Monitoring Module

Introduction

The Monitoring module is the home to Watching That’s Anomaly and Validation Monitors.

Anomaly Monitors track real-time output data (e.g. log files, pixels, plugins) sent to Watching That during ad playback. They help identify unexpected behaviour in live delivery.

Validation Monitors check input data by calling your ad server’s API every 10 minutes to confirm that key settings and tag configurations are correct.

Getting Started

To access the Monitoring page, use the navigation bar at the top of the webpage. The two types of monitors are accessible on seperate pages within the Monitoring space.

Open

To create a monitor, select the pink button on the top right hand corner of the page.

Key Features

Anomaly Monitors

Anomaly monitors can be configured for any metric you have running through the platform. There are four different formats of anomaly monitor that you can leverage. Monitors are run in 10 minute or 1 hour intervals.

Flatline

This triggers when the selected metric ceases reporting data - this can be across a certain field or for the entire data source. For example triggering when impressions have stopped delivering. To mitigate noise in the alerts, you can choose to ignore dimensions that don’t meet a certain volume of the chose metric as well as setting filters to customise your alert further.

Open

Flatline Example: I want an alert to trigger when any Series (programmeseriesid) flatlines gross impression, however I only want this alert to trigger for Series that typically serve at least 100 impressions per 10 mins

Relative Change

This triggers an alert when there is a significant deviation from established normal trends. The monitors work by selecting your metric and dimension first, then calibrating the type of change you want to be alerted for. You can choose to be alerted for increases or decreases, define the deviation percentage, set the time granularity, and specify the comparison window size.

When creating a Relative Change monitor, the platform displays a graph of the previous 24 hours and shows how many alerts would have been triggered, this is especially useful for calibrating the monitor to avoid excessive noise from too many alerts. You can also configure alerts to ignore dimensions that trigger too many alarms within a 24 hour period (for example, dimensions with high volume that naturally fluctuate throughout the day).

Open

Relative Change Example: in this example I wanted an alert for gross impressions at a Series level when there is a change of 10% or more. The comparison window is set to 6, this means the previous 6 values will be averaged, and if the new value is >10% deviated from this, an alarm will trigger.

Threshold

This triggers when metrics exceed defined upper or lower limits (similar to Flatline, but the threshold can be above zero). When setting a Threshold Monitor, consider the level of severity you want to be alerted to. You can set thresholds for both upper and lower limits, or just one. The graph shows whether any triggers would have occurred in the past 24 hours based on your settings. Time windows can be set to 10 minutes or 1 hour.

Open

Threshold Example: in this monitor I have configured the alerts to trigger when impressions exceed 6000 or drop below 1500 for a 10min window and to alert me after 2 consecutive breaches.

Lookback

This triggers when there has been a deviation from expected daily or weekly seasonal patterns. This is useful when volume naturally fluctuates throughout the week , for example, a spike in impressions during weekend sports broadcasting.

Lookback Monitors learn typical behaviour over time. A daily lookback monitor needs 1–2 full days of data to learn and establish a daily baseline. A weekly lookback monitor requires at least one full week to start detecting meaningful patterns based on historical weekly trends.

When evaluating the data, the monitor compares the same time periods, for example, Tuesday at 6 PM this week will be compared to Tuesday at 6 PM last week. This makes it particularly effective for spotting irregularities in recurring traffic patterns, while reducing false positives caused by normal time of day or day of week variation.

Choosing between daily and weekly depends on the kind of fluctuations you're trying to track. If your traffic varies by time of day, a daily monitor is likely best. If your traffic patterns repeat across days of the week, a weekly monitor will offer more relevant comparisons.

Open

Lookback Example: in this example I want to be alerted if there is a 10% drop week-on-week in gross impressions.

Validation Monitors

Validation Monitors use your ad server’s API to check for configuration issues before ad delivery begins.

They’re especially useful for businesses with regulatory or compliance requirements, for example, ensuring that all creatives include a Clock Number or that every piece of programming content has an appropriate Rating.

Open

Available dimensions for customers using Freewheel.

These monitors can be configured to flag both missing values (e.g. no brand name assigned) and specific conditions you want to track, such as campaigns with budgets below a minimum threshold, or high budget campaigns that may require closer oversight.

Validation checks run approximately every 10 minutes. However, the monitor will wait for the previous check to complete before starting the next one, so while it’s designed to run frequently, the actual timing depends on how long each validation takes.

Open

Validation Monitor Example: This monitor will check that all Videos have a rating applied - if a Video does not have a rating, this will trigger an alert via Email or Slack to the notification channel selected.

For more details see the Validation Monitors page.

FAQs

Q: Can I use validation monitors on any adserver?
A: Right now validation monitors are only available for customers using Freewheel or GAM. If you would like to leverage validation monitors but use a different adserver, please reach out to your customer success representative.