Forecasting and MRP

Forecasting in the material plan

Forecasting in the material plan is coded as order category 010. As opposed to other transactions, the forecast transactions are not read from a file, but calculated on-line in (RPS950) as you open the material plan. Two factors have the greatest effect on the forecast in the material plan. They are:

• Period frames
• Forecast logic

The period frame set in 'Planning Policy. Open' (MMS037) is used to set the time buckets into which the forecast should be placed. The period frame is updated for each item in 'Item. Connect Warehouse' (MMS002). Due to the dynamic possibilities in creating time buckets, you may have anything from daily to yearly buckets. The quantity you have forecasted is always broken down to daily quantities and then added to the number of working days in the bucket. In the bucket, the position of the forecast quantity could be either in the beginning, middle or according to a point of time given by the transaction type.

The distribution point parameter in 'Forecast Logic. Open' (FCS305) determines where the forecast will end up. The forecast must be put on a defined delivery day in the system calendar. If the forecast quantity in the bucket belongs to a day that is not defined as a delivery day, it will be placed on the previous defined delivery day as a separate record.

If the forecast quantity is not divisible in even units for the days, and the number of decimals for the item's basic unit of measure is less than the decimals for the forecasted quantity, the number must be rounded off. Rounding is done according to following rules:

• If the remaining decimal is more than 5, the number is rounded up. Then, starting from the first day in the bucket, the total forecast quantity is distributed until the entire forecast for the period is used up. Consequently, the last period(s) might not have anything left to distribute.
• If the remaining decimal is less than 5, the number is rounded down. The leftovers are placed in the last bucket of the forecast period.

Forecast logic

The forecast logic is used to control the forecast implementation in the material plan. Some important time fences make this possible. The picture below illustrates these fences.

Forecast consumption is a key issue when setting parameters. The theory of forecast consumption is basically quite simple. If you have a forecast in a period, the customer orders consume the forecast as they are entered. Since the forecast is just that, it most likely will not match the customer orders 100% in any period. Businesses treat the difference between actual orders and forecasts differently. They either decide that the market has a memory — for example, if you surpass or do not consume as much as the forecast in one period, you assume it will even out in the following periods. This is then considered in your planning. Conversely, you can say that the market has no memory and therefore don't consider over or under consumption of the forecast.

M3 offers several possibilities with parameter settings for forecast consumption. This will help you create a setting that matches the business requirements for many companies.

• Consumption History Fence (2) – the number of days in the past (defined in forecast logic) which will be compared to current forecasts and consumption values. This allows forecasting to adjust to actual consumption for certain periods.
• Current day – the system date on the computer which is always used as a reference when calculating the other fences.
• Demand Time Fence (1) - The demand time fence (DTF) is entered in Work with Item/Warehouse (in calendar days) and controls some of the forecast consumption functions. The theory is that customer orders, not the forecasts, matter in the short term. The DTF ignores the forecast when calculating availability. (This depends on the parameter setting in M3.)
• Forecast Consumption Fence – The forecast consumption time fence is entered in 'Forecast Logic. Open' (FCS305). Consumption stops at the end of the defined consumption fence.
• Planning Horizon – The planning horizon is entered in 'Item. Connect Warehouse' (MMS002) and controls how far into the future (in calendar days) material plan transactions are displayed. This means that a forecast is not displayed outside the planning horizon.

Parameters in forecast logic

• Demand time fence management

  This parameter controls how the forecast and the consumption are treated between the current date and DTF. Valid values are:

  0	The forecast, the consumption, and if included, the history, are implemented into the material plan at the current time. In this case, the forecast surplus from history (59-50 = 9) is implemented into the material plan in the first bucket after the current date. The forecast in the material plan before the first bucket becomes 14 (25+9-20).
  1	The forecast, the consumption, and if included, the history, are implemented into the material plan after the demand time fence date. No forecast is implemented inside the DTF. When the parameter is set to 1, the forecast may not be implemented into the material plan until after the DTF. Therefore the history surplus of 9 is put into the memory of the first bucket. The memory is then implemented into the forecast in the first bucket after the DTF. The forecast becomes 39 (25+14) in the material plan.
  2	The forecast, consumption and if included, history, are implemented into the material plan at demand time fence date. But in this case the forecast, consumption and history are not carried over to the first date after the DTF as above. This function requires that parameters 7 and 8 are set to zero, otherwise 1 and 2 are the same.

  Note: Recommendations using Demand time fence management

  • Consider F/C consumption in past day

    Enter the number of calendar days that should be the time span for the Consumption History fence.

    Before the given time period, the system will calculate both the forecast and the consumption. The difference between these two may be included as a plus or minus in the next period. The consumption history fence is calculated either from the current date or from the DTF. The one used is determined by parameter 1.

  • Drop F/C consumption in past with period break

    Instead of using only the days to define the "look back" horizon, you may also consider a number of period breaks. If the given "look back period" period crosses one period break date line, consumption will stop. The number of period breaks to be crossed may be defined in this parameter. Up to 9 breaks may be defined with this parameter. The relevant periods are the MPM-periods defined by division.

  • Break date calendar for historical forecast consumption

    The period breaks described above are defined by period frames. With this parameter, they may be overruled with a unique table. By prompting on this field, you may select or create a table of break dates that should be used instead of the "normal ones." These break dates are only used for historical consumption.

