The short answer:
Yes. We use a standard Machine Learning performance measure called Precision/Recall.
The long answer:
What does Precision/Recall mean?
This measure looks at how well our system is able to select relevant links through two lenses:
Precision: How many of the selected links are relevant? (How many clips were duds?)
Recall: How many links were selected out of all the relevant links? (How many relevant links did we miss?)
A system with perfect precision would be able to select ONLY relevant links, while a system with perfect recall would be able to select ALL the relevant links. Unfortunately, precision and recall work against each other, so it’s not really possible to have it all.
The trade-off between precision and recall
If our precision was too high we would only be able to provide a few links. While these few links will be very relevant, many other equally relevant links will be left behind. If our recall was too high, we would be able to provide all the relevant links, but there would be a lot of junk links in between.
The trick is to find the perfect balance between the two, where we provide a set of targeted links that are relevant enough to increase the campaign’s performance and large enough to ensure high view volumes. But, it also means that the final outcome will contain a small ratio of noise, usually about 5% of the clips.
Although we have a default balance that works very well, we can change the balance for individual campaigns. If you have a special requirement to increase the precision of a specific bucket of placements, we can set things up to return a lower number of very accurate placements.
Feel free to contact us about your specific media buying optimization requirements.