Forecast the demand for your resources to resolve your ongoing influx of tasks

[Rekha Raman]

Have you wondered how you can staff the right number of agents to meet your expected demand to resolve tasks over time? 

SUMMARY 

In the Workforce Optimization for ITSM application, using historical data from multiple sources, you can accurately forecast the resources you’ll need for the work your teams do. After you create the forecasts, you can calculate the resources based on performance analytics data and work schedule trends. Once you determine the number of agents required, you can schedule them on the team calendar to ensure that they are always available. You can also manually tweak the forecast, analyze the pattern in a time-series visualization, and use that data for predicting your resources. 

DEMAND FORECAST FLOW 

Here’s the flowchart that shows how to implement Demand Forecast. 

Introduction

As a manager with the Shift Planning Admin role (sn_shift_planning.admin), you can create schedules for your team and monitor them. When you want to foresee the number of agents you need for future shifts, for example, how many agents you would need during the Christmas season for the upcoming year, is there a way to predict the expected demand?  

Demand Forecast in Workforce Optimization for ITSM lets you do just that! 

This article walks you through implementing Demand Forecast in Workforce Optimization for ITSM. 

As an admin, you configure demand forecast. 

1. Activate Demand Forecast
2.  Configure Demand Forecast
   2 a. Define historical data
   2 b. Add formula parameters 
   2 c. Add resource conversion formulas
   2 d. Associate the configured formulas to groups
   2 e. Run schedule jobs

After the admin configures Demand Forecast, as a manager, you can tweak the forecast parameters and visualize the forecast in a time-series chart. You can also make manual adjustments to fine-tune them for greater accuracy. 

3.Tweak forecast parameters
4. Manually adjust the forecast

1. Activate Demand Forecast

You must activate the following plugins to use Demand Forecast: 

1. Workforce Optimization for ITSM plugin [sn_wfo_cfg_itsm] — Allows you to configure Demand Forecast
2. MetricBase plugin [com.snc.clotho] — Allows you to collect data over time. 

2. Configure Demand Forecast

Predicting the number of agents that you need to staff for upcoming expected demand is based on the conditions you set to capture historical data.  

For example, if you want to predict the number of P1 incidents that will be assigned to Team A for the next 3 months, you must first collect the historical data on how many P1 incidents were resolved in the past 3 months by Team A. This will help you plan how many agents you need in future to resolve P1 incidents for the same time period. 

2 a. Define historical data 

To define the historical data that you want to collect, in this topic, follow the steps in the Define the data that you want to collect for forecasting the agents that you need section. You’ll use the Data Collection Definitions application to define the data you want to collect for forecasting agents. 

2 b. Add formula parameters

Add parameters used in the formula to calculate the resource count per hour in a day. For example, you can add Average P1 Incident Work Time as a parameter and add a value to calculate the parameter that you will use in the next step when you create the formulas. Follow the steps in the Add parameters for the formula to calculate how many agents you need per hour in a day section. 

2 c. Add resource conversion formulas

Create formulas to convert forecast to resources. For example, you can create formulas to convert the expected number of P1 incidents into the number of resources in your team you would need to respond to those incidents. Follow the steps in the Configure the resource conversion formula section. 

2 d. Associate the configured formula to groups

• Associate the configured formula for resource conversion to assignment groups. For example, if you create a formula to convert the expected number of P1 incidents into the number of resources in your team, you must associate the formula with the assignment group that will work on the P1 incidents. Follow the steps in the Associate an assignment group with a resource conversion formula section.  
• You must also define the minimum or maximum number of agents required per hour so that you always have the desired staffing coverage. Follow the steps in the define the minimum or maximum number of agents required per hour topic. 

2 e. Run these scheduled jobs

Schedule job name	What does it do?	Example
Collect daily data for automated forecast configurations	Gathers the data for the metrics defined in the data collection definitions.  Auto scheduled to run daily.  Fetches the records from the previous day for each hour and for every assignment group and stores it in the MetricBase. You can access this data using the MetricBase list on the Group [sys_user_group] table.	Running this job collects the data on all resolved P1 incidents daily.
Collect historical data for automated forecast configurations	On-demand job to collect hourly historical data.  Runs once per job.  Collects data for the past three years.	Running this job gets all the P1 incidents resolved in the last 3 months.
Forecast resources for future	Forecasts resources based on the collected data.  Auto-scheduled to run daily.  Stores data in the Agent Forecast metric in the MetricBase.  Set the collection frequency using the forecast properties. You can use this information to calculate the forecast or the time period for which you want to store the data.	Running this job forecasts the resources based on the data you’ve collected.
Note: The time series metrics created for data collection definitions use the WFO Forecast retention policy. The forecast intervals can be configured in 15, 30, or 60 minute intervals. This policy stores data at a one-hour interval for the past three years.

You can see the forecasted demand on the calendar.

3. Tweak forecast parameters. 

Tweak the forecast parameters, such as the number of P1 incidents, to see how the forecast behaves when you vary the period length, periods to forecast, or algorithm and publish them. You can view the modified forecast on the time-series visualization.  
Note: The start date for a forecast configuration is based on the number of historical days you want to consider for data visualization. This is set using the sn_agent_forecast.number_of_historical_days_in_timeseries_chart system property. 

Follow the steps in Modify forecast parameters to visualize forecast data topic to tweak the parameters and visualize the forecasts. 

4. Manually adjust the forecast. 

Adjust the forecast manually to fine-tune them for greater accuracy and see how it impacts your forecast. 

Follow the steps in Create a manual adjustment for a forecast to create the manual adjustments. 

Conclusion 

Now you can predict the right agents you need at the right time to resolve the constant stream of incoming work items. 

Troubleshooting

When you encounter this issue	Follow these steps for resolution
Cannot create forecast configurations.	Make sure your application is in Global scope.
The calculated forecast is inaccurate.	If the resource conversion formula uses two different forecast configurations from a data collection definition, the calculation considers all forecast parameters set for each of the forecast configuration in a resource conversion formula and uses the one with the lowest forecast parameter values for the calculation. For more information, see forecast parameters in demand forecast.
Historical data not displaying in the forecast.	Make sure the Collect historical data for automated forecast configurations scheduled job is run. For more information, see Demand Forecast reference.
Historical data not displaying in the forecast after running the scheduled job.	Make sure that the resource conversion formulas for the data is associated with assignment groups. For more information, see forecast parameters in demand forecast.
Forecast data is not calculated even if the historical data is available.	You need at least three times the historical data available for forecasting. For example, if you want to forecast for the next 30 days, you must collect at least 90 days of historical data.
Unable to edit a published forecast parameter.	Each forecast parameter is created for a specific time period using a specific algorithm. To preserve the history of all configurations used in each forecast parameter, you will not be able to edit a published forecast parameter. You can create a new forecast parameter with the desired values and publish it, which makes the current, published forecast inactive. For more information, see forecast parameters in demand forecast.

References 

• Demand forecast in Workforce Optimization for ITSM (Documentation) 
• Model demand scenarios (YouTube) 
• Workforce Optimization Essentials – Utah release (Training)