Unit 4: Forecast

Lesson 1 · Explaining Basics of the Forecast

Prerequisites

• Forecast parameters in the material master (Forecasting view) – forecast model, period indicator, number of periods, etc.
• Historical consumption values. They are updated automatically by goods issues; you can check and correct them before forecasting and enter history manually for a first forecast.
• Whether a movement type updates planned, unplanned or no consumption is defined in movement type Customizing (OMJJ). E.g. a goods issue to a cost center (201) updates unplanned consumption if it was not reserved.
• Without history, use a reference material (another material, or the same/another material in another plant) for consumption values.
• The period indicator sets the forecast period: monthly, weekly, daily or per fiscal year variant (posting periods).
• Forecast results only affect MRP if the forecast indicator of the MRP type is set (OMDQ).

Forecast models

Model selection

Lesson 2 · Executing and Postprocessing the Forecast

First forecast

Model initialization

Determines model parameters (basic value, trend value, seasonal indices). It happens automatically in the first forecast and must be repeated after a structural break when the previous model no longer fits. The number of initialization periods can be entered in the Forecasting view.

Ex-post forecast

A forecast for past periods. If more history exists than initialization needs, the history is split: the older part initializes the model, the newer part is "forecast" in parallel with actual consumption, and the comparison tunes the parameters to recent behaviour. It is also used when skipping periods, evaluating forecast quality and optimizing parameters. Only possible with exponential smoothing models.

Exponential smoothing

The constant model uses first-order exponential smoothing: the newer a value, the more weight it gets. The new basic value is computed from the previous basic value, the latest consumption value and the smoothing factor alpha: new basic value = previous basic value + α × (latest consumption − previous basic value). With α = 0.5 the most recent value weighs 50 %, the one before 25 %, then 12.5 % and so on. The forecast value equals the basic value of the last ex-post period.

Parameter optimization

If set in the material master, the system simulates ex-post forecasts with gradually increasing smoothing factors (step size = optimization level). A second pass refines the area with the lowest mean absolute deviation; the combination with the lowest MAD wins. Only for exponential smoothing.

Executing the forecast

Postprocessing

• Online check: errors/exceptions in the total forecast are logged as exception messages assigned to error classes; faulty materials are marked for postprocessing and can be selected by error class.
• List check: print the forecast results online (MP39) or in background (MPDR); select "Log" to include procedural and exception messages.
• Example exceptions: no history; safety stock not calculated because service level is 0; consumption pattern changed.

Influencing the result

• Correction values adjust historical values (e.g. outliers) or forecast values; corrected forecast values are firmed automatically and not overwritten by later runs. Forecast values can also be fixed manually.
• Number of historical periods: default and maximum 60; fewer periods limit the history considered.

Error measurement

Lesson 3 · Performing Forecast-Based Planning

Key concepts

In forecast-based planning (standard MRP type VV) the forecast values become requirements for MRP. Proposals are created when cumulative forecast demand up to a date exceeds cumulative firmed receipts (stock, production orders, POs, firmed planned orders and requisitions) up to that date. No planned independent requirements need to be maintained. Forecasts are re-run regularly, so demand follows actual behaviour; every goods withdrawal reduces the forecast requirement so the consumed part is not planned twice. Make sure all past consumption is posted before forecasting new periods.

Forecast requirements in MRP

• Each period's forecast requirement is dated on the first working day of the period.
• The splitting indicator breaks coarse forecast periods into a finer MRP pattern (e.g. monthly into weekly) and defines how many forecast periods are included in MRP. Reductions are spread proportionally. Defined per plant and periodicity in OMDG and assigned in the material master.
• Only forecast requirements count as issues; sales orders and reservations are displayed but not netted.
• The share of safety stock available for netting is set in the MRP group (OPPZ).
• Dynamic safety stock via a range of coverage profile in the material master.
• MRP Live hands VV materials to classic MRP. Forecast demand is semantically close to planned independent requirements, which can be created from it.

Backward scheduling

If backward scheduling yields a start date in the past, the system switches to forward scheduling, unless the plant parameters allow start dates in the past. As an alternative to safety stock, a safety time (MRP 2: safety time indicator + days) brings requirement dates forward.

Lesson 4 · Performing Time-Phased Planning

Key concepts

Time-phased planning plans a material only on fixed days – ideal when a supplier delivers on, say, every Tuesday. In the material master you enter a time-phased MRP type and a planning cycle (a planning calendar defining planning days). Planning calendars are maintained with MD25 and can also be used for scheduling (lot-size PK).

The planning file holds a planning date for each such material, set when the material is created and after each run, calculated from the planning cycle. The run plans the material only on that date. You can also enter a planning date in the run to plan earlier (e.g. Friday instead of Monday).

Net requirements interval

Planning date + planning cycle + purchasing processing time + planned delivery time + GR processing time

Forecast requirements in this interval are compared with stock plus fixed and firm receipts, and the order quantity is calculated. Lot-for-lot is the usual lot size; optimizing procedures are also possible. Because fixed receipts are assumed somewhere within the interval, temporary shortages can occur.

Combined with a reorder point

The material is planned on its planning date or when a goods issue drops stock below the reorder point. In the latter case the interval runs from the shortage to the availability date of the next regular planning date; afterwards regular cycles resume.

Lesson 5 · Performing Automatic Reorder Point Planning

Key concepts

In manual reorder point planning (VB) you enter reorder point and safety stock yourself; in automatic reorder point planning (VM) the forecast run calculates both and writes them into the material master. Because the forecast runs regularly, both values adapt to current consumption and delivery behaviour – which tends to reduce stock.

Reorder point = safety stock + average daily requirement × RLT

Safety stock is calculated from the RLT, the service level (converted into a factor of the normal distribution) and the mean absolute deviation. The service level is the share of demand that stock should be able to cover; the higher it is, the larger the safety stock. Allowed range 60 %–99.9 %; 90–95 % is a good starting point to experiment with.

A minimum safety stock in MRP 2 sets a floor: if the calculation results in less, the minimum is used.

Unit summary & exam focus

Forecasts need parameters and history; models (constant, trend, seasonal, seasonal trend) are chosen manually or automatically, initialized, tuned by ex-post forecast and optimization, and monitored with error total, MAD and tracking signal. Forecast-based planning nets forecast requirements with backward scheduling; time-phased planning plans on calendar days over a defined interval; automatic ROP planning lets the forecast compute safety stock and reorder point.

T-codes in this unit

Self-check questions

1. How many historical values does the seasonal model need at least?
   Answer

   13 (seasonal trend: 15).
2. What is an ex-post forecast?
   Answer

   A forecast for past periods based on older history, compared with actual consumption to tune model parameters; only for exponential smoothing.
3. How is the tracking signal calculated and what is the default tracking limit?
   Answer

   Error total divided by mean absolute deviation; default limit 4.
4. Which indicator defines how many forecast periods enter MRP in forecast-based planning?
   Answer

   The splitting indicator (OMDG).
5. In forecast-based planning, are sales orders netted?
   Answer

   No; they are displayed but only forecast requirements are considered as issues.
6. Give the interval formula for time-phased planning.
   Answer

   Planning date + planning cycle + purchasing processing time + planned delivery time + GR processing time.
7. Which data does the system need to calculate safety stock in automatic ROP planning?
   Answer

   Replenishment lead time, service level and mean absolute deviation.