Google Analytics Intelligence is an umbrella term that refers to a number of different features within Analytics. All of these features are powered by machine learning and are designed to turn your website’s data into actionable insights without any additional human input.
For the purposes of understanding what Google Analytics Intelligence is and how we can use it effectively, we can group the different features into two categories: insight and conversion.
The ‘insights’ features
The insight features can be found in the Analytics Intelligence sidebar that you can access by clicking on the blue logo in the top right of your Analytics window. These features provide ways of accessing different data quickly. The two different tools are:
- Answers: you can ask a question like ‘why did my traffic change today?’ or ‘which browser do most of my visitors use?’ and Analytics Intelligence will source a graph or table that gives you your answer. It also has suggested questions to try.
- Insights: even without being asked questions, Analytics Intelligence will display insights from your data that it predicts will be useful to you.
The ‘conversion’ features
These features can’t be found all in one place and don’t have any functional connection to each other. The reason I’ve grouped them like this is that they all predict how likely your site’s visitors are to convert in a future session, albeit in different ways. These features are:
- Smart Goals: if you’re currently unable to measure conversions from AdWords clicks, Smart Goals identifies which sessions are the most likely to lead to conversions, so that you can use this information to optimise your ads instead. Smart Goals needs to be linked to your AdWords data and only affects PPC efforts.
- Smart Lists: Analytics Intelligence identifies users that are likely to convert in their next session and compiles this information into a ready-made remarketing list. According to Google, it’s designed to give marketers a successful start to remarketing. It’s not designed to provide permanent help.
- Session Quality: A report within Analytics for sites that have at least 1000 monthly ecommerce conversions. It provides a score that helps webmasters to identify how close different users are to making a purchase.
- Conversion Probability: A dimension (conversion probability) and metric (average conversion probability) that provide an estimate of how likely different users are to convert in the next 30 days.
How does Analytics Intelligence work
Before we dive into further applications, we need to take a step back and look at the principles behind how these features work. They are powered by machine learning, which, at its most basic level, is a way of getting a machine to crunch large volumes of data in response to a problem and spit out what it thinks is the most likely solution. There are a few different examples of this at play in Analytics Intelligence:
- Natural language understanding is foundation of the answers feature. Using Google’s vast natural language databases and new data from Analytics users, Analytics Intelligence can take a question that could be phrased in any number of different ways and return the data that it has decided will be most likely to satisfy the intent behind the question. For the SEO folks reading this, this is a small scale application of the kind of machine learning we see in RankBrain.
- Predictions based on your activity powers the Insights feature, where Analytics highlights trends and data it predicts that you’ll be interested in. The insights that it provides will become more useful the more you use Analytics. There is also a feedback ‘yes/no’ option for both insights and answers that help refine the algorithm further.
- Predictions based on site visitor data is the core of all of the ‘conversion’ features mentioned above. All of Analytics’ scores and predictions about who will convert in the future come from an analysis of existing user data. The more data you have, the more accurate the predictions will be.
Analytics Intelligence’s inherent limitations
Given that Analytics Intelligence uses the basic principles of machine learning, it’s no surprise that it’s also subject to the same limitations. There are two things that you have to bear in mind when evaluating the usefulness of Analytics Intelligence: data quality and programmer bias.
Data quality
“Machine learning algorithms are only as good as their data,” is a phrase that you’ll very quickly come across in some form if you read into machine learning or AI. It’s a fundamental limitation inherent in the way machine learning works. These programs use training data to understand trends and inform the predictions they can make. If they have a lot of high-quality training data, they’ll make more accurate predictions. If their sample size is small or even worse, not representative of the wider data, their predictions will also be skewed and inaccurate.
This is why many of the Analytics Intelligence features, especially in the conversion bracket, require a certain number of visitors or conversions before they provide data. Any lower than the minimum threshold and the AI insights just aren’t useful. Even if you meet that minimum, you should bear in mind that the smaller your data set is, the less you should trust Analytics Intelligence. In any case, you should approach the information the conversion features give you with healthy skepticism, especially if they seem off based on your business experience.
The issue of data quality is also why the features in the insight category come with the option to select whether or not the data presented was useful – this provides additional data beyond what the algorithm can gather from your browsing patterns.
Programmer bias
While I wouldn’t expect this to be as much of an issue for machine learning features dedicated to this kind of task (as opposed to more sensitive subject areas), we need to remember that machine learning is not yet at the level where the AI can decide for itself all the different factors to take into account. Programmers still need to set the algorithm up in such a way that it can filter useful and not useful information and variables.
Thus, what the Analytics Intelligence features deem to be important when calculating conversion stats or showing you insights will be, at least in part, determined by what they were programmed to look for. Google is, of course, going to get a lot of that right, but, for different businesses and industries, it might not be 100% on the money.
The crux of the matter is that you shouldn’t prioritise Analytics Intelligence over your own experience. It’s not that clever (yet).
Using Google Analytics Intelligence
Even with those caveats mentioned above, the various features offered by Google Analytics Intelligence can be very helpful. At the very least, they provide a new way to access data that we might not have come across ourselves. At their best, they can inform our strategies and help make both paid and organic marketing efforts more successful.
Smart Goals and Smart Lists
These features are particularly useful if you’re setting up various paid campaigns, whether that’s an initial PPC campaign or a remarketing campaign further down the line. The data provided by Analytics Intelligence can point you in the right direction, giving you something to build upon as you gather data and practical experience. If you have a lot of experience running paid campaigns, or you’ve been running campaigns for a while, it’s likely that you can use the data at your disposal to make more comprehensive lists and strategies anyway.
Session quality and conversion probability
In general, the best way to use the information provided by these features is to evaluate it alongside all the other information you have available. It would be a mistake to assume that these estimates are 100% accurate, but not as big a mistake as it would be to ignore them completely. Given that the figures produced by Analytics Intelligence are drawn from user engagement data, you can see them as a summary of the performance of a particular channel or demographic. They are helpful in identifying trends and in showing you the areas of the site that are the least likely to convert visitors. In this way, these figures can inform your strategy and give you areas to focus on in future PPC, SEO and CRO efforts.
Insights and answers
The features within the insight category are most useful for giving webmasters quick access to data that’s otherwise hard to find or produce. They can also give a snapshot overview of trends that you should be aware of within your own data. The insights and answers produced may not always be exactly what you’re looking for, but they should improve with use and if you give feedback. These features can highlight areas to focus on in reports and strategies and can increase the breadth of data that you’re aware of going forward.
What’s your experience with Google Analytics Intelligence? Was it helpful, or is human insight always better? Let us know in the comments!
