Set Aggregate Thresholds

Hide aggregate values for small populations from users.

Note: Setting aggregate thresholds may result in a large number of metric values to be hidden. Before you enable this feature, contact Visier Technical Support for more information about user impact. There may be alternative ways you can hide certain details from users through data security.

An aggregate is a whole value made up of several units. For example, the aggregate compensation data for women in the New York office is made up of each individual woman and her base pay. Sometimes, there are very few members in a population, such as women in the New York office, which makes it possible for users to discover sensitive information from the data.

You can set an aggregate threshold to require a minimum number of members that must be in a population for aggregate values to be displayed. Aggregate thresholds make it harder for users to infer specific data values for members of small populations from aggregate values.

Example: Using aggregate values to infer individual data values

Rachel Chu is the only woman in the New York office. Users with this knowledge can apply filters to the data to uncover sensitive information about her. If users have aggregate access to compensation data, they are restricted from seeing the Base Pay value for specific employees. However, users can easily filter their analysis population to women in New York to see an aggregate value for Base Pay. From there they can conclude that the aggregate value displayed for the Base Pay metric is actually Rachel's Base Pay value.

Note: Even if aggregate thresholds are applied, users can still infer specific data values from aggregate values using math and filters. To ensure that users cannot access sensitive data, such as compensation, we recommend that you restrict their data access. You can do this by changing the access level for attributes from Aggregate to None. For more information, see Data Access Sets.

Aggregate threshold options

The following options are available when configuring the aggregate threshold:

No threshold: Aggregate values are displayed even if the population consists of only one member. No threshold is applied by default.
Set threshold for subjects: Aggregate values are hidden if the population falls below the aggregate threshold. You can set thresholds for individual or all subjects.
Set threshold for a subject property: Aggregate values are hidden if the population falls below the property's aggregate threshold.

Note: You can also set thresholds for event properties and multi-value properties that are associated with a subject.

Using the Employee subject as an example, you can set a threshold for:

Relocation Cost, an Employment Start event property.
Compensation Amount, Compensation Item, Converted Amount, Currency Code, and FTE Weighted, the simple properties that make up the Employee Compensation Items multi-value property.

What happens when thresholds are not met

Users should expect the following chart behavior when their analysis population does not meet the aggregate threshold:

Aggregate values are hidden and N/A values are displayed.
In a Trend visual, data points are removed which may result in gaps in the trend line.

How thresholds are applied

For metrics that are calculated using average headcount, the threshold is also applied to the population that contributed to the calculation.

Example:

Turnover Rate is calculated as the Turnover Count divided by Average Headcount. The aggregate threshold for Employee has been set to 3. The Turnover Rate will be hidden if one of the following criteria is met:

The Turnover Count is less than 3.
Over the selected time period, the total number of unique employees is less than 3.

Note: The threshold is not applied if a user has detailed access to the fields within the aggregated records.

Set aggregate thresholds

In a project, on the navigation bar, click Model > Settings > Aggregate Thresholds.
Click Enable aggregate thresholds.
Optional: Set the default aggregate threshold value which will automatically be applied to all and any new subjects.
To set the aggregate threshold value for individual subjects, click Add.
Select the subjects that require individual aggregate thresholds.
For each subject, set a threshold.
Optional: To override the subject's threshold at the property level, click Add Property and define a new threshold for the property.
1. Define property overrides for related events in the Event Properties tab. Using the Employee subject as an example, you can define a threshold for Relocation Cost, an Employment Start event property.
2. Define property overrides for related multi-value properties in the Related Objects tab. Using the Employee subject as an example, you can define thresholds for Compensation Amount, Compensation Item, Converted Amount, Currency Code, and FTE Weighted, the simple properties that make up the Employee Compensation Items multi-value property.
When finished, publish your project to production. For instructions, see Publish Project Changes.

Example:

Let's say you've set a default threshold of 5 for all subjects. This means that any population must have a minimum of 5 members to display values to users with aggregate data access. However, you want an increased threshold for your employee population, specifically at the gender and ethnicity level.

You can add two property overrides and set their thresholds individually. For example, a threshold of 30 for gender and 50 for ethnicity. This means that users with aggregate data access will not see values for populations that have fewer than 30 members for gender-based metrics or filters, nor will they see populations below 50 if they are looking at ethnicity-based metrics or filters.

Note:

If a user has mixed access to a population that provides them with both detailed access to some individuals and aggregate access to others in that population, they cannot view groups that fall below the defined aggregate threshold.
It is important to understand that the threshold is based on the number of unique keys found in the processed records, not by the value of the results.

Examples

Mixed access

To explain how aggregate thresholds are applied when users have aggregate access to all properties and detailed access to certain subsets, we are using the following permissions, users, and employee data in the examples below.

