We’ve launched some major new features which will give you an intent score against each keyword to help you with your SEO and content marketing campaigns! These new User Intent scores are available for each monitored keyword in each project and are also available for every SERPs API request.
In this post, we share some details of what the new features look like and how SEOs can use them to good effect for their SEO and content marketing campaigns. We’ve also shared some of our thinking behind our approach to analysing user intent in search and how we’ve constructed our model(s) and have contrasted this to some limitations we have experienced using other keyword intent models promoted by leading SEOs.
Why is User Intent important in SEO today?
Clearly, there’s been a heightened interest in this whole topic area by SEOs well before the BERT update; even long before the Hummingbird algorithm update. If you want chapter and verse on this then you could start with this post on Searchenginejournal.
Suffice to say, I think the bi-directional nature of the BERT algorithm is interesting as it shows the additional lengths Google is striving to go to in its quest to properly understand user intent – and let’s face it – they’re only just getting started. Where are we going to be in 10 years’ time?
So, my advice is pay attention. We can clearly see why Google wants to understand user intent – because it wants to be the world’s best search engine and that means being the best at knowing what users are looking for and serving it to them quickly.
As marketers, this means we need to pay attention to the different types of competing pages that Google is rewarding with top organic ranking and featured snippet positions for related clusters of keywords. We also need to ensure that our landing pages closely follow the type and format of content of the competitors’ top ranking pages and then try and enhance your content, if you can, so it’s a little bit better still than the competition.
At the risk of sounding pedantic; we prefer using the term ‘User Intent’ to ‘Keyword Intent’. This is because ultimately users are all that matters. Google cares about users and we care about attracting as many users as possible to our websites and apps.
Whilst keywords are useful in helping us understand one or more users’ intents, two different users may start their search using the same keyword but have very different purposes in mind. Likewise, two users with the very same intentions may start their search journeys by describing what they are looking for in very different terms.
Google is very, very good at understanding this, making sense of it and serving what it deems are relevant results – and if real-world user behaviour shows that the results are not as good as expected then it will change its algorithms.
Granted it is not perfect yet (nor is Bing or any other search engine) – but what Google thinks the user intent is of paramount importance to us as SEOs if we want to rank for a bunch of relevant terms around that intent.
For this reason, we have built a classification model that analyses Google’s search intent as a proxy for user intent. We’ve been experimenting and training models for well over a year now and have manually reviewed at least 40,000 different keywords and SERP results.
To spare you more technicalities at this stage; what you see is Phase 1; the beta launch of our user intent features. This means for every single keyword ranking check, we analyse the make-up of the first full page of the SERP (and ignore the rest of the top 100 results). We look at the prominence and order of the organic and paid SERP results, the Universal SERP results featured, the navigational elements and the types of ranking sites. This allows us to develop a user intent model that works in the markets Google serves irrespective of language.
In phase 2, we’ll go one step further by visiting and analysing the top ranking pages, but for now we think we have a model that should give you pretty accurate results internationally.
Pitfalls of common approaches to SEO Intent Methodology
As you may have gathered by now, we have eschewed the dictionary based approach to understanding search intent promoted and favoured by some SEOs.
Typically, the well-meaning advice is to do your keyword research and then classify your keywords based on whether they contain terms like the ones in the table below.
For a rough estimation exercise it’s not a terrible starting point but it has many limitations and problems. Here’s just a few:
This is a rather limiting set of keywords – how far do you go?
This does not help you resolve queries including ambiguous terms? e.g. ‘CFD’ could mean ‘Contracts for Differences’ and have some commercial intent or it could mean ‘Computational Fluid Dynamics’ and be informational in nature.
It forgets implicit intent, e.g. where the user’s intent is implied due to the nature of the query, e.g. “Dry cleaners”, “used car”, “post office” and the type of ‘near me’ queries users ask without typing ‘near me’, where Google assumes you are looking for something in your vicinity.
You cannot simply say a user’s intent was to make a purchase (or a keyword phrase is transactional in nature if you prefer) just because they have used the word ‘buy’ in their query. You don’t have to think for more than a few seconds to find queries where this doesn’t work well;
‘Best Buy’ – Branded.
‘How to buy a house’ – Informational
‘What not to buy for baby’ – Informational
‘When is a good time to buy shares’ – Research
Some keywords will appear in multiple categories, e.g. ‘Compare Sony TV prices and deals at Best Buy’.
It is not scalable across languages; and even in the same language certain keywords can have different user intent at different times of year (e.g. Search for ‘Cheltenham’ around March and you are more than likely looking for information related to the famous horse race meeting, not the pleasant county town.)
It ignores context and ignores a user’s location or device; Google calls this Visit-in-person vs Non-visit-in-person intent. e.g. “Cinnamon” could be a nearby restaurant.
It ignores the changing dynamics of intent, e.g. User Intent may change over time due to timeliness, seasonality or freshness.
Pitfalls of common approaches to SEO Intent Models
Some SEOs only split keyword intent into three broad categories (Navigational, Informational and Transactional) but we don’t feel this is granular enough and in practical terms it doesn’t help you differentiate your landing pages to attract users and buyers at different stages of the buyer journey.
The 3 types of Search Intent
Navigational – information relating to a company, person or brand
Informational – general or factual information about a topic
Transactional – information to aid a purchase decision
i.e. The extension of this simplistic approach is that all terms that are not branded or informational in nature are transactional which is clearly not the case.
We believe there are at least two major pitfalls of splitting your keywords into just 3 types of Search Intent.
It over-simplifies and over-generalises the type of queries and intentions a user has and dumps anything that is not obviously a Navigational or Transactional query into one big useless Informational bucket. As an SEO, you don’t know whether a commercial keyword with high commercial intent has been sent into the Informational or Transactional bucket.
