Anomaly Alerts
Geetanjali Tyagi avatar
Written by Geetanjali Tyagi
Updated over a week ago

On this page:

What are they?

Anomaly Alerts, also called Auto Alerts, are not set up by you but are generated automatically by the Optmyzr system.

They can be generated for any Google Ads, Microsoft Ads, or Facebook Ads account linked.

Note: Anomaly alerts are currently not supported for Amazon and Yahoo accounts.

They are set up by default for the following metrics: Cost, Impressions, and Clicks. You will receive a notification whenever the performance of either of the three metrics changes by 75% or more on a particular day, both positively or negatively.

Note: The anomalies will start the calculation at 2 PM in the account time zone.

The system considers the last 3 weeks' data and estimates the value for today. With that, it considers the difference between the estimated value and actual value. If there is more of a difference than expected, then the system will trigger the alert.

These alerts are generated and monitored by the Optmyzr system, so you will be automatically notified over email (once a day, evening time), on the Alerts page, and on Slack if you have enabled this option.

Note: Please note that there can be some differences between the alert you see here and the information you received over email. If you wish to be alerted immediately whenever those automatic alerts are triggered, we recommend setting up the integration with Slack or Zapier.

You can find these alerts under the filter "anomaly" in the alerts side tray (go to the Alerts page from the Insights menu or from the bell icon on the top right corner of the Optmyzr interface):

If you are not able to find an anomaly on the triggered alerts page, it can be because it is already resolved, and so it can be closed automatically by our system.

How are these Alerts Calculated?

For anomaly alerts, we consider the last 28 days as the current value and the previous 63 days as the target.

For example, if we were to look at May 30th, the current value would be the total cost from May 2nd to May 30th (28 days). The target, in this case, would be the average from May 2nd to the previous 63 days (February 27th to May 1st). This means that the target is the cost of those 63 days/63.

To compare this data with the current 28 days, the system will multiply the average of the above-mentioned by 28. These values are compared, and then, if required, an alert is triggered accordingly.

Note: If the account is new, then the target will be 0, so the system won't trigger alerts until it has enough data.

Who receives these alerts?

When an anomaly alert is triggered, the system will notify the account owner, as well as whoever has starred the account on the MCC Dashboard. If there is no account owner and no team member has starred that account, by default, the Optmyzr subscription owner will be notified (over email and slack if enabled).

Important note: If the value remains unchanged during the next 2 or 3 days after the sudden drop, this will be considered the “new normal” by our tool, and alerts or reminders will not be triggered anymore.

To avoid this kind of situation, the Rule Engine can help you set up recurring custom alerts that would notify you if an account stops receiving traffic suddenly (using the “If and Then” statements and automating the strategy to run daily).

Don't hesitate to reach out to if you need help to get started. We can create a sample strategy in your account that you will then be able to customize and automate.

Why am I seeing that the anomaly alert was last updated x days ago?

This can happen more multiple reasons:

  1. Data is the same for the past x days, and the anomaly alert is still not closed due to that.

  2. It could be that we lost access to the account.

  3. In the case of Facebook, it could be an exemption thrown by the Facebook API, and we are not able to retrieve the data.

For example, if the cost is 0 for the past 3 days, the last updated date will remain 3 days ago. If the access is lost, we still show a trend chart with previous data.

Did this answer your question?