Let’s Talk About…Video Fill Rates

recoverstrong logo - rgb
Watching That’s “Let’s Talk About…” series, part of our #RecoverStronger programme, delves into the depths of the most interesting, poignant & fundamental concepts in video advertising – demystifying & clarifying for the industry. Check out our #RecoverStronger hub for more guides, tips, tricks and expert advice.


When it comes to video advertising, fill rate is the most commonly used measure of revenue success.  

Yet ask 10 people what fill rate is and you’ll get 10 different answers.

With fill rate such a fundamental concept and metric to get right, we thought we’d spend this Let’s Talk About… post breaking it all down:

  1. What exactly is the video fill rate? 
  2. Why do so many teams not know what their fill rate is (but might think they do)?
  3. Why is it so important to measure fill rate as accurately as possible?
  4. What can you do to improve the accuracy of your fill rate?
  5. And, finally, what is a good fill rate?


What is the “Video Fill Rate”?

It’s most telling that if you google “IAB Definition Video Fill Rate” the results don’t actually contain any IAB pages on the subject! 

Yet it is one of the most commonly used measures of video advertising success and business performance monitoring. 

In a previous post, where we looked at how to Troubleshoot A Low Fill Rate in 3 Steps, we recommended you adopt a very technical definition:

Fill Rate = impressions rendered / ad requests sent by the client where a play request is present.

But let’s back it up a step and talk about what this really means.

Every day an editorial team produces and publishes video across their sites and apps.  The purpose for this is to attract, win and retain the attention of an audience.

Once gained, the commercial team steps in and attempts to monetise that attention. This is typically done by interrupting what the person wants to watch right at the beginning by showing an ad before the content. (Although with the rise of OTT platforms it’s increasingly likely to happen as ad breaks scattered throughout the playback.)

If this all works, then the publisher makes money.  If this all works at scale then the publisher makes lots of money.

To know if they are succeeding, the revenue operations team will measure and monitor their Fill Rate: ie how well they are managing to get ads delivering into the content stream (the FILL) and if that success is going up, down or staying steady over time (the RATE).

By knowing the dynamics of their fill rate, operational teams become highly efficient as they can focus their energy and (limited) resources exactly where needed – taking the guesswork out of their decision making. 

The dynamics of a video fill rate

The trick to getting an accurate fill rate measurement is to determine the right data points for both the ad REQUEST and the ad IMPRESSION.  

Once you have that, you then determine the difference between the number of times there is an opportunity to serve an ad (ie the REQUEST) against the number of times there actually is an ad served (ie the IMPRESSION) over any given time interval (a minute, an hour, a day, a month etc).

As simple as it sounds, it’s actually very easy to get wrong. And many companies do.


Why do so many teams not know what their fill rate is (but might think they do)?

It’s a pretty controversial statement to say that many teams do not know what their fill rate is.

Indeed a lot of Watching That’s customers claim they know it when we first meet.  To prove it they often simply point to a monolithic spreadsheet that has a column labelled Fill Rate (%). 

But very soon we agree that what they have is not really the true fill rate, although it can come close. 

The reason is because the data points used are not actually the correct ones required for an accurate measurement.  They usually originate from only one system, the ad server, and that system can only see a part of the overall ad sequence – and in the case of a fill rate calculation it can’t see enough of it. 

In essence it’s a tail wagging the dog situation.

To understand this a bit better we need to take another step back and look at the lifecycle of a video view and how the video ad fits into that puzzle.  

A video view is a compound of several different parts:

  • It’s part environment (the code that makes the page, the device used, the network that delivers the video streams);
  • it’s part content consumption (how much the viewer watches, does the viewer skip through the content, is the volume changed);
  • and it’s part monetisation (does the viewer pay for the view, is advertising used, does the viewer leave before the ad finishes, who served the ad, how much did money did it make your business).

Makeup of a video view-1

Each of these parts are typically powered by different systems.  The environment is powered by a web server and page code (or the app store and app code in the case of mobile / OTT).  The content is powered by production crews, cameras, editors, encoders, CDNs. The monetisation is powered by the ad server, the SSP platforms, DMPs, verification vendors.

Somewhere in all that mess of wires, blood, sweat and tears operational teams need to find the key data points to use for their fill rate analysis. 

Which brings us back to the problem with how fill rates are typically calculated. 

The REQUEST data point that is needed for a Fill Rate calculation originates from the Environment part of the video view.  From code on the page or in the app that sends the request to the primary ad server.  

The IMPRESSION data point comes right at the end of the monetisation flow and typically after the ad server has handed off to a 3rd party programmatic partner, which can then hand off again and again (and again!).

video with ad playback breakdown

Faced with many degrees of separation and data silos, teams have traditionally had to make do with what data they had, and that was the ad server data.  After all everything goes via the ad server, so it’s better than nothing, right? Although arguably not by much.

In the end, with a REQUEST number that is too low and an IMPRESSION number that is too high, what operational teams are left with is a fill rate that is not real and therefore – at scale – highly misleading.