This also ignores the possibility that keyword queries don’t fit neatly into one bucket at all, as transactional and informational keywords often have brand modifiers.
Authoritas User Intent Model
Keywords can be ambiguous in nature and if you read Google’s Search Quality Raters guideline you will see that Google recognises that the SERP intent can often be mixed.
We therefore do not just give you a single overall intent classification for a keyword unless it really is an absolute 100% nailed-on certainty.
At Authoritas we have therefore developed an alternative 4 Type Model of Search Intent that facilitates a more practical implementation of User Intent for SEO:
The 4 types of Search Intent
Navigational – the user is looking for a specific entity: e.g. A company, website, brand, product, event, location, social media handle, etc.
Informational – the user is looking for general facts or information about a topic that has no commercial intent.
Research – this is ‘commercial research’ where the user is researching a product or service and is looking for relevant data about what to buy and where to buy it to inform their decision.
Transactional – the user has narrowed their research and is now looking to purchase a specific product or service.
We refer to this as NIRT.
Our intent model classifies each keyword into four categories and importantly a score is given for each category ranging from 0 to 100.
The categories are probably very familiar to you if you are an SEO; Navigational, Informational, Research and Transactional.
This way, for a particular keyword (or better a group of related keywords for the same landing page) you can develop a feel for the spectrum of user intents and if there is a mixed or overall Dominant Intent it will be obvious.
We also separately show the Dominant Intent (N, I, R, T) where there is one – in some cases this can be evenly split between two categories (e.g. R/T).
By splitting queries that have a commercial or monetizable informational search from a search that is more factual and non-commercial in nature, you can actually focus on optimising for the terms that are likely to convert into sales and not just the keywords that you feel might be relevant or have a high search volume.
Furthermore, by making a distinction between Commercial Research and Transactional you are getting a better picture of the buyer journey and the content that a user is seeking out at different stages of their buyer journey. This will help you when you start planning your content strategy.
This new feature in Authoritas will go live w/c 18th May 2020. It is available both in the platform and also in the SERPs API.
V7 SEO Platform – User Intent is now in the Keyword Ranking module
Please navigate to the Keyword Ranking module and you will now see a User Intent bar next to each keyword. This will show you the intent score for each category next to each keyword.
Here’s an example.
You can also click on the column selector to add the ‘Dominant Intent’ column.
Hopefully, it’s fairly self-explanatory.
You can also use the filters to filter your keywords by tag, user intent and/or landing page.
Practical Implementations of User Intent for SEO:
Here’s some suggested use cases and ways of utilising these great new features in your SEO and content marketing campaigns.
Review your top landing pages with good potential traffic.
Compare the average ranks of the ranking keywords by intent category.
e.g. ‘R’ keywords average rank = 8, ‘T’ keywords average rank = 25, ‘I’ keywords average rank = 76 and no ‘N’ keywords rank for this page.
Determine what this is indicating:
You’re on the right track: high ranking keywords appear to have consistent intent. If it’s fairly consistent for the keywords you are ranking well (1st page/2nd page) then you’re probably on the right track. To optimise, it’s probably now just a question of comparing your page with the top competing pages to ascertain manually the differences between the main content types and formats and optimise your page accordingly. You may of course need to look at backlinks and technical SEO factors.
There’s a mis-match: e.g. Is your page ranking well for ‘Research’ phrases and poorly for ‘Transactional’ phrases or vice-versa? In which case you might want to split your content into two separate pages. Time to invest in creating new content and optimising your content.
You can also use this data earlier in your SEO strategy for keyword research. If you research thousands of keywords for new content ideas you can automatically organise them and segment or cluster them to fit them to your buyer journey. This allows you to match them to the types of content you should create. You can also filter out keywords with navigational intent which may be harder for you to achieve top rankings for if Google is rewarding a different site with top branded results like site links and Knowledge panels. Sometimes this is not always obvious to you, since there are niche brands in every market that have generic names, e.g. ‘Little Black Dress’ is a high volume search term but also a nice brand with an exact-match domain (EMD).
SERPs API Response
You will now see some additional data for every keyword request. This will include;
A score from 0 to 100 for each category Navigational, Informational, Research and Transactional
A Dominant Intent flag – this looks at the scores and gives you a letter or letters (in the case of mixed results) indicating what the dominant user intent looks like
A Local Intent flag – this shows by detecting the presence of local result types that indicate there is some local intent
Remember, this is just Phase 1 of our User Intent features. We’d love to hear from you with suggestions about how you would like to use this data and any suggestions to help us improve this further.
Feel free to contact us or leave a suggestion or feedback in the comments below.
The API supports all major search engines in many of the large markets and languages globally (desktop or mobile rankings). It is also geo located to decimal lat/long and can be used for high volume queries across all the major search engines.
Looking for new ways to produce more content, more efficiently? (Who isn’t?) Authoritas has designed a new module to help with just that. Our new AI Content Assistant tool will help you create new content with only a few keywords or phrases.
Things used to be a lot simpler. Google would just serve up approximately 10 blue, organic links per page (plus some Ads at the top), so to work out your organic rank, you simply had to see how far down that list of links you were.
Back in April 2015 Google released an update that started to penalise sites in mobile search which weren’t ‘mobile friendly’. This meant that some sites were more likely to rank differently in mobile search compared with desktop.
We made a small release today to add support to the platform for keyword rank tracking in Yandex. This might seem like a small update, but it actually consumed quite a lot of development time, as we found the Yandex markup to be quite a bit different to other search engines’ markup.