Data Monitor and Alert 🔔
Monitor metrics or react to data events and alert when needed. Ideal for creating actionable alerts in response to business risks or opportunities and for monitoring data validity.
⚙️ Technical Examples
Monitor Data Assets: Alert when new data is available.
Specific Data Events: Alert on specific data events within a monitored asset.
Metric Monitoring: Define trigger conditions using anomaly detection or rules like moving average and period-over-period comparisons.
📈 Business Use Cases
Marketing: Monitor new leads table and alert on new high-priority leads.
Fin-Ops / Data Quality: Monitor daily total transactions per customer and alert when anomalies are detected.
Customer Success/Ops: Monitor customer weekly app token consumption and alert on a declining trend of over 15% compared to the previous week.
Simply set your monitoring conditions and alert criteria to stay informed and take timely action based on your data insights.
In this article we are going to cover the following topics:
In this article we are going to cover the following topics:
There're 2 types of data triggered alerts:
1️⃣ Data events
Use data events to:
Monitor a truncating table and alert when it contains any values
Monitor new rows added to a table that follow some specified conditions defined on the columns
Monitor new records added per business stakeholder and personalize their alert trigger and content accordingly
Any data is available
This option means that the alert will be triggered when ANY results are returned, according to the asset and applied filters. (For example, when a specific table is not empty)
New data, since last sample, is available
This option means that the alert will be triggered when there are NEW results, according to the asset and applied filters. (For example, when new purchases are added and these are attributed to the TikTok campaign)
Primary key(s)
Define a field, or more, to compare vs. last sample. This is how Rupert will know when new data is added and the alert should be triggered.
For example, you can use the id field of your table as the primary key for comparison. This means that each sample containing new ids will trigger the alert. Primary key can also be a timestamp column or a combination of a few columns.
Filters:
After selecting / creating the asset to build the trigger on, you can add filters (optional). For example: if you want the alert to trigger when new purchases are added but only if are attributed to the TikTok campaign
Dynamic value comparison filter
With this option, you can set filters to compare two dynamic values within your data. This powerful enhancement allows for more nuanced and accurate data filtering.
Example Use Case: By comparing the fields number_of_credits_used and number_of_credits_allowed, you can identify accounts that have exceeded their credit limits and take appropriate action.
Dynamic date filters
Dynamic dates - common use cases
1. Time Interval Definition:
Purpose: Monitor activity and ensure timely follow-ups
Use Case: Check if more than a specified number of days, months, or years have passed since an event occurred.
Examples:
Account Activity Monitoring: Identify accounts that have not logged in for over 30 days to send re-engagement emails.
Payment Compliance: Detect users who have not made a payment in the last month to prompt them for renewal.
2. Dynamic Date Fields Comparison:
Purpose: Enable nuanced and accurate data filtering by comparing two dynamic date fields.
Use Case: Add a time interval between two dynamic values to refine data analysis.
Examples:
Trial to Paid Conversion: Filter results to show accounts that moved to a paid version less than a week after starting a trial.
Service Upgrade Timing: Identify users who upgraded their service within a certain period after account creation to analyze upgrade patterns.
3. Historical Data Evaluation:
Purpose: Detect significant changes or trends by comparing historical and current data.
Use Case: Evaluate previous alert or sample values compared to current ones.
Examples:
Credit Usage Trends: Assess if there has been an increase in credit usage over time and investigate potential causes.
Performance Tracking: Compare past and present performance metrics to identify improvement areas or emerging issues.
Is null filter
The "Static Values" option enables you to filter trigger results using the "is" or "is not" operators. For filtering based on null values, click on ⚡Dynamic Values to select the null option (see image below).
Note: This option is only available when selecting a text field.
Custom expression - advanced filtering
As a filter
An a Aggregated Filter After Group By
ℹ️ Learn more about how to work with expressions and see common use cases and examples in this article
Group by
Along with aggregated conditions, now available for SQL sources using data event-based triggers. This powerful addition allows you to group your data based on specific fields and apply aggregate functions to these groups. Summarize and analyze large datasets by grouping data and applying aggregate functions like COUNT, SUM, AVG, MIN, and MAX.
Example Use Case: A classic application of this feature is to alert when a certain aggregated measure exceeds a defined threshold for each user or customer. For instance, you can:
Group By User or Customer: Organize data by user or customer ID.
Apply Aggregate Conditions: Use aggregated functions to calculate the total / average amount spent by each customer or the time elapsed since the last event for each user or customer.
Set Up Alerts: Trigger alerts when the total amount spent by a user exceeds a specified threshold, enabling targeted follow-ups and personalized engagement strategies.
