SEOs traditionally say that a specific search query can be classified as either navigational, informational, or transactional. The categories were originally published in 2002 in a peer-reviewed paper by Andrei Broder who worked for Altavista (remember them?!) at the time.
The categories that Broder came up with have been invaluable to SEOs for many years, helping many of us explain the different types of search query that we should consider.
However, it’s time to revisit these categories to see if we can improve their usefulness in a world of direct answers, apps, intelligent personal assistants, and other developments.
Recap
For those who haven’t reminded themselves recently, here is a quick recap of the three categories:
- Transactional – Here the user wants to get to a website where there will be more interaction, e.g. buying something, downloading something, signing up or registering etc.
- Informational – This is when the user is looking for a specific bit of information.
- Navigational – The user is looking to reach a particular website. There’s only one likely destination that they’re looking to reach.
Google, in their human rater guidelines, call these three categories:
Google also widened the definition of the categories slightly from their original paper. Interestingly, Andrei Broder, who created the original categories, now works at Google.
Beyond the Web
However, what is important to understand is that Broder’s was proposing “A taxonomy of Web search” — i.e. the categories were designed for web searches. Even though Google widened the definition of these queries in their Do-Know-Go framework, they also still discuss them in terms of web search.
Here’s the thing — so many of the searches we do nowadays are not web searches at all. Various papers have estimated differently, but most estimates (here, here, and here) are that around 50–80% of searches fall into the “informational” category, which is the category that often lends itself to a direct answer:
Furthermore, we are more and more frequently searching, not via a web-powered search bar on a desktop machine, but instead via an Intelligent Personal Assistant app (IPA):
In the most recent revision of their human rater guidelines, Google introduced a new category — Know Simple — for informational queries that can be answered with a short (< 2 sentences) answer that has an uncontroversial answer, i.e. the type of answer that is suited for these direct answer-type responses. You should read this great post on takeaways from the latest quality rater guidelines from Jennifer Slegg to read more about Know Simple and other interesting tidbits from the Google document.
New considerations
Google’s addition of Know Simple indicates that Google understands that the Do-Know-Go