judgment and marketing events (the 4Ps of marketing) at the levels where they work, such as by market, market channel, product family within market, or by other levels as appropriate.
Marketing and high-level planners can have full visibility of current statistical forecasts, field intelligence, total market demand forecasts, projected market shares, market and economic indicators, new product plans, strategic initiatives, key performance indicators, expected economic, competitive, and promotional events, and all assumptions.
FD7 provides special functions for New Product Forecasting, Substitutions, Obsolescence, and Product Transitioning.
FD76 provides for entry, maintenance, and application of economic, competitive, and marketing events and assumptions. Events and assumptions are typically maintained by the people best suited to gauging the timing, size, duration, and probability of an event's impact on volume and sales revenue.
Managerial input is lastly captured in final demand manager edits and maintained separately as the 'management forecast'. The management forecast area is informed by all prior steps in the demand planning process. Management edits are often influenced by considerations from supply, capacity, cost, YTD performance against the budgeted financial plan, and company strategy. The management forecast is where demand management and sales and operations planning collaborate to yield a company-wide consensus on future sales expectations.
Each period FD7 can archive each set of forecasted numbers as well as the assumptions and thinking behind the numbers. This archiving can take place at every level of detail and aggregation and for each set of forecasts produced as part of the customer's demand management process: statistical, sales force, marketing, sales management, and demand management final edits.
Each period FD7 can flag many types of exceptions, such as accuracy and bias, and it can prioritize them as feedback for improving forecast quality and accuracy. In the practice of forecasting the areas of largest error need to be appreciated as the areas of greatest potential improvement. Using exception flagging constructively, FD7 provides an efficient, prioritized process for getting better at setting expectations, gradually learning from mistakes, gradually calibrating the factors that drive demand and performance, and gradually reducing forecast error.