Optimized rotation

Optimized rotation is an advertising strategy in which the best-performing rows within a feed are served most frequently. Optimization is based on click-through rate (CTR) and click-through-conversion rate (if your site has Floodlight tags set up).

How Google's content optimization engine works

Optimized rotation is achieved using Google’s content optimization engine. When multiple rows are eligible to be served, the content optimization engine uses a very complex model to determine the best row to serve on each and every impression. Based on the large scale machine learning systems developed by Google Research, the content optimization engine continually adjusts the predicted performance of element values based on predicted CTR.

The engine takes many different factors into consideration, just one of which is the targeting criteria (e.g. placement, geo, etc.), to predict which row will lead to a click or conversion. Each of these factors is given a weight based on historical data. Over time, this allows the content optimization engine to identify the best-performing row(s) and serve those more frequently. To account for new and/or updated content, rows that are not performing as well continue to be served, just less frequently. The result is the improved overall performance of the ad serving the creative. 

Dynamic optimization vs. Campaign Manager 360 creative rotation

Dynamic optimization controls the selection of content from a dynamic feed. Depending on the type of optimized rotation, the best-performing row is served. Campaign Manager 360 creative rotation controls the rotation of individual Campaign Manager 360 creatives assigned to an ad, some of which may or may not be dynamic creatives. Each feature is set up and controlled separately, so there should be no concern about an overlap of strategies.

Turn on optimized rotation in Studio

To optimize dynamic content rotation, follow these steps:

1. Open your dynamic profile in Studio, and navigate to Step 3: Manage Rules.
2. In the Rotation dropdown, select Optimized.
3. Proceed with your profile setup.
4. To optimize based on conversions, there are a few more setup steps in Campaign Manager 360:
   1. Set up the Floodlight activities on which you want to base your optimization. If you assign weights to these activities, they'll be taken into account by Studio.
   2. Set up optimized rotation in the campaign, in the “Creative optimization” section. For the metric, use either Click-through conversions or Click-through / View-through conversions. (Note: If you use view-through conversions or video completions, Studio will only optimize your dynamic creative for click-through rate.)
   3. Assign the creative to an ad in Campaign Manager 360. Open the ad, then in the “Creative assignments” section, open “Creative rotation” menu. Select the conversion metric you set up in the previous step.

Frequently asked questions

At what level does optimization occur (e.g. within an individual placement, across all placements, etc.)?

Optimization occurs at the impression level. The content optimization engine takes many factors into consideration, some of which are attributes specific to the impression. Each of these factors is weighted based on historical data. Therefore, optimization really occurs at the impression level, ensuring that the optimal row is served based on the algorithm and historical data available for each impression.

For optimization purposes, is it better to comma-separate multiple targeting values or list the values out in individual rows?

You can use whichever option works best for your strategy/feed structure. As far as optimization is concerned, comma-separating or listing out individual values achieves the same results.

How quickly does optimization start working?

There is no set number of impressions. If a particular variant has much higher performance than others, that information will be picked up and optimization will kick in sooner. If the difference in performance between the best and worst variants is small, optimization will take longer to identify differences. If all variants perform exactly the same, optimization will never kick in, because there's nothing to optimize on.