Alerts: what they are and how they work
crystal is always on and ready to give you insights on your data. You can ask your virtual advisor to send you customized alerts, in order to instantly receive notifications on the most significant information for your business.
To start off, keep in mind that currently alerts can only be created for topics that either have a single line, multi-line or scaled-up number visualization. Line charts monitor numerical data over a period of time and a scaled-up number displays a single value. Check out more about visualizations here.
Alerts appear to the user as notifications when certain events and/or anomalies are detected. Please, keep in mind that if you use alerting for multi-line visualization, the alert will be triggered once the anomaly or event is detected for any of the lines.
By tapping the notification, the related topic is shown, with a brief introductory message.
The user will also receive a notification when the set value goes back to normal.
But how can we define an anomaly? An anomaly is a numerical value that can be perceived as unexpected. There are two ways of identifying such a situation. One is by employing a static threshold (the technology behind the Threshold Alerting feature), while the other considers a statistical method to automatically and adaptively recognize anomalies, and in this case we will refer to adaptive alerts (the technology behind the Anomaly Detection feature) .
Let's see in detail how alerts work.

Behind the scenes: how alerts are processed and sent

To detect possible anomalies and, consequently, send out alerts, crystal periodically checks the data connected.
As of now, the possible intervals for data checking are every minute, every 15 minutes, every hour, every 6 hours, every day, every week.
When an alert is set, crystal starts to check every minute for changes in the data.
After 15 minutes, if no change is found, the check is delayed to every 15 minutes.
If after an hour no change is found, crystal will check every hour and so on.
If a change is found in any of the intervals mentioned above, crystal will again within the same interval. If another change is found immediately after, it will reduce the check to a smaller interval.
For example, if a difference is found, and, after a day, another one is also found, the system will start checking every 6 hours.
This algorithm should guarantee a regular, dynamic, check of the data sets, to ensure the users’ anomalies are always spotted on time.

Threshold alerts

This type of alert informs you when a topic’s data goes above or below a specific numerical value threshold set by you.
The most important factor with this type of alert is that you know exactly what data change should trigger a notification.
For example, if your topic is Total sales growth rate you can set a static alert to know when the growth rate goes above a given threshold value.
Depending on the column type selected in the configuration of the topic, the threshold value can be an absolute value (e.g. “tell me when total sales growth rate exceeds 1 million”) or a percentage (e.g. “let me know when my total sales growth rate goes above 10%”). The value type can not be changed through the advisor: it will always be the type that was originally configured.
This type of alert will send you a notification every time the threshold has been reached and then again when the value has gone back to normal.

Anomaly detection alerts

Anomaly detection alerts inform you when an anomaly or irregularity occurs, compared to the usual behavior.
This type of alert works best when you don’t have a precise data change in mind, but instead are interested in general to monitor a specific topic.
For example, you can ask the advisor to monitor International total sales and an alert will be sent every time a significant change occurs.
The more data you have, the more precise crystal is in understanding when an irregularity occurs.
Over time a continuous estimate is created. Each time new data is checked, if it has a very small probability to belong to that estimate, an alert is generated.

Qualitative alerts

Qualitative alerts inform you of a status change with your data, including all non-numerical but qualitative changes.
For example, you can ask the advisor to alert you when a product goes out of stock or, more generally, when a product changes status.
Qualitative alerts vary from case to case and are flexible. Focus on a specific metric and a relevant status change to monitor.

How often could I receive an alert?

An alert’s time range and its granularity are inherited from the topic’s default time filter, selected during its configuration.
For example, if you have a single line chart that by default shows the last 30 days and data points are broken down by day, those conditions will be applied to its alert and you can not modify them through the alert set up. Only an admin that has access to the Self-Service Console can change its configuration.
Time range and granularity are important to know because crystal checks your topic’s data based on that and not on how often the data source is updated.
Whether that be a specific threshold or for an adaptive alert, data is monitored every time a new data point is created for the topic.
For example, if a single line topic is configured to show a data point for each week, the data will be checked on a weekly basis. If, at this point, the data has arrived at the threshold indicated, or, in the case of an adaptive alert, if the data is not close to the trending estimate, a notification will be generated. In this case, a static threshold alert generates a notification you will also receive a notification once the value has gone back to normal.
So, for example, if your threshold, “above 100” was reached and a notification was sent, the next time a data point is created, if the value is below 100, another notification will be sent stating that your topic is “back below the threshold!”.
It is better if the data source granularity has a higher frequency than the topic to avoid having partial values for new data points. For example, if the data source granularity is daily and the topic granularity is weekly, the last point risks to be a partial value.

How to set an alert

Setting an alert is really simple and direct: all you need to do is to ask the advisor in a natural way, for example: “Please, tell me when the number of items produced goes above 100 pieces”, or “Can you monitor the trend of national sales by month?”
The following are the primary items to include in your request:
  1. 1.
    The request has to start out by including something like, “notify me” or “advise me about” and, in the case of static threshold, also “set threshold”. This way crystal knows you are indicating an alert.
  2. 2.
    Next you need to include the name of the topic. For example, Sum of orders over time by day, so that crystal knows which topic to monitor.
  3. 3.
    You can optionally indicate filters that are associated with that topic to narrow down the insights.
  4. 4.
    If you know exactly what data change you want to be alerted about, you need to specify the numerical threshold and the direction, either above or below that number.
When asking crystal to set a static threshold alert, the numeric value needs to be specified in full. You cannot include “,” or “.” in the number, you cannot abbreviate the number and you cannot include a currency symbol.
For example, you can not specify a number like $ 3,4K: instead, it needs to be 3400.
Note that alerts are filter sensitive, meaning that if a filter value is specified in the related topic, you can add it to the alert.
For example, if you want to be notified when the average engagement score goes above 7.5 points and the topic has a filter country set up, you can also ask “Notify me when the average engagement score in Italy is above 7.5”.

Where to find active alerts

All of your active alerts can be found in one place. From desktop, click or tap on your Profile in the menu bar and on mobile from the settings icon in the upper right corner of the app.
Use the search bar to look for specific alerts: it helps a lot when you have set up several alerts.
For every active alert, you will find the title of the topic, a threshold value and its direction if it was a static threshold alert, and filters if any had been applied.

Delete an alert

You set up an alert a while ago, but your business objectives have changed and you don’t want to receive notifications on that insight anymore.
No worries: you can always delete your alerts.
To delete an alert, just click on the corresponding bin icon in the alert management section. You will get a message to confirm you really want to delete this alert.
Click on Yes, delete to confirm you want to get rid of the alert.

Edit an alert

There is no way to change an alert once is set. You must delete it and then replace it with a new one.

Be always updated

Alerts are crucial to staying up to date on your data and being able to make data-driven decisions in a timely fashion. Setting up a strategic alert system will allow you to keep your business under control and be constantly updated on any significant changes.
If you are using crystal app, be sure to turn on push notifications, so that you can receive real-time alerts even when you are not working with crystal.
If you are using crystal on desktop, or in case you haven’t activated push notifications on the app, you will receive notifications when you open crystal.
We hope this article was helpful! Check out more articles from the advisor section.
If you have any questions about crystal’s features, you have encountered a problem or you would like to share your feedback, contact us using this form.
Copy link
On this page
Behind the scenes: how alerts are processed and sent
Threshold alerts
Anomaly detection alerts
Qualitative alerts
How often could I receive an alert?
How to set an alert
Where to find active alerts
Delete an alert
Edit an alert
Be always updated