Analytics

Filtering runs

  1. In a runnable’s run list page, users can filter runs using date range and values in input and output. Click “List Runs” in a runnable’s detail page
_images/filter_runs_1.png
  1. For nested data, use double underscores (__) to represent an index or key. For example, if a typical data looks like {“key1”: [{“key2”: 3}]} and you want to search where the value for key2 is 3, you would enter “key1__0__key2” for path value and 3 for value
  2. Note that a data range cannot contain more than 30 days and if it contains more than 30 days, it will be shortened to the last 30 days before the end date.

Setting Up Alarms

  1. You can configure to receive alarms when variable distributions change. The platform offers two types of alarms:

    1. Numeric Simple Alarm: if a percentile aggregate of a variable’s values is too high or low compared to a threshold, send an alarm
    2. Categorical Simple Alarm: if a relative frequency of a variable’s value is higher or lower than a threshold, send an alarm.
  2. To create a numeric simple alarm, click “Create Numeric Simple Alarm” button in a runnable’s detail page

  3. Provide appropriate values to the form.

    1. Json field: whether the variable exists in input or output
    2. Path: path of the variable
    3. Percentile: a value between 0 and 100 indicating which aggregate to look at. For example, 0 is min, 50 is median and 100 is max
    4. Number of periods and period unit: How long to look back in calculating an aggregate
    5. Threshold direction: the condition to trigger this alarm (lower or higher?)
    6. Threshold: the threshold value
    7. Evaluate frequency and unit: how often to evaluate this alarm
    8. Minimum sample count: if the number of runs in the period is less than this value, an alarm won’t get triggered
    9. Recipients: a comma-separated list of emails to receive alarms when triggered
    10. For example, if you want to receive an alarm every hour when the 95th percentile of a “score” variable in output in the last hour exceeds 90, then you will see the above parameters as follows: Json field – output, Path – score, Percentile – 95, Number of periods – 1, Period unit – hour, Threshold direction – Target is greater than threshold, Threshold – 90, Evaluate frequency – 1, Evaluate frequency unit – hour, Minimum sample count – 30, Recipients – your email.
  4. Creating a categorical simple alarm is similar. In the form specify the value of a variable that you are interested in checking a relative frequency and when you want to get an alarm.