Pricing Viewability: The Shift from Placement to Ad Call
By Asaf Shamly | January 28, 2021
– Viewability is a key metric in our industry today.
– Advertisers will pay more for inventory that has a higher likelihood of being viewed by users.
– Publishers should be able to set higher prices for that inventory and increase yield.
– Browsi research directly connects our engine’s ability to predict viewability on the ad call, and to connect that prediction to improved bottom line.
An ad unit is the sum of its parts, and in those parts are gold. In terms of viewability, the ad requests within a single ad placement hold a treasure trove of opportunity. Specifically, they enable pricing according to ad calls, connecting viewability directly to the bottom line. Let’s explore.
Yield more revenue on current inventory by setting tiered floor prices that sync with higher predicted viewability.
What is viewability prediction?
Viewability prediction is a measurement that states the likelihood of a single ad call (future impression) to be viewed by a specific user on a specific page.
Browsi viewability prediction leverages over 50 user-engagement and page-structure data points to create an accurate viewability prediction on each ad call in real time – which is then delivered via key values to the publisher ad server for targeting campaigns.
To illustrate, on the far left of the table below we see impressions divided into 10% increments of viewability prediction. In the second column we see ad calls that have been predicted, and next to that the actual viewable impressions. As the actual viewability (right side) falls within the prediction range (left side), the predictions are accurate and can be used for setting smarter floor prices that leverage this granularity and increase revenue.
Browsi’s Test setup
To test the correlation between predicted viewability and advertiser willingness to buy higher cpms on inventory with a high likelihood of viewability, we took a two-step process:
Setting a baseline
To establish a clear starting point, we attributed a single pricing rule (Branded 0.45$ and Anonymous 0.1$) to all inventory on a single site.
We ran this setup for 10 days and analyzed the data, establishing a benchmark for measuring future optimization.
Creating the test
After running the baseline rules for 2 weeks, we created three new pricing rules:
First pricing rule – predicted viewability of 10% – 60% (Same floor prices as prior to test)
Second pricing rule – predicted viewability of 70% – 100% -(Doubled the previous floor price)
Control – 10% of all ad requests (Same floor prices as prior to test)
We ran the new floor prices for 10 days and then analyzed results against our baseline:
The outcome is clear:
20% uplift in total eCPM*fill rate on the high predicted viewability tier.
15% increase in revenue overall.
The way it works is simple:
Advertisers pickup the new floor price and analyze whether it’s worth paying more for that inventory (as opposed to the previous price).
Because this process mimics a simple update on a floor price, the change is natural to an advertiser as they check the coverage and eCPM of any pricing after any update.
The difference here is that only the most premium ad requests are adjusted with higher price; after about 2 days, advertisers not only bid more but also cover more ad requests.
– Publishers can align with the advertiser buy process by using technology and Machine Learning to better sell their inventory.
– Viewability is a key metric today, and clearly defining the value of higher viewable inventory must be communicated to advertisers via aligned pricing.
– With a 15%-20% uplift in revenue with no changes to your code or any need for a site redesign, the opportunity for leveraging viewability at the ad request level – via prediction – is real.