Data cube measures are numeric data values derived from your original table. For example, a measure can be an aggregation or a function output.
A measure that represents the event count is often created by default, depending on the data. For example, the Number of Events measure:
You can edit measures or add new ones in the Measures tab of the data cube edit view.
You can configure the measure format in the Format tab.
Here you can select from a list of abbreviation sets that can be applied to format the measure, and adjust the decimal precision that will be displayed.
Additionally, when applying time comparisons to measure values, you can configure the display coloring for increased and decreased values.
Measure fill options
You can specify how to fill a measure on continuous visualizations, such as line charts.
A data set can have gaps in data, for example, due to down time for the systems that generate the data. There are a few ways to handle the missing data in charts, as determined by fill options. The example shows four measures that are differentiated by differing fill options.
The possible options are:
- Zero filled (default) – fill missing data with zeros. This is most suitable if the measure represents an additive quantity.
- No fill – leave missing data empty. This is suitable when missing values indicate that the data is not collected.
- Previous value – fill missing data with last value seen. This is suitable for sensor type data.
- Interpolate values – interpolate the missing data between the seen value. This is suitable for sensor type data.
In the Advanced tab, you can transform a measure to display as
Percent of parent segment or as
Percent of total instead of the default measure display.
Create a measure
To create a measure, click the New measure button in the Measures tab.
Use the Basic tab to configure simple measures consisting of a single aggregate function over a single column (with an optional filter).
A measure can also represent some post aggregation or computation (see the specific measure types section below). In that case you would use the Custom measure tab, where you can enter any supported Pivot SQL expression as the measure's formula.
Imply can also provide suggestions about simple measures. Polaris scans the schema of the underlying table and automatically suggests a measure with the appropriate aggregate for any column that is not already represented by an existing measure.
This is particularly useful if you have added a new column to your table after creating the data cube and now want to represent it in your views.
Custom measure examples
This section describes some of the specific measure types.
The sample expressions are in Pivot SQL format.
Filtered aggregations are very powerful. For example, if your revenue in the US is a very important measure, you can express it as:
SUM(t.revenue) FILTER (WHERE t.country = 'United States')
It is also common to express a ratio of something filtered vs. unfiltered.
SUM(t.requests) FILTER (WHERE t.statusCode = 500) * 1.0 / SUM(t.requests)
Ratios surface relationships in your data.
Here's one that expresses the computation of CPM:
SUM(t.revenue) * 1.0 / SUM(t.impressions)
A quantile can be a very useful measure of your data. For large data sets it is often more informative to look at the 98th or 99th quantile of the data rather than the max, as it filters out corrupted values. Similarly, a median (or 50% quantile) can present different information than an average measure.
To add a quantile measure to your data, define a formula like so:
It is possible to fine-tune the accuracy of the approximate quantiles as well as pick a different algorithm to use.
To learn more, see the Druid documentation.
In some cases, you may need to determine the rate of change (over time) of a measure.
There is a special function provided by Pivot
PIVOT_TIME_IN_INTERVAL('SECOND') that gives the corresponding number of time units (first parameter) in the time interval over which the aggregation is running.
Therefore, it is possible to define a measure such as this one:
SUM(t.bytes) / PIVOT_TIME_IN_INTERVAL('SECOND')
This gives you the accurate rate of bytes per second regardless of your filter window or selected split (hour, minute, an so on).
Nested aggregation measures
It is possible to define measures that perform a sub-split and a sub aggregation as part of their overall aggregation. This is needed to express certain calculations which otherwise would not be possible.
The general form of a double aggregated measure's formula is:
PIVOT_NESTED_AGG(SUB_SPLIT_EXPRESSION, INNER_AGGREGATE(...) AS "V", OUTER_AGGREGATE("V"))
SUB_SPLIT_EXPRESSIONis the expression on which the data for this measure is first split.
INNER_AGGREGATEis the aggregate that is calculated for each bucket from the sub-split and is assigned the name
V(which is arbitrary).
OUTER_AGGREGATEis the aggregate that aggregates over the results of the inner aggregate, it uses the variable name declared above.
Examples of double aggregated measures:
Netflow 95th percentile billing
When dealing with netflow data at an ISP level one metric of interest is 5 minutely 95th percentile, which is used for burstable billing.
To add that as a measure (assuming there is a column called
bytes which represents the bytes transferred during the interval), set the formula as follows:
PIVOT_NESTED_AGG(TIME_FLOOR(t.__time, 'PT5M'), SUM(t.bytes) * 8.0 / 300 AS "B", APPROX_QUANTILE_DS("B", 0.95))
Here the data is sub-split on 5 minute buckets (PT5M), then aggregated to calculate the bitrate (8 is the number of bits in a byte and 300 is the number of seconds in 5 minutes). Those inner 5 minute bitrates are then aggregated with the 95th percentile.
Average daily active users
When looking at users engaging with a website, service, or app, we often need to know the number of daily active users.
You can calculate it as follows:
PIVOT_NESTED_AGG(TIME_FLOOR(t.__time, 'P1D'), COUNT(DISTINCT t."user") AS "U", AVG("U"))
Here the data is sub-split by day (P1D), and then an average is computed on top of that.
Switching metric columns
If you switch how you ingest your underlying metric and can't (or don't want to) recalculate all of the previous data, you can use a derived measure to seamlessly merge these two metrics in the UI.
Let's say you had a column called
revenue_in_dollars, and for some reason you will now be ingesting it as
Furthermore, right now your users are using the following measure:
If your data had a "clean break" where all events have either
revenue_in_cents with no overlap, you could use:
SUM(t.revenue_in_dollars) + SUM(t.revenue_in_cents) / 100
If instead there was a period where you were ingesting both metrics, then the above solution would double count that interval. You can splice these two metrics together at a specific time point.
Logically, you should be able to leverage Filtered aggregations to do:
SUM(t.revenue_in_dollars) FILTER (WHERE t.__time < TIMESTAMP '2016-04-04 00:00:00') + SUM(t.revenue_in_cents) FILTER (WHERE TIMESTAMP '2016-04-04 00:00:00' <= t.__time) / 100
Other custom measure examples
Here are more example expressions that you can use to build custom measures:
Counts all the rows in the data cube.
Counts the distinct values of the dimension.
Counts how many rows match a condition. This example counts how many rows in the dimension are true.
SUM(CASE WHEN (t."dimension"='true') THEN 1 ELSE 0 END)
Returns the sum of all values for the dimension.
Returns the smallest value of the dimension.
Returns the largest value of the dimension.
Returns the average value of the dimension.
Returns the x quantile of the dimension. You cannot use a compound expression in this function.
Limits the number of records to apply a given aggregate. This example returns x or fewer results.