From Overwhelmed to In Total Control. How a Billion a Month Broadcaster Transformed their Ad Monitoring.

“What do you mean we haven’t been serving ads on [a very popular platform] for over a week?!?!”

“How did this happen?!”  “Why didn’t we notice it sooner?!”  “This is going to seriously cost us….”

Sound familiar? The above were all from a streaming broadcaster who generates over a billion of ad impressions per month across over 25 playout endpoints (of which 90% of them are distribution partners).

They had just noticed that one of them had stopped serving ads.

To understate it the situation wasn’t good.

What made it more intolerable was that it went unnoticed for so long, even though they are using some of the best ad tech and platform monitoring tools around.

Something had to be done and quick.  Because it wasn’t a matter of IF but WHEN it would happen again.

Job #1 was to take stock of their situation.

Despite having Google Ad Manager and Datadog in their tech stack, why did this still get missed?

Venn diagram that shows the massive part of the playout network that is not covered by existing ad and platform tech for monitoring.

Every streaming business worth their salt will have a premium ad server at the core of their ad tech stack.  And they will have a premium platform monitoring service in their application stack.

The ad server takes care of the mammoth task of serving ads at scale, and the platform monitoring service takes care of making sure their application code is continuously up and running.

Great, but who’s taking care of making sure all the ads are playing back as you’d expect on code and platforms that are not yours?!

  • The ad server takes care of delivering ads but is not built to track and monitor them.
  • The platform service is taking care of your owned and operated apps but knows nothing about the distribution end points you don’t own or the intricacies of the dynamics of ad serving.

As the streaming industry becomes more distributed (with the rise of FAST and YouTube as major audience pools) the gap in the coverage between the ad server and the platform monitoring service grows wider and wider.

It is this gap that the broadcaster inadvertently fell into.  And it’s a very long drop indeed.

Job #2 was to get delivery signals back from every ad.  Not just the ones on their owned and operated apps.

Deploying a pixel in the ad server was the first step to getting in control.

Screenshot of the Google Ad Manager interface that shows where in the UI you add tracking pixels

The easiest bit was to take a Watching That Pixel and push it through the ad server.  For GAM this was deployed at the ad unit level.  No need to fuss about the creative level. This is something ad servers support as standard so that box was quickly ticked.

Once in place, it was sent in every ad response, whatever it’s destination, delivering a real-time heart beat of every ad.

This data was then immediately and continuously available in a centralised control dashboard.

But that is a tonne of data to keep across and they have other things to do and focus on.

Job #3 was define the data expressions and signals that are meaningful.

Not All Data Is Equal.  To Change the Game They Needed To See The Data In “Their Way”

Screenshot of a sample Watching That table widget that shows Ad unit level 2 and level 3 together

“I did it my way” is the hallmark of every successful company.  To be competitive is to do something unique and, in many cases, novel.

When it comes to the expression of data that can mean using proprietary metric definitions or defining common concepts in unique ways.

In the case of this billion impression strong publisher it meant the most interesting data signals came at the intersection of two tiers of their ad units.

Ad unit level 1 represents their over 50 content channels and Ad unit level 2 represents their over 25 distribution end points.

On their own they are interesting but not particularly actionable.  Paired together, the channel on the end point,  they provide critical life signs from their revenue generation engine.

But with over 1,250 permutations that would need to be monitored for anomalies continuously 24/7, they needed something special.

Job #4 was to deploy Watching That’s Monitors with Dimensional Support.

One Monitor That Watches Over Them All

Screenshot of Watching That Monitor List page

This was the easiest job of them all.  In Watching That they simply created a monitor, configured it to watch a specific channel over all endpoints for any anomalies.

“SWEET!, This is perfect” was the direct quote from the head of ad operations.

Job done!

Now In Control of Over A Billion of Impressions a Month Across 50 Channels and 25 Endpoints

“Hey did you know we had a drop in impressions on our news products 10 minutes ago across our distribution partners?”

“Yes, I did. We were alerted as soon as it happened. It’s fine and expected.”

“Ok great.”