How search queries work with Google Analytics' question answering feature

29 October 2019

Did you know that the searchbox at the top of Google Analytics doubles up as a so-called 'question answering feature'?

Yes, OK, you can use the searchbox to search for reports. Start typing a keyword, such as 'Goals', and it will show you all the matching reports: Goal Overview, Goal URLs, and Goal Flow. That's pretty handy.

But it is also possible to use it to ask questions and request information in natural language. Examples that Google Analytics itself suggests are:

  • Any anomalies in sessions this month?
  • How to set up a property
  • Top countries by users
  • Pageviews in last 30 days

Depending on your query, Google Analytics will then provide matching insights.

Google recommends asking questions that are as specific as possible (with a defined metric/dimension), and that use the exact language for custom events and goals.

This can lead to the natural language search requiring language that is somewhat... unnatural in order to work as expected. For example, another question that Google suggests is 'Worst pages by load speed'.

Worst pages by load speed

This works, as do all of the following variants:

  • worst pages by load time
  • slowest pages by load time
  • slowest pages by load speed
  • slowest loading pages
  • pages with longest load times

But NOT simply 'slowest pages'. This gives you, slightly surprisingly, insight into 'bottom page by pageviews'.

So it looks like you need a combination of:

load + page + slow/bad/long (ie a negative term)

With 'load' being the metric ('Average Page Load Time'), 'Page' being the dimension, and 'slow/bad/long' being the sort order.

Hence, omitting 'load' causes Google to misunderstand the request. In this context, Google sees 'slow' and 'bad' as synonymous, so 'slowest pages' gives you the same insight as 'worst pages' - which, in the absence of a defined metric, is assumed to relate to pageviews. Hence, 'bottom page by pageviews'.

What you don't know won't Bert you

I expect the question answering feature will improve over time as Google makes use of the search data to increase the number of questions that GA can answer accurately.

It's also possible that the technology behind the very recent BERT update to Google's search ranking system will come into play here. As Google says,

BERT models can... consider the full context of a word by looking at the words that come before and after it — particularly useful for understanding the intent behind search queries...

Particularly for longer, more conversational queries, or searches where prepositions like “for” and “to” matter a lot to the meaning, Search will be able to understand the context of the words in your query. You can search in a way that feels natural for you.

But for now, I will leave you with this short video on Google Analytics' question answering feature and how best to (currently) use it.

Leave a Reply

Your email address will not be published. Required fields are marked *

Copyright © James Clark
linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram