Our Automations feature lets you set up notifications to alert your about changes in the data set that meet predetermined conditions. Automations allow you to specify exactly what changes you’re interested in tracking, the cadence of updates, and who to notify about these events.
Best Practices for using Automations
The ability to set up notifications is based on predetermined conditions or changes in a data set. There are several best practices that we suggest following, listed below:
1. Be specific in the alerts you set up to maximize their impact (e.g. delta’s, increases, decreases, etc.).
Logic should account for a targeted insight - alerts should be specific (e.g. isolate specific taxonomy labels, bottom 5% sentiment scores, etc).
Logic should be isolated to one model and structured data at most - to avoid model processing limitations, we suggest users select logic based on only one model and structured data. With multi-model filtering, errors can occur based off models finishing at different times.
2. Consider the underlying data stream when setting up alerts.
In order to get the maximum value to be sure to:
Base alerts on when actions take place (e.g. alerts for compliance breach, alerts when sentiment score changes, etc.).
Base alerts on how data is updated (e.g. if the data stream is updated weekly, set up weekly alerts, monthly alerts, or for when data changes, NOT daily).
Base alerts on your analytical goals (e.g. in time for a monthly leadership report (monthly cadence), or in time for a daily report to management (daily cadence)).
3. Set appropriate alert responses based on the conditions prompting the alert.
Let’s imagine I’ve set up alerts to notify me any time there’s a breach of compliance based on an increase in the “number of records” for a given taxonomy label. To get the maximum value out of the Automations feature, the alert should:
Be sent to the appropriate email accounts
Provide a description of what the label is and what it represents
Provide a description of who has broken compliance.
NOTE: This may require further investigation within the workspace.
4. Setup a single dashboard to isolate a specific alert use case:
Alerts work best when you have one dashboard using one model and a specific use case. This way the alerts trigger quickly (when the model completes) and the logic is straightforward.
Using this approach is the best way to avoid processing errors.
How to Configure Automations
1. Start by clicking the bell button and plus button to create the Automation.
2. Select a numeric dimension to track (for this example score increase)
3. Select an aggregation method and logic for your alert (for this example average score that increases by more than 1. You can also select % vs set numerical value)
4. Select how you want to receive notifications.
5. Choice your alert frequency and continue.
6. Input the recipients, the message, and selected fields for the alert.
6. Select fields for sample data that will be included in the alert to help determine if you want to go to the workspace to review full insight.
6. Enter a topic or subject name and short description for the alert. Click submit to active your alert.
Alerts will not send until all models are finished processing in your dashboard
We have set up an alert when a new record arrives for a taxonomy label. In the workspace, a Theme Analysis and a Taxonomy Analysis are deployed. When new data comes in, the taxonomy model finishes first and then the theme/topic analysis finishes 30 minutes later. The alert will be triggered 30-40 minutes later.
Alerts should only be set up using one model and/or structured data
Processing errors can occur for alerts using a filter on a taxonomy label AND a sentiment score > 0.
Emails that do not yet exist in the Stratifyd system must first be added. If you need assistance, contact your Stratifyd account manager or submit a ticket to get a new team member's email added
If using field selection, the first two fields MUST be a small character field or number. Both of these fields will be formatted in the top left and right of the alert template (see below for example)
In the future we plan on addressing each of the limitations above. As we make progress towards fixes for these issues, we will update the content of this page. For now using the following best practices is the best way to ensure success.
We're here to help! Don't hesitate to contact us for further assistance via chat or submit a ticket!
Please sign in to leave a comment.