Applying time series aggregations

  • Release version: Washingtondc
  • Updated February 1, 2024
  • 7 minutes to read

    • You can aggregate changes in indicators into discrete time intervals. These aggregations can make trends more easily visible, or help track progress against a target.

    A time series aggregation consists of a math function, such as AVG or SUM, combined with a time formula, such as By quarter. In Platform Analytics, you can select an aggregation when exploring an indicator in KPI Details or when creating a data visualization for an indicator data source, such as a time series visualization. In the Core UI, you can select an aggregation either in the Analytics Hub or when creating a Performance Analytics widget such as a score widget.

    To understand time series aggregations, first consider a watch, phone, or fitness tracker that keeps tabs on how many steps they walked each day. These devices all collect and save your measurements on a minute by minute basis. However, the apps and reports are often designed to show how your activity is progressing compared to last week, last month, or last year. It’s one thing to see if you were more active at noon versus 10 am. What you really want to know, however, is whether you’re continuing to improve over time. Aggregating your steps into weekly, monthly, and yearly scores gives you an easier way to identify trends in your activity.

    While daily indicator scores are foundational and almost always desired, sometimes you also want to have the scores available by week, month, or quarter. In other cases, you want to have a month/quarter/year-to-date number that shows cumulative progress up until the current point. Instead of defining multiple indicator sources and indicators to track each interval, Performance Analytics natively allows you to capture the data once and then adjust the view. Use a time series aggregation for any of these scenarios:

    • Aggregate the data to a less frequent period with a Period time series.
    • Smooth out spikes in the data to reveal trends with a Running time series.
    • Determine a period-to-date cumulative score with a To date time series.
    Note:
    Business and fiscal calendars do not support time series aggregations. You cannot set a time series aggregation on a widget or Analytics Hub entry for an indicator that uses such a calendar.

    For examples and a deeper exploration of using time series aggregations with indicators, see this Performance Analytics Academy video, Leveraging Time Series with Performance Analytics.

    Partial periods

    Some time series include indicator scores from incomplete collection periods. These periods can include the current period and the period from the beginning of data collection. A plus sign in the name, +, identifies these time series.

    A time series that does not include data from partial periods must have data from the beginning and the end of the period. For example, a By month SUM time series aggregation requires scores from the start and the end of the month to be present. Otherwise that month is not included. A time series that includes data from partial periods, such as By month SUM +, needs only data from one day in the period.

    Warning:
    Partial periods can skew the results of certain aggregations, such as averages.

    Aggregation definitions

    Performance Analytics comes with default SUM, AVG, and other time series aggregation definitions.

    Danger:
    Do not alter these definitions. Do not add new definitions. Any changes to aggregation definitions can have unexpected results. Creating a new definition can prevent the data collector jobs from running.

    Disallowed time series aggregations

    You are not allowed to put a SUM aggregation on an indicator with percentage % as the unit. It is not meaningful to sum percentages.

    Validation of time series aggregations

    Time series aggregations are validated on the Analytics Hub and in KPI Details but not in widgets. Therefore, you might navigate from a working widget to the Analytics Hub and get a warning that the indicator has an invalid aggregation.

    Excluding time series aggregations for an indicator

    Some time series aggregations, while technically allowed, are not helpful to apply to an indicator. For example, a SUM of percentage values is unlikely to provide useful insight. You can exclude these time series aggregations from the indicator. For more information, see Exclude time series from an indicator.

    Displaying time series aggregations with a date range selector

    If on a widget you combine a time series aggregation such as By month SUM with a date range selector, you may show one more time period than you expect. The reason is that when a period such as 3m is selected in the date range selector, the instance interprets that as 0-3 months and shows four months of scores.

    Use cases for time series aggregations

    Performance Analytics offers four different types of time series. Understand their use cases to know which type to use.

    Rounded calculations based on averages may be off by a small amount due to rounding errors.

    Table 1. Use cases and examples for each type of time series
    Type Examples Use Case
    Running

    28d running AVG/SUM

    30d running AVG/SUM

    7d running AVG/SUM

    12m running AVG/SUM

    3m running AVG/SUM

    6m running AVG/SUM

    13w running AVG/SUM

    4w running AVG/SUM

    4q running AVG/SUM

    		 Smooths out spikes in the data to make trends easier to spot. For example, looking at daily incident counts may show a drop every weekend, but a 7-day running average smooths out those drops.
    To Date

    Fiscal quarter-to-date AVG/SUM

    Fiscal year-to-date AVG/SUM

    Month-to-date AVG/SUM

    Quarter-to-date AVG/SUM

    Week-to-date AVG/SUM

    Year-to-date AVG/SUM

    		 Shows cumulative scores for the period. These time series aggregations are very useful if you have a monthly target to hit, but you need to also see the velocity throughout the month.
    By Period

    By week AVG/SUM

    By month AVG/SUM

    By fiscal quarter AVG/SUM

    By quarter AVG/SUM

    By fiscal year AVG/SUM

    By year AVG/SUM

    		 Shows the cumulative scores for entire periods. While you may want to track the number of P1 incidents daily, the frequency is too high when scores are considered daily. Instead, you can set a target at the monthly level with a "By Month” time series. The current period will never appear in the results because it is incomplete.
    By Period (Including Partial)

    By week + AVG/SUM

    By month + AVG/SUM

    By fiscal quarter + AVG/SUM

    By quarter + AVG/SUM

    By fiscal year + AVG/SUM

    By year + AVG/SUM

    		 The “+” version of the “By” Time Series includes partial periods, so a score is always provided for the current period.

    Indicator frequency limitations on time series aggregations

    The frequency with which scores are collected for the indicator determines which time series are applicable. Some time series include data from partial collection periods.

    When you select a time series aggregation, the frequency with which indicator scores are collected limits which time series you can choose. You cannot select a time series aggregation that is applied to scores more frequently than those scores are collected. For example, the By week SUM time series aggregation can apply to an indicator with a daily frequency. However, By week SUM cannot apply to an indicator with a weekly, monthly, quarterly, or yearly frequency.

    Note:
    • Only weekly indicators support the 4w running and 13w running time series. Weekly indicators support only weekly and yearly time series.
    • Bi-monthly and yearly indicators do not support any time series aggregations.

    The following table shows which time series are supported for which indicator frequencies. These relationships are independent of which aggregation (AVG, SUM, or custom) is combined with a time series, and therefore only the time series are shown.

    Table 2. Time series and associated indicator frequencies
    Time series Indicator frequencies
    Daily Weekly Monthly Quarterly (Fiscal Q, 4-weekly, Bi-weekly)
    Daily Yes No No No
    7d running Yes No No No
    28d running Yes No No No
    30d running Yes No No No
    4w running No Yes No No
    13w running No Yes No No
    3m running No No Yes No
    6m running No No Yes No
    12m running No No Yes No
    4q running No No No Yes
    By week, by week + Yes No No No
    By month, by month + Yes No No No
    By quarter, by fiscal quarter, by quarter +, by fiscal quarter + Yes No Yes No
    By year, by fiscal year, by year +, by fiscal year + Yes Yes Yes Yes
    Week to date Yes No No No
    Month to date Yes No No No
    Quarter to date, fiscal quarter to date Yes No Yes No
    Year to date, fiscal year to date Yes Yes Yes Yes