Why you must ensure you’re measuring fill rate as accurately as possible

So… it may seem obvious that getting the best fill rate measurement possible is a good thing. But this is more than just an academic exercise. It is so critically important to your business.

Let’s take a moment to examine two real world scenarios that underscore the need:

Example 1: 25% discrepancy between the ad server and SSP fill rates

In this real world case the publisher, using Google Ad Manager, noticed a big discrepancy between the fill rate they calculated for their SSP partner vs the the fill rate supplied by the SSP.

Data table that shows discrepancy between ad server impressions and SSP impressions

At its worst, the discrepancy approached 25%, which had a major impact on their revenue forecasting models as well as the ad server priority algorithms (the ad server thought the SSP partner was performing 25% better than it was).

The cause was that the ad server only reported on the successful handoff of the impression to the 3rd party in question.  However the SSP was awarding impressions based on certain viewability criteria above and beyond the standard viewable impression definition.  So their impression counts were much lower than those in the ad server.

To be able to get the right fill rate and understand exactly what is going on the publisher needed to extend the IMPRESSION data point to the SSP’s marker, and not that in the ad server. 

Example 2: 22% inflated fill rate due to ad blockers, poor network conditions and more

The drop off funnel of the video ad sequence

Because the ad server can only report on what it receives, revenue operations teams are usually working off of inflated fill rates.  The impact of ad blockers, poor network conditions etc are rarely taken into account when analysing video fill rates.

In one real world case this meant a publisher overstated its fill rate by 22% leading the teams to believe they were performing much better than they were.  A false positive that was only felt at their Profit and Loss reconciliation. It showed they were spending much more on video delivery than they were making, which contradicted the fill rate they were using.

There are many things you may or may not want to do in these conditions to recover this lost inventory but what is critical is you understand the size of this drop off.  

Because the publisher only used the ad server numbers (which only counts those requests received) they never got the measure of the drop off; they didn’t know that they had a problem until well after the fact; and they didn’t have the right info to tackle the issue. 


What can you do to improve the accuracy of your fill rate?

Example of increasing fill rate visualisation

Well it goes without saying that you need to make sure you’re getting the right data in through the door to make this all work. 

You want your REQUEST marker to be captured from the client page / app so you have a true measure of total requests being sent; and you want your IMPRESSION marker to be as far down the chain as possible. 

Now that is not always possible so at the very least understanding what do you have and then applying proper multipliers will go a long way in making sure you’re working from a source of truth.

Unfortunately this approach will have little impact on ad server algorithms that are used for forecasting and pacing purposes but it’s a good start.

It’s also important to retain the data points that show drop off within the sequence.  They allow for deeper analysis of each stage making up the ad sequence.

For example, the Watching That platform includes a metric called Use Rate. This measures the drop off of ads that load but the creatives fail to start which marks the hand off from the Request Zone to the Playback Zone of the Video Ad Flow.

Some of our customers call this an Impact Score.  Essentially it allows you to establish where ads are failing to deliver due to creative issues and it is a major source of errors.

Using this and other measures that segment the ad lifecycle can really help in linking overall performance (established by a proper Fill Rate measurement) down to segmenting individual stages and zero-ing on where the drop off occurs. But that is a topic for a future post! 


And, finally, what is a good fill rate?

You may not be surprised to know this is the most frequently asked question we get at Watching That. 

You may however be surprised when the answer we give is “it depends..”!

Not exactly a cop out as there is no one definitive answer here for every business. Although we can range it:

  • 100% fill rate is perfect, and like perfection in life more broadly it’s totally unattainable;
  • 0% is D.O.A and you wouldn’t be reading this article if that were the case;
  • 50% is a great starting point and you should be aiming above that;
  • 80% is ambitious but reserved for specialised inventory groups (specific markets, content verticals campaigns);
  • 40% is likely if you are a volume based publisher with more supply than demand;
  • 25% is a sign that editorial output is out of balance with commercial demand;
  • 60% is where you want to be if you’re optimising your output;


That said it is also important that you carry your video fill rate ambitions through as a function of profitability /yield and weighted against other strategic priorities.  

For example, an editorial-first strategy (where priority is given to increasing the audience size at the potential cost of not being able to monetise all of it) will naturally have a lower fill rate.

Whereas a highly programmatic and open market setup should see higher fill rates but might deliver lower yields.  

So setting the right fill rate target for your business requires an understanding of what is achievable given your setup, balanced with your overall strategic objectives.


So in summary
  1. Measuring video fill rates accurately is critical; 
  2. Many companies try to measure fill rate, get close but ultimately come short;
  3. This is because the data needed for an accurate measure cannot come from just one system (the ad server);
  4. And relying on just one system can cause discrepancies of up to 25% – which is bad;
  5. But the good news is getting a better measure of fill rate can be done ! It just requires moving the goal posts further up stream to the client and further down stream to the programmatic partners. It’s possible with the right setup.

And now you know.


This article is part of Watching That’s Recover Stronger programme – check out our #RecoverStronger hub for more guides, tips, tricks and expert advice for the video industry. And subscribe below to get new updates.