Distributions

After the forecasting methods have determined the volumes for each day of the week, the system needs a way to spread that volume across the hours of the day. Distributions are created to fill this need. The system comes with two types of ready-made distributions:

  • Flat distribution evenly spreads the volume across the day.
  • Best Fit distribution places non volume-driven workload onto the schedule where the workload fits best into the schedule. Best Fit allows the system to schedule that workload at times when the scheduled shifts are not covering volume-driven work.

In addition to these distributions, these distributions can be created within the system.

Automatic distributions

Automatic distributions analyze historical results data to predict how a driver performs during each day of the week and time interval. These types of automatic distributions are supported:
Machine learning
The machine learning model generates distributions during training. Drivers must be configured to use machine learning to use this type of automatic distribution. The distribution is generated by analyzing the past 8 weeks of historical data. The distribution provides a ratio for each interval on each day of the week. These ratios are used to distribute daily forecast volumes. The machine learning model can compensate for gaps in historical data when creating automatic distributions.
Note: Inverse distributions are not supported for drivers using machine learning.
Moving-average automatic distribution

This type of automatic distribution is generated for drivers using any forecast method other than machine learning. It calculates an automatic distribution from a moving average of the past 8 weeks of historical data. The results of this moving average are used to create a ratio. The ratio spreads the forecast volumes within each day of the week.

The Normal distribution mode uses the curve of the workload through the day to schedule the work. This mode is the most common distribution mode because the workload follows the curve of the sales, traffic, and other factors.

The Inverse distribution mode flips the curve to schedule workload at the times when traffic is low. The Inverse mode can be used to schedule cleaning or recovery tasks that should happen when fewer customers are in the store.

Day Part automatic distributions

These distributions are required for use with the Workload by Day Part page. This distribution is treated the same way as a flat distribution unless edits are made to the Sales information by day part in the Workload by Day Part page.

After edits are made, the distribution details are automatically updated to maintain the distribution of the driver's forecast across the day parts. When selecting this distribution type, the distribution's location must equal the sales driver with which the location is associated. The distribution’s location can not be associated to the store level location.

Historical results distribution

There are times when the automatic distribution method may not be an accurate form of distribution. For example, the day after Thanksgiving (Black Friday) may have a very different distribution pattern than the previous 8 weeks of October and November. In cases like Black Friday, use a distribution based on a day known to have a distribution pattern similar to what you expect for Black Friday (for example, the previous year’s Black Friday). The Historical Results distribution method calculates a distribution just like the automatic distribution, but uses the dates that you enter instead of using the past 8 weeks.

Manual distributions

When none of the automatic distributions can be used, a manual distribution may be the best option. This option allows you to specify the percentage of workload manually for each interval of the day. This option is rarely required.