The data access granted by each permission:

Permission	Subject	Population access	Data access
Permission 1	Employee	All	Aggregate access to all properties
Permission 2	Employee	Organization: IT	Detailed access to Organization, aggregate access to the rest of the properties
Permission 3	Employee	Gender: Female	Detailed access to all properties

The permission assignments for each user:

User	Permission
User 1	Permission 1, Permission 2, Permission 3
User 2	Permission 1, Permission 2
Super User	Detailed access to all properties

The employee data loaded in Visier:

Employee ID	Organization	Gender	Location	Hourly wage
E1	IT	Female	United Kingdom	21
E2	IT	Female	United States	22
E3	HR	Male	United Kingdom	23
E4	IT	Female	United Kingdom	24
E5	HR	Male	United States	25
E6	IT	Male	United Kingdom	26
E7	HR	Female	United Kingdom	27
E8	IT	Male	United Kingdom	28
E9	IT	Male	United Kingdom	29
E10	IT	Female	United States	30
E11	HR	Male	United Kingdom	40
E12	IT	Male	United States	34

Headcount grouped by Gender, employee aggregate threshold 6

Gender	Super User	User 1	User 2
Female	5	5	N/A
Male	7	7	7

In this example we are looking at how headcount, grouped by gender, is displayed based on different access levels when the aggregate threshold is 6.

User 1: The female headcount shows as 5. This is because they have detailed access to the female gender field from Permission 3, overriding the aggregate threshold.
User 2: The female headcount shows as N/A. This is because they have aggregate access to the female gender field, and since the number of unique keys (E1, E2, E4, E7, and E10) is below the aggregate threshold of 6, N/A is displayed.

User 1 and 2: The male headcount shows as 7. Both users have aggregate access to the male gender field, and because the number of unique keys (E3, E5, E6, E8, E9, E11, and E12) meets the aggregate threshold of 6, 7 is displayed.
Super User: Both values are shown as they have detailed access to all properties.

Headcount grouped by Gender, employee aggregate threshold 6, gender aggregate threshold 8

Gender	Super User	User 1	User 2
Female	5	5	N/A
Male	7	N/A	N/A

In this example we are looking at how headcount, grouped by gender, is displayed based on different access levels when the employee aggregate threshold is 6, and the gender aggregate threshold is 8.

User 1: The female headcount shows as 5. This is because they have detailed access to the female gender field from Permission 3, overriding the aggregate threshold. The male headcount shows as N/A because User 1 has aggregate access to the male gender field, and the number of unique keys (E3, E5, E6, E8, E9, E11, and E12) is below the gender aggregate threshold of 8.
User 2: Both the female and male headcounts show as N/A. This is because they have aggregate access to the gender field, and both values are below the aggregate threshold for gender, so N/A is displayed.
Super User: Both values are shown as they have detailed access to all properties.

Headcount grouped by Organization filtered by Location United Kingdom, employee aggregate threshold 4

Organization filtered by Location (UK)	Super User	User 1	User 2
IT	5	5	5
HR	3	N/A	N/A

In this example we are looking at how headcount, grouped by organization, filtered to the United Kingdom, is displayed based on different access levels when the employee aggregate threshold is 4.

User 1: The HR value shows as N/A. This is because they have detailed access to one HR employee (E7) from Permission 3, but they do not have detailed access to the remaining HR employees (E3, E11). The number of unique keys (E7) is below the aggregate threshold, so N/A is displayed.
User 2: The HR value shows as N/A. This is because they have aggregate access to all of the filtered records, and the HR value is below the aggregate threshold, so N/A is displayed.

User 1 and 2: The IT value shows as 5. Both users have detailed access to the IT Organization field, and aggregate access to the location field. The value is displayed because the number of unique keys (E1, E4, E6, E8, and E9) is above the aggregate threshold, however if the aggregate threshold was increased to 6, N/A would be displayed.
Super User: Both values are shown as they have detailed access to all properties.

Average hourly wage grouped by Organization filtered by Location United Kingdom, employee aggregate threshold 4

Organization filtered by Location (UK)	Super User	User 1	User 2
IT	25.6	25.6	25.6
HR	30	N/A	N/A

In this example we are looking at how the average hourly wage, grouped by organization, filtered to the United Kingdom, is displayed based on different access levels when the employee aggregate threshold is 4.

IT: The average hourly wage is showing 25.6 for IT. There were 5 unique keys used to process the value, which is above the aggregate threshold of 4. Both User 1 and User 2 have detailed access to the IT Organization field, and aggregate access to the location field. If the aggregate threshold was increased to 6, N/A would be displayed.
HR: The HR value shows as N/A. There were 3 unique keys use to process the value, which is below the aggregate threshold. Both User 1 and User 2 have aggregate access to the fields used in the calculation, so N/A is displayed.

Events

In this example, we are looking at a basic count of event survey data. Some surveys within the dataset share the same event ID, and the user consuming the data has aggregate access to all records used in the calculation.

Event survey data sorted by ID: E1, E1, E1, E1, E2, E3, E4, E5.

Threshold	Count	What value is shown
5	8	8
6	8	N/A

If the aggregate threshold is 5, the value shows as 8. This is because the number of unique keys (E1, E2, E3, E4, and E5) meets the aggregate threshold of 5. However, when the aggregate threshold is increased to 6, the value shows as N/A, since the number of unique keys is below the aggregate threshold.