Concept: Time Anomalies

Time anomalies are time blocks that, for one or more reasons, are atypical for a worker. Workday uses machine learning to determine workers' normal time entry patterns from historical data based on:

Time entries.
Time off.
Work schedules.
Whether time approvers approved, denied, corrected, or sent back similar time blocks.

Workday identifies time blocks that violate these patterns and assigns the time block a score. This score quantifies the likelihood that the time block is anomalous. When Workday identifies a time block as anomalous, it doesn't mean that the time block is invalid. Rather, the time block is atypical and might require additional attention.

Once Workday identifies anomalous time blocks, you can:

Create custom reports to view time anomaly scores and top time anomaly reasons for time blocks.
Access the
Time Anomaly Trends
report to view time anomaly trends by supervisory organization.
Enable managers and timekeepers to see and take action on time anomalies within the
Review Time
report.

Time anomaly detection is meant to supplement time entry validations. Time entry validations and rule conditions are the best way to catch known errors. Alternately, time anomaly detection helps you identify unusual time entries that don't violate validations.

The machine learning model observes the actions (approving, denying, correcting, sending back) taken by managers, administrators, and workers. Using this data, the model learns what kinds of time entries are acceptable or unacceptable. Example: A manager regularly approves a worker’s time blocks from 8:00 AM-5:00 PM on Monday through Friday. Over time, Workday learns that 8:00 AM-5:00 PM blocks are acceptable. Alternatively, a worker repeatedly enters time blocks from 7:00 AM-11:00 PM, and the manager repeatedly sends back the time blocks. Workday learns that this behavior isn't acceptable and assigns a higher anomaly score to future similar time blocks.

As Workday observes more and more actions, the machine learning model more accurately identifies anomalies. However, this means that Workday isn’t able to detect anomalies if a manager or administrator has never taken action on them before. Example: A worker incorrectly submits time on a Saturday from 1:00 AM-5:00 AM. Since enabling time anomaly detection in the tenant, workers haven’t entered any similar time blocks, and administrators haven’t taken action on similar time blocks. Workday doesn’t flag this time block as anomalous. Once the manager takes action on the block, Workday will begin to learn that this block and similar blocks are unacceptable. Workday then assigns a higher anomaly score to similar time blocks.

The model trains on historical time data for each worker. To have sufficient data for the model to analyze, Workday recommends you deploy Time Tracking at least 1 year before implementing time anomaly detection. We recommend you opt in to the workforce management machine learning service and data categories in a Production tenant so that the model can train with live data for at least 3 weeks. You can monitor the results in the

Time Anomaly Trends

report. Wait until anomaly results are consistent before you make them visible in the

Review Time

report. To hide time anomalies from managers and time approvers during this period, don’t select the

Include Time Anomalies

check box on the

Edit Time Approval Template

task.

Testing this feature in a nonproduction tenant has limitations because the model needs a constant stream of live data to learn the patterns of acceptable and unacceptable time blocks. When you opt in to time anomaly detection in your Production tenant, the machine learning model can access more data and quickly improve accuracy. If you need to test time anomaly detection in a nonproduction tenant, you can enable the feature in Implementation tenants. To generate time anomalies in a nonproduction tenant, you need to submit and either approve or deny time blocks in that tenant.

While there might be multiple factors that contribute to a time block being anomalous, Workday identifies only 1 top possible anomaly reason for each time block. You can view anomalies that a manager can take action on in the Review Time report. Possible anomaly reasons include:

Reason	Description	In Review Time
Calculated Quantity	The quantity of time that the worker entered is greater than or less than their scheduled quantity of time.	Yes
Check-in	The check-in occurs outside of the worker's usual check-in time (based on their work schedule calendar).	Yes
Check-out	The check-out occurs outside of the worker's usual check-out time (based on their work schedule calendar).	Yes
Day of Month	The worker entered time for a day of the month when they typically don't work.
Day of Week	The worker entered time for a day of the week when they typically don't work.	Yes
Employee Type Location Location Type Job Profile Supervisory Organization Worker Type	The worker belongs to an employee type, worker type, job profile, or supervisory organization that frequently has anomalous time entries. Example: Members of the Sales organization have higher levels of time anomalies because they enter time on mobile devices while traveling. Workday takes this factor into account when determining the top possible anomaly reason.
Modified by Others	Someone other than the worker modified the time block.
Position Worker Time Block	A multiposition worker entered 1 or more positions that are different from the positions they typically work.
Project Worker Time Block	A Projects worker entered 1 or more projects that are different from the projects they usually work on.
Reported Quantity	The quantity of time that the worker entered is greater than or less than the worker's typical time entry pattern.	Yes
Reported & Scheduled Quantity Difference	The quantity of time that the worker entered is greater than or less than their scheduled quantity of time.	Yes
Shift Date	The shift date on the time block might be unusual.
Time Anomaly Count	The worker has a history of multiple anomalous time entries.
Time Calculation Tags	The worker entered time for 1 or more time calculation tags that they typically don't use.
Time Clock Event History	The time clock event history for the time block might include a missed check-in, check-out, or both.
Time in Position	Workers who have been in a position for a certain amount of time can enter more anomalies than others. Example: A worker just started in a new position and entered an anomalous time block. Their top possible anomaly reason could be Time in Position.
Time Type	The time type entered is different from the worker's usual time type.	Yes
Worktags	The worktags the worker entered are different from the worktags they typically enter.	Yes

Workday uses an anomaly confidence score to quantify the likelihood that a time entry is anomalous. The anomaly confidence level is:

Low when it corresponds to 50 to 80 percent.
Medium when it corresponds to 80 to 90 percent.
High when it corresponds to 90 to 100 percent.

You can include the anomaly score in custom reports using the

Time Anomaly Score

report field on the Time Block business object.

You can enable time anomalies to display in the

Review Time

report. When you display anomalies in Review Time, managers and timekeepers can quickly identify time entry issues that need their attention. Workday displays a

Workers with Possible Time Anomalies

filter on the summary page of the

Review Time

report. Workday also displays these columns:

Possible Time Anomaly Count
on the summary page.
Possible Time Anomaly Reason
in the worker details view.
Time Anomaly Confidence Level
in the worker details view.

The

Review Time

report only displays time anomalies that a manager can take action on. Examples:

Managers can only approve submitted time. Workday won’t display time anomalies related to approved or unsubmitted time.
Time needs to be ready to approve without any errors. Managers can’t approve time with errors.
Workday won't display time anomalies related to job profiles because managers can't modify job profiles from the Review Time report.