Chapter 2 - How User Intent Can Help in the Future

After a Google engineer (let’s call them Stevie) has decided on a hypothesis for improving search results based on a machine learning system, they will most likely pre-train the model with known data.

Basically, Stevie will give the system large volumes of data with known results. In the case of search, they would give the system queries, results, and a rating of the results on a numeric scale of its success.  Is it a good result? Does it fulfill the needs of the user?

So already out of the gate we see the impact of user intent coming into play.  The entire system is built on the likelihood of a query meeting it.

One of the likely ways the data would be gathered for this training stage would be to pull queries and their results, determine what a likely successful results page has as its engagement factors. A SERP for “weather victoria bc” would have very different success metrics than “best rowing machines for a home gym”.

They would need vast amounts of data for large numbers of queries, a grading of various resulting elements on a page including features snippets, knowledge panels, those ten blue links, and any other elements.

This would be fed into the system to pre-train it.

What is essentially going on during this stage is that Stevie will have created a hypothesis around improving search results using machine learning – let’s say, in what entities are used on a page. For example, if a page is about tourism in New York, what entities should exist on a page likely to satisfy a user?

The hypothesis that certain entities being present on a page or site could suggest that those resources will provide a higher level of user satisfaction with a result might have started when Stevie noticed while planning a trip that most of the successful pages related to tourism in NYC include the boroughs of Queens, Bronx, Brooklyn, Manhattan and Staten Island. From there Stevie may have considered that there are likely other entities common to satisfying users for the query, and that this is likely true in other cities and perhaps for other queries and query types.

From there an algorithm would be trained with data like:

What you see here is simply all the entities that exist on a resource sent to the algorithm along with the resulting score. The machine learning system would create an algorithm to weight the entities by their type, frequency, location, etc. in a way that reliably produced the resulting score.

This would be repeated thousands … hundreds of thousands … maybe millions of times.

With enough data, and the system set to understand how the input (entities) can get as close to the desired output (the known score).

We can see how user intent drove the initial formation of the system itself. But that’s rather indirect.

In stage two there’s a simple adjustment to the way the system is trained. During this stage both the system and the engineers get to learn.

 The system gets to continue to refine its understanding of the path to optimal results, and the engineer gets to understand how well (or poorly) it’s doing.

The illustration for this stage would look something like:

In this stage the system is given the data, processes it based on the pre-training, and that data is then compared with the known-right answers and StevieBrain in continued to be trained and the engineer is kept aware of how well or poorly the system is doing.

The goal in this stage is to get it as close to producing results that are better than those that are produced by current algorithms. Once this threshold is crossed and consistently confirmed, the system can be deployed.

At some point we run out of training material, and even if we didn’t – training sets can only go so far.  You can’t improve by only trying to do what’s been done.

The system needed to be taught and trained to recognize what we’re looking for, but the only ongoing measurables that can be brought in at the scale required are from the system itself.

Core success metrics then, needs to be built into any deployable machine learning system that might be integrated into search.  While these metrics are complex because they involve user interaction with a results page and variables as broad as not just a query but the user’s device, location, time, and countless other features, they can be produced.

A separate ML algorithm can be trained and created simply to monitor a user interaction with a result page and produce said score based on the myriad of signals it’s being given. That’s what machines do. This is likely (though not necessarily) what produced the score in our first two stages.

This score would then be used by StevieBrain to judge whether it is improving on itself consistently. To judge its ongoing success once deployed as queries, environments and the world at large changes.

Google has said that user behavior isn’t a ranking factor. User behavior is much more powerful than that.  It doesn’t just impact the rankings of your site, that’s too easy to game and too onerous for Google.

John Mueller eluded to the impact of user intent when he answered a related question in 2018:

User intent impacts the entire algorithm.

A poor user experience with your site won’t result in your site being “penalized”. However at scale it will result in an algorithmic shift that would see your site negatively impacted.

So, it’s not a penalty, but it sure may feel like it.

You’re probably expecting us to say something cliché like, “Just build content that people like.” But we need to think like machines, because machines understand people at scale as opposed to people living in their own bubble. We have all been surprised by the results of A/B tests because people liked something we could not have predicted, so too are they likely to search with an array of needs, wants and desires that’s well beyond what we can predict.

However, a machine can. Or at least, over time it can learn to. So, let the machine tell you what to do. When you’re thinking about the content to produce, it’s always good to start  with the SERP.

What formats and information does Google feel are relevant to the topic.  This is what they and their billions of queries think fulfill the user intent. There just might be a bit of insight there. Another wonderful free (for now) tool is AlsoAsked.com. It follows the path of what people also ask on a topic like:

Not everything might be relevant to you, but it’s relevant to somebody and almost always worth knowing. You will never stay ahead of the machines, at least not for long. With ML, they are changing too fast, but one thing is constant, the target answer.

Features, factors, signals and indexes change – but the target remains the same. The user, and the SERP that best meets their intent.

It may seem cliché to say that that needs to be the target but, that’s the only target we have that we can keep up with.

User intent and search intent has been spoken about for some time, but it looks like 2021 will be the year where more resources are dedicated to understanding it and building websites with that in mind.

Therefore it is important to write with the user intent in mind and not focus on the keyword itself. Pages which have tailored their content towards a certain search intent, such as navigational, informational, commercial research or transactional stand a better chance of  appearing for a variety of relevant long tail terms that answer the searchers’ queries and their intent.

We have seen that users can often spend a lot of time in the informational phase of their buying journey.  But just focusing on producing content that satisfies Informational Intent isn’t going to help your business grow revenue (in most cases).  Of course, not everyone who is searching online is in a position to buy either, so you may not get very far if you just focus on content that caters to Transactional intent.

If you are interested in Navigational Intent and improving your brand’s SERP then a good starting point is to review the Tea Time SEO Show on How to get and manage Knowledge Panels with Andrea Volpini and Jason Barnard (The Brand SERPs guy).

We would recommend a balanced approach.  For example, structuring one’s content strategy around user intent across the buyer journey helps with allocating resources to the key pages which will be the most relevant for the predominant intent type you are focusing on in each stage of the journey.

This guide to user intent is designed to help you understand further the importance of this aspect of SEO and bring that to the forefront of your SEO efforts in 2021 and the years beyond.

Thank you again to our co-authors for sharing their expertise with us. If you would like to contribute to this or any of our other Ultimate SEO Guide series then do get in touch.