Fast and awesome way of doing bid only RLSA

Fast & awesome way of doing bid only RLSA

Since recently there is a new feature in Google Analytics; Session Quality. You can use this to make RLSA audiences in AdWords, and it is easy to set up! In Google Analytics, you can find it under Audience > Behaviour > Session Quality.

What does it do?
Google measures the engagement in a session and scores it from 1-100, with 100 being the closest to a transaction/sale. You’ll need a minimum of 1.000 transactions a month to use this feature. Google uses multiple metrics to decide session quality, like bounce rate, time on site, page visits, visit depth, recency, etc. and all of that is off course calculated with the magical use (and black box) of machine learning and AI.

Using it for remarketing
Session quality can be used for Remarketing audiences in AdWords. It is pretty easy to create for example 5 audiences; the first audience being very unlikely to convert, and the last audience to be very likely. For example, for one of our clients we created the following audiences:

  1. Session quality | 1
  2. Session quality | 2 t/m 5
  3. Session quality | 6 t/m 20
  4. Session quality | 21 t/m 50
  5. Session quality | 51 t/m 100 

See screenshot 2 below for an example of how to set up the remarketing audience.

Please check if these audiences are grouped correctly in terms of quantity (the audiences should have sufficient users and traffic) and quality (the audiences must have a similar grouping of quality).

Now test
If you have created your audiences, you can add them to some campaigns. Here you can test the following hypothesis;

  • Do the audiences correctly predict actual conversion rates? 

So you should be seeing the actual conversion rate increasing with the quality of the audience moving upwards.

Below you can find the results for one of our iProspect clients. The data is made up, but the performance of the audiences relative to each other are real:

As you can see, the audiences show exactly what we expect; the conversion rate increases with each better audience. Even CTR increases with each step. Also, in terms of conversion quantity, the audiences are evenly distributed. This is important, because it makes sure that all of your audiences will have sufficient data to set bid modifiers.

Why is it awesome?
The awesome part is that you get very sophisticated remarketing lists which would be impossible to create manually.

The disadvantage of current RLSA audiences is that you often group sessions based on one metric, e.g.: the depth of a URL (homepage, category page, check out. etc.), or time on site (<15 sec, 15-50 sec, 51 sec > etc.), recency (15 days, 30 days, 45 days etc.). However, it is super useful to combine all of this information (e.g. “Depth” x “time on site” x “recency”). Doing this by hand is basically impossible and would result in thousands of audiences. With this feature, these metrics are taken into account, but with a manageable group of audiences.

Also, it is pretty nice to make segments in Google Analytics and do some extra analyses:

  • Which of your keywords and campaigns deliver highly engaged users?
  • Which conversion paths are most effective?
  • What percentage of your users conduct low-quality sessions?
  • Which conversion paths do the lower-scoring users follow? Are there opportunities along those paths to deliver more effective marketing?

Why is it lame?
The lame part is that it is super black box; we have to trust Google on their explanation on how it works (more details here). As such, it is important to test; are the audiences really predicting conversion rates correctly? Are they better grouping sessions than your current audiences? Is it stable over time? Is it predicting conversion rate evenly over time? Does it fit your client’s objectives? Etc.

Don’t miss the opportunities of machine learning
Google is more and more heading in the direction of machine learning and AI. Session quality is just the latest example of this trend of new possibilities. Although it is black-box, it does offer some really interesting opportunities. We advise testing this on a small scale on a number of campaigns. Only by testing you can find out which of these new developments are worth moving forward in.

Screenshot 1: