Basic Auto Optimization

• Introduction
• For Creatives
• For Sites

Introduction

Auto-optimization was designed to automate the decision-making process when optimizing a campaign. Based on the optimization rules that you define, the system will make decisions on your behalf. You can think of it as a transparent algorithm that you configure and control.

The optimization algorithm evaluates by zone (not by site). A zone, also known as a placement, is a combination of Network, Site ID, Dimensions, and Page Position. For each zone, the algorithm starts by looking at the current hour's data, which is the smallest piece of statistical data available. If a sufficient sample can not be made from the current hour, then we will add the previous hour's data, continuing in this fashion until we find a sample meeting the Sample Size/Metric criteria. If, after looking back 1 full week, we have not found a sufficient sample, then the zone is skipped. If a good sample is found, then we calculate the Optimization Metric, and if it does not meet the Threshold, we set the zone offline. If the Site to which the zone belongs does not have any more zones online, then we offline the site itself (this is really just a visual cue for the user, since all the zones are off anyway).

The optimization job runs constantly. It will run through every campaign with optimization enabled, then wait a couple minutes and start at the beginning again. Thus, campaigns are continually polled throughout their lifetime. Note that this makes it possible for the same impressions to be included in more than one optimization run, or for some to not be included at all. This would be dependent on the sample size, how long it takes the campaign to reach it, and how often the campaign is evaluated by the optimizer.

For Creatives

Auto-optimization is intended to distribute the majority of impressions to your best-performing creatives. By adjusting the metrics used and the optimization settings you can customize the way that your creatives are optimized:

• Optimization Metric – Choose your optimization metric. Highest CTR? LP CTR? Conversion Rate? By default, creatives are optimized by CTR, because it’s typically the first metric with which creatives are judged and optimized. It also doesn’t have any extra requirements – like our LP Click-Thru code for LP CTR, or our conversion pixel for Conversion Rate.
• Evaluation Period – The Evaluation Period dictates the length of time with which the Auto-Optimization uses to make decisions and determine confidence. Typically, the longer the Evaluation Period, the more accurate the optimization. If the optimization metric is CTR – the Evaluation Period is measured in 6-hour increments. For LP CTR, the Evaluation Period is measured in 12-hour increments. For Conversion Rate, the Evaluation Period is measured in 24-hour increments. By moving the slider left (faster), decisions will be made quicker, which might skew the accuracy of the optimization since the sample period is too short.
• Evaluation Group – Prior to reaching the minimum Sample Size, ads must be evaluated in a default group. By default, all new creatives are evaluated in the Control Group, so that they can receive enough SOV, and thus impressions, to reach the minimum Sample Size.
• Sample Size – You can specify what you believe to be a significant sample size for optimization. This number should theoretically be higher for evaluating LP CTR or Conversion Rate, but for evaluating CTR, 10,000 is generally a good sample size for banner ads.
• Control Group Size – “Control” is another word for highest-performing creatives. By default the slider is set 20% of the way, which means that the best 20% of creatives will be considered the “Control Group”. Sliding it further left will reduce the amount of ads in the Top-Tier group, resulting in fewer “control” ads.
• Control Group SOV – The Control Group SOV (share of voice) controls what percentage of impressions will be given to your best creatives. By default, 80% of traffic will go to the top-performing 20% of ads.

For Sites

Auto-Optimization for Sites allows you to set a CPC or CPA goal that any site or placement in your campaign must meet. If the goal is not met, after your sample size is reached, individual placements or entire sites will be automatically set offline for you.

Optimization Target – Choose your optimization target (eCPC or eCPA) and set the goal value. For eCPA, you would set the maximum cost-per-acquisition that your campaign is allowed to reach on any particular site. For eCPC, you would set the maximum cost-per-click that your campaign is allowed to reach on any particular site.

Sample Size – This is the “look back window” that will be used to evaluate the site. Sample size is continuous, meaning that it is always looking at the most recent impressions to optimize.

The sample size can be defined by three different measurements:

• Spend – Set a dollar value as the sample size. The system will continuously look at the last $XX.XX spent to determine your eCPA or eCPC.
• Impressions – Set a number of impressions as the sample size. The system will continuously look at the most recent # of impressions bought and determine eCPA or eCPC based on those impressions.
• Hours – Set a number of hours as the sample size. The system will continuously look at the past X hours to determine eCPC or eCPA.

If you are managing a wide variety of sites/campaigns, we suggest you give this feature a try.