Campaign or bid strategy bid optimization is governed in Filter settings.
None (No Minimization or Optimization)
When selected, the system will always bid with your set max CPM. This may be useful e.g. when targeting deals that have a fixed floor price.
With bid minimization, the system will continuously minimize bid CPM levels without sacrificing your campaigns or bid strategy's delivery targets. This protects the buyer from paying more than needed for targeted impressions.
If "Bid Optimization" is set, the system will score the bid opportunity with respect to the set campaign KPI. This score (0 - 100%) will be multiplied with your set max CPM to get the bid CPM.
Adjusts bid proportional to likelihood of impression being clicked
Adjusts bid proportional to likelihood of impression being viewable
Adjusts bid proportional to likelihood of impression leading to a conversion.
Requires conversion tracker to be set up, and a minimum of 100 conversions tracked. Model trained on 30 days of data.
If the optional Min. Score setting is set, you require the score to be at least this setting. This can be useful if you'd like to only bid on highly scored opportunities. The downside is that the pool of available inventory is limited, see Optimization and Pace below.
Optimization and Pace
By default, the system will maintain a minimum bid CPM floor to keep your campaign pacing towards its goals. For opportunities with low score, the bid might be raised to this minimum.
If the Min. Score setting is used, no minimum bid CPM floor is used, this means that your campaign or strategy risks not meeting its goals.
Things to think about when using Bid Optimization
Using bid optimization will on average lower you eCPM with 10-30%. This means that you should set a higher max CPM than your wanted average CPM, to facilitate buying the most valuable impressions.
Also, keep in mind that the optimization does not exclude poor performance completely, but rather places lower bids, making you pay less per unit of performance.
How does bid optimization work?
Machine learning models are used to predict the effect of bid opportunities on some common campaign KPI's. The models are trained from known effect of previously bought ad impressions. The machine learning algorithms used to train the models consider attributes such as when (weekday, time of day), where (site, category, ...), what (ad size, category, ...) and who (device type, previously clicked ads, ...)
When running, the ML model will score how well a given bid opportunity fits the campaign KPI. The score between 0 and 1 will be multiplied with the set max CPM bid, to get the final bid CPM.
Advertisers with more than 1 million impressions per week will get models trained specifically with their own buying performance data. Models are trained and updated once each day.