Over the course of the last few weeks, Google has rolled out a new report to certain accounts with eligible traffic. The new report is called “Conversion Probability.” This can be found in the Audience > Behavior section.
What is Conversion Probability and how does it work?
Google is using the same general machine learning tools that power smart goals and smart lists, but instead of engagement metrics such as pages per session, time on site, etc., Google Analytics has focused on likelihood to conversion.
Right now, Google has explicitly said that the predictive likelihood for conversion is within the next 30 days—regardless of the date range that you choose. As Google develops this tool more, we speculate that the date ranges will become more flexible and the tool will have greater capability.
How do we use Conversion Probability and what are we looking at?
As Google Analytics rolls out Conversion Probability in beta, the interface seems rather confusing at first glance.
The top row of data is the historical totals for the date range you selected. Note that this is NOT projected sessions for the following 30 days. In the figure above, the past week saw 20,975 sessions total. This is consistent with the rest of the reporting in Google Analytics for that time period. The gray table below the high level data contains those 20,975 sessions distributed throughout their conversion probability scale.
The chart is divided into rows of “% Conversion Probability.” This is a score from 0-100 (0 being extremely low probability and 100 being extremely high probability) that a user will convert. One thing to note is that the collapsible menus are NOT evenly distributed. The 2-5 collapsible menu only contains 4 points of spread, whereas the 51-100% category contains 50. It’s easy to glance at the report and not realize that these rows are not equal. In the screenshot above, the majority of the 20,975 sessions will not convert and fall into the 0-1% conversion range.
To the right of each probability (whether it’s a grouping or a specific percentage does not matter), there is the ability to create segments, which can then be used in AdWords. Much like the machine learning algorithms it is likely based off of, these segments are imported into AdWords similar to smart lists and smart goals.
If a deeper dive is desired, the default drill down dimension is default channel grouping. This can be changed and filtered much like any other report — the resulting table readouts will also update as you change your filtering criteria.
Things to watch out for
Since this is a new report as of writing (April 20, 2018), the jury is still out on effectiveness and whether or not it’s worth targeting based on Google’s conversion predictions. Certainly, it’s an interesting look into your users, and allows you to compare different cohorts against each other to see how Google’s machine learning tool views them.
That said, this is an output of Google’s machine learning “black box.” While we can surmise that this is based on a handful of principles and methodology such as (at a high level) lookalike modeling, nearest-neighbor analyses, support vector machines/clustering, etc. the exact methodology and code is unavailable to engineers and data scientists outside of Google. This lack of transparency is certainly not unexpected (as this is proprietary), but something to keep in mind as you build and base your AdWords budgets off of this data.
As a new tool, it will be interesting to see how accurately these predictions verify. Since Google is a leader in deep learning and AI development, we’re staying tuned to see how widely adaptable and accurate these conversion predictions verify across various industries. Also, in true Google fashion, expect consistent updates and upgrades to the tool itself, as well as the interface. Here at Elite, we’re excited to get our hands on the new Conversion Probability report and discover its possibilities as we delve further into the age of machine learning within the context of marketing.
Interested in learning more about increasing conversions on your site? Read our Top 10 CRO Tips to help you convert more of your visitors on your site.