Proactive Insights is a set of features that uses machine learning and pre-defined conditions to help you understand and act on your data. All insights are automatically generated every day and when available you can see them in the Insights feed.
- Please note, Proactive feature is available only for Dealtale customers in the Pro and Enterprise Packages.
Anomaly Detection Insights act like a dedicated, 24/7 AI analyst. It tracks unusual changes or emerging trends in your data and notifies you automatically through the Insights feed. Moreover, for certain Insights, we can indicate their possible cause, so you can take rapid actions to fix it.
All active Anomaly Detection Insights can be identified with the monitor icon [icon] that will only be displayed when the insight is relevant for you to take action upon. You can review all past Anomaly Detection Insights in the Insights archive.
Getting started is easy, as long as you have created charts for metrics you are interested in and added these charts to a dashboard, we automatically start tracking your data’s behaviour for you. No additional actions are required.
Missing a chart for one of your important metrics? No problem, just create a chart for this metric, add it to one of your dashboards (or create a new relevant dashboard for it) and we will notify you of any anomalies from this point on.
*Please note, in order to determine if an anomaly has occurred, 60 days’ worth of historical data is needed for each metric. Please make sure all your data sources are connected to your DealTale account and working as expected in the data source tab.
The Insights feature provides continuous monitoring of business metrics to identify deviations from normal behaviour and generates focused explanations on affected segments to facilitate further analysis.
Based on state-of-the-art time series forecast algorithms and 60 days of historical data, a normal range of values for each metric is calculated. If the actual value of a metric falls outside of the predicted range, an anomaly alert is created. The statistical significance of obtained results is determined based on a historical lagged differences statistic.
When a global level anomaly is detected, an in-depth analysis is performed to identify affected segments. The affected metric is aggregated on multiple segmentation levels based on provided dimensions for a metric. Each segment is tested for anomalous behaviour based on the same approach that was used on a global level. Information about affected segments is aggregated and summarized to produce useful explanations about an observed anomaly.
The anomaly detection algorithm runs on your data once a day (after ensuring all data from the previous day has been synced from your data sources) and looks for anomalies that occurred the previous day.
Did you receive insights on metrics that aren’t relevant to you? Just turn off the insights for the chart which triggered the anomaly in the “learn more” screen, and you won’t be notified about anomalies in this chart’s metrics anymore. You can always revert back to receiving the anomalies in the settings screen.
Don’t want to receive anomalies for a certain period of time? Just head over to the insights settings screen and click on the “mute” button. This will stop notifications on anomaly detection for 7 days. You can “unmute” before the 7 days are over.
Don’t want to get any anomaly detection alerts? You can deactivate the feature with the settings toggle and reactivate it at any time.
We notify you of anomalies as soon as they are detected. We usually wait for the last sync of the previous day to ensure all data is available for review, to avoid false alerts.
All active anomalies can be found in the Insights feed, if they don’t appear anymore, it means we determined based on the data they are not active anymore (e.g. if the number of leads went from 10 leads per day to 100 leads per day, and it stays like this for a few days, the model we’ve built and trained will no longer consider it an anomaly, but the “new normal” and therefore it won’t be presented in the feed). However, you can access all anomalies that have ever been detected in the Insights archive.
Updated 4 months ago