  • Classify old reservations as non-consumable

    The parameter controls whether old transactions in the material plan should consume the forecast.

  • Past consumption quantity accumulation method

    Choose how to calculate the quantities that have consumed the forecast in the past. Your options are:

    • Use the item statistics data (MITSTA-file) and calculate an average usage during the considered history period. Example: Current date: 981101 – Look back period 10 days. Accumulated sold quantity for 9810 is 100. The consuming quantity will then be 100/31*10=32. 31 for the number of days in October and 10 from parameter two in the forecast logic. This may of course be different if break dates or some other parameters are turned on to perform other functions that will put this out of order.
    • Read the history file in M3 (MITTRA) and add the transactions that should consume forecast (over the consumption history fence). Transactions that should consume forecast are defined at the items 'Forecast Method. Open' (FCS300). Also, consuming transactions are only those that are coded as representative usage by order type.
  • Implement possible F/C surplus from past

    If the history bucket gives a forecast surplus (forecast quantity-consuming quantity), decide whether this should be added to the forecast memory and added to the total forecast.

  • Implement possible consumption qty from past

    If the history bucket gives a consumption surplus (forecast quantity-consuming quantity), decide whether this should be added to the forecast memory and added to the total consumption in order to lower the forecast for the upcoming periods.

  • Set forecast memory to zero – days

    When we build up a forecast memory this may be set to zero with a given day interval. Specify how often you would like to set the memory to zero.

  • Forecast memory if period break

    This controls whether the forecast memory should be set to zero. You options here are 0 and 1. With 1, memory is reset to zero at every period break.

  • Forecast consumption - look back adjustment

    If you have an even forecast but a uneven order quantity situation this parameter helps you reduce forecast in close time due to an order that is further away. Your options are:

    0. No adjustment made

    1. Adjust by reducing forecast starting from the current period and onward.

    2. Adjust by reducing forecast starting from the given period and going backwards.

    Example: Current date 980930

    	Period	Period	Period	Period
    	9810	9811	9812	9901
    Forecast	100	100	100	100
    Consumption	0	0	0	250
    Memory (1)	100	50	0	0
    Net forecast	0	50	100	0
    Memory (2)	0	50	100	0
    Net forecast	100	50	0	0

    With the code set to 1, 250 will reduce the forecast in earlier periods starting in period 9810.

    With the code set to 2, 250 will reduce the forecast in earlier periods starting in period 9812.

    Note: The period that holds the order in this case 9901 will always reduce its full forecast. Whatever is left will then be reduced in accordance with the given rules.
  • Production program - forecast consumption logic

    The transactions that reduce the forecast are determined by the forecast method. With this parameter you may overrule that setting and report only manufacturing orders that reduce the forecast. This may be used if you use the forecast as a production program and if your customers call-off their orders.

• Forecast consumption time fence

  Enter the number of days after which customer orders will not reduce the projected on-hand balance in 'Forecast Method. Open' (MMS080). Entering 0 or leaving the field blank means you don't have consumption after the fence. The transaction will then be displayed in the material plan but with no reduction of projected on hand balance. If you prefer not to see these transactions, set the Display Reservations after Consumption Time Fence parameter. You may also turn the consumption on and off with the No Forecast Consumption parameter.

• No forecast consumption

  Setting this parameter to 1 will set the forecast consumption fence to 0 days and hide the transactions.

• Display reservations after consumption time fence

  With this parameter set to 1, you will display transactions in the material plan even though they don’t consume the forecast.

• Source of demand - material plan calculation

  When the material plan is putting a forecast into to the plan, it may be retrieved from three different places. Here you specify in what order to look for the forecast. The valid options are:

  0. No Source

  1. Manual forecast

  2. Automatic forecast

  3. The blank MPS-version

• Distribution point

  The field indicates where the forecast requirement should be placed within the current time period. That is determined by the applicable period template.

  The valid alternatives are:

  1. Start of period

  2. Middle of period

  3. End of period

• Weekend management

  The parameter controls if the forecast (and all other transactions in the material plan) should be placed on Saturdays and Sundays in the material plan. The parameter is overruled by the delivery day parameter in the system calendar (only one way, meaning that transactions can never occur on a day if this is not a distribution day in the system calendar). If you have forecasts on Saturday and Sundays but these are not delivery days in the system calendar, the forecast is added to the previous delivery day.

• Daily forecast distribution table

  The field indicates the ID of the table for a daily forecast distribution. The table is connected to items through the forecast consumption method in 'Forecast Logic. Open' (FCS305), which is entered per item/warehouse. Valid IDs are defined in 'Distribution Table. Open' (FCS330).

  The entered table ID automatically distributes a weekly forecast, based on the percentages in the table, to the daily level in the material plan. You can override this by defining the distribution of weekly forecast to the daily level in 'Manual Forecast/MDS. Open' (FCS350), sorting orders 5 and 6.

  If nothing is entered, it is assumed that daily distribution is not allowed for this forecast consumption method.

• Activate net change interval

  The field indicates the range of days to be used to calculate forecasted requirement planning items, if no transactions have been made since the last calculation. If transactions have been made, calculations will occur automatically without consideration to the set range of days.

  The valid values are:

  0. The function is not used. No recalculation due to time passed will occur.

  1-99. Recalculation of MRP made according to the specified range of days due to time passed.