SEO’s (and now some SEO tools) tend to conceive of search intent as directed to one of four or five basic motivations or “intents”:
- Informational
- Commercial
- Transactional
- Navigational.
These intents were described in a 2002 paper by Andrei Broder, “A Taxonomy of Web Search“, which has since become the foundation for the SEO industry standard practice in discussing and classifying intent.
Overlooked in discussions of Broder’s taxonomy and how to identify it in keyword research is a section in Broder’s paper where he referred this classification to “The evolution of search engines.” Broder here describes an evolution from information retrieval models focused exclusively on text within a document (until 1997), to models focused on ranking algorithms using cumulative linking data on the model of academic citations (PageRank) to inform ranking. At the time he wrote his paper, in 2002, Broder heralded the rise of third-generation retrieval. This paradigm (for which he recommended his taxonomy as a solution) would be making efforts in the direction of “blend[ing] data from multiple sources in order to try to answer ‘the need behind the query'”.
This is a critical point, as it actually pointed the way beyond the relevance of Broder’s taxonomy, which I argue is by now too dated to be relied on for keyword research. And indeed, search engines have evolved far beyond merely accounting for raw link-score data and term frequency in and across documents (tf*idf). Now search engines are in fact able to return results which demonstrate a much more sensitive understanding of intent.
The “need behind the query” is now served by search engines in ways which are more nuanced and multidirectional than the classification into four possible intents. While building out full awareness of this reality into a search optimization and content strategy requires a deeper process (and will always be quite limited in scope and brand-specific), a few takeaways about intent and strategies to tackle it are available.
Intent is about a searcher first, and a search engine’s ability to rank the right content second.
Intent refers firstly to the intent of a searcher, and can be seen in query disambiguations; what is returned by a search engine is usually our best hint on what to optimize for, but even that depends on the number of quality results for a keyword on the top of the search results.
Intent can’t be simply captured by the big 4/5 intents. Or it could, but this in itself won’t tell you that much. Consider the example of related queries for the search “ETL”:
Intent is defined here by query patterns (e.g. does it have a brand name in the query, the words “buy” or “cheap”, the phrase “how to”, or the comparative “vs”, as in “X vs Y”); as well as by which pages are actually ranking on search. For higher volume and more generic queries, understanding these intents may be enough to reliably optimize homepages, product pages, and so on. But as you conceptualize your keyword strategy as a funnel, and move down-funnel to serving increasingly more specific and customer-relevant intents, it can be helpful to begin to define them beyond the big four or five.
To begin to enhance the four most conventional intent labels, it is helpful to consider the breakdown used by the SEO tool Sistrix. For all keywords, Sistrix returns the following intents:
- Know
- Know Simple
- Visit
- Website
- Do
You can find Sistrix’s definitions of intent here. I find their breakdown valuable because it hews more closely to guidelines that Google uses for manually evaluating the quality of results. I also prefer it because it splits informational queries into know and know simple. Informational queries make up huge volumes of search data, but without mindful targeting can lead to low value traffic. Together with “do”, we are already able to better classify this huge intent segment of search results into more basic or definitional types of information needs, more complex needs (I need to know how something is done), and more proactive needs (I need to know how to do this).
With this in mind, I prefer some version of the following breakdown for higher level intents:
Understand intent yourself to do what Google can’t (or competitors won’t)
There is a certain awkward reflex SEOs sometimes engage in when speaking of intent, by describing what is currently ranking on Google as more or less the equivalent to a declaration of searcher intent, rather than what Google simply happens to rank. Google itself enforces this: in the guidelines cited above Google actually connects result types (e.g. blue link, featured snippet, map pack, video, PLA) to intent. But guessing which media or rich results satisfy a query (and with plenty of rank testing behind it) isn’t the same as satisfying the need behind it.
While Google’s rankings are generally a pretty reliable guide to what will rank (and are therefore critical to observe for the mapping and strategy stages), the outcome often becomes less obvious with longer, more ambiguous, less-searched queries for which related and quality content is not necessarily available or perceived by Google. In these cases, Google may simply be backfilling content from strong and/or seemingly relevant pages.
This is where taking a more strategic view to search data can help you better understand your customers’ likely intent as searchers – especially valuable for companies in a tight space which need to produce content, or companies with a PLG strategy looking to both exploit and wedge open new search spaces.
I suggest tackling this by adding specific intent labels to your search data which aren’t just thematically oriented (“they want to learn about this specific ETL integration”) but also understand the information need being expressed in a query. It is of course always integral to try and separate signal from noise on search: volume, query length, theme, and mode of expression are all ways to filter your way towards that need.
Finding new quadrants (up and to the left!) for active, directed searches which are trying to do, implement, compare, fix quickly, debug, solve a problem (which your product solves) – this sort of intent labeling can help you to better hone in on real user needs. Once you have parsed queries with advanced intent labelling, you can apply the following rules of thumb to identify when a certain label points to actionable insights in terms of your ability to serve your customers’ needs in ways which are unavailable through the generic taxonomy:
If a query has the following features, there is likely no benefit in straying from generic labels
- large search volume (over 100 monthly searches)
- is not very expressive (e.g. does not contain modifiers like “how to”, “connect with”, etc.)
- returns a page in the top position which you as a searcher feel basically satisfies the intent behind that query
If however a query has the following features, intent labels might be able to help you surface high intent keywords:
- lower search volume (under 100 monthly searches)
- is expressive (contains modifiers)
- the top ranking pages seem to serve your customers’ intent quite directly; or they seem to be fairly poorly selected.
In the last condition, familiar-looking results on the top of the SERPs are probably a safe bet that you are heading in a viable direction. But, especially if the top ranking results don’t seem to have much salience for the keyword selected, that might indicate that Google is backfilling results with some nearest neighbors. In other words, it is not always a good idea to take the backfill as canon, especially when you think you can spot high intent searches which are currently underserved by Google.
In these cases, you may have an opportunity to produce net-new content which is original in being able to satisfy this currently unsatisfied search intent. Intent labeling can help to surface and organize this research and give shape to your content production efforts.
To get here, some spreadsheet work is necessary. You can sift your search data in this way by filtering for search volume, wordcount, modifiers, or other attributes, and then examining keywords and results one by one.
This process introduces more precision to search, which is always to some degree a vague space. And ultimately, it helps serve the real business case for SEO, which is optimizing a channel for your users, and supporting content and web strategy which tries to meet your searchers where they’re at.