2️⃣ Metric Over time
Use metric over time to:
Monitor hourly metric and alert when it drops below some KPI by 10%
Monitor metric with some filters and breakdowns and compare each broken down metric to the same time last week and alert in case of a large drop
Monitor weekly top company metrics and trigger an alert when anomaly is detected in the metric value
Define metric
After selecting / creating the asset to build the trigger on, set the aggregated measure to monitor and the time period. The time period range from hour to month.
After setting the aggregation + time period, you can add a filter and/or a breakdown
Custom expressions
Define a metric using an advanced custom expression
Or add an advanced filter using a custom expression
ℹ️ Learn more about how to work with expressions and see common use cases and examples in this article
Breakdown metric
You can preview the different breakdowns in the metric preview chart. Use the Breakdowns dropdown to select one or more breakdowns to preview.
🟣🟣🟣
Trigger conditions
Check if your metric meets a specific condition (or more). In this section you can define rules and thresholds or choose the anomaly option to help you detect, monitor and alert on any anomaly in the data.
Rules
Rules can contain a static threshold value to check or a dynamic one:
Static value
Use one of the comparison operators to check if the metric meets a specific threshold value.
For example, trigger when the sum of revenue is lower than 50,000
Dynamic value
Use one of the comparison operators to check if the metric meets a dynamic threshold.
For example, trigger when the sum of revenue is grater that the moving average in the last 30 days.
There are a few types of dynamic values you can use to compare your metric with
Trigger settings
There're 2 settings you can use to help configure the trigger:
Delay - alert only after N triggering samples
Snooze breakdown repeated values
To configure the settings, after defining your metric, click the "Trigger settings" button:
1. Snooze breakdown repeated values
Enable snooze, in the snooze settings, to temporarily pause alerts from triggering repeatedly for the same breakdown value.
Use-case example
Say you are tracking the spend by different account. If the same account exceeds spending limits twice consecutively, it will be snoozed on the second occurrence, preventing multiple alerts for the same issue.
If the issue is resolved and the account stops triggering, it will trigger alerts again in the future if issues recur.
2. Delay alerting: alert only after N triggering samples
When dealing with noisy data or situations where occasional random deviations occur, you may want to configure alert delay to prevent unnecessary alerts.
To set up alert delay, specify the number of consecutive triggering samples required before an alert is sent.
Use-case example
If you're monitoring the number of user clicks per customer account using anomaly detection, and you set the consecutive trigger count to 3, here’s what happens:
The first time an anomaly is detected in the click count, it will be recorded as the first triggering event, but no alert will be sent.
If the anomaly persists in the next sample, it will be counted as the second triggering event, but still, no alert will be sent.
On the third consecutive sample, if the anomaly is still present, an alert will be sent, along with any other triggering events in that sample.
To configure this, expand the "Delay settings" option, check "Enable," and enter the number of consecutive triggering events required before an alert is sent.
Track snoozed and delayed data
You can track data sampling results by downloading the sample data CSV from the Alert Manager. This file is available next to each specific sample or delivery line. The CSV includes a status column that shows the status for each recipient and breakdown (if applicable).
The sample status column will display the following:
Triggered: The data met the conditions and triggered the alert.
Triggered X/Delay Number: The data triggered but is still within the delay period and has not yet been sent.
Snoozed: The breakdown is snoozed and will not send alerts during this period.
Not Triggered: The data did not meet the conditions to trigger an alert.
By reviewing this CSV, you can easily monitor the status of your alerts and understand whether they are triggered, delayed, snoozed, or not triggered.
Metric preview
This chart will display a preview of your trigger configuration which include the most recent data considering the filters, breakdowns and rules.
Here you can see the trigger points and when the conditions are met and get a noise level indication and prediction based on previous data
Control the plot preview period - deep dive into recent values or display historical view to learn about the metric's trend. You can modify the start date for the preview by clicking the edit icon
Anomaly detection
This option allows you to trigger an alert once an anomaly is detected.
The model in which Rupert uses to determine and predict data anomaly is based on past data from your source.
** If there's not enough data for the model to predict the anomaly, no results will be shown.
The advanced options allow you to adjust the prediction model and make it more / less sensitive to various variables.
The default settings will trigger for every deviation, above or below, the data range
Choose “Above” to only track cases where the metric value is more than the upper bound
Choose “Bellow” to track metric values under the lower bound. For example, if you're monitoring retention, you can set an alert to trigger only when retention drops more than usual
🟣🟣🟣
What’s next? Check out these articles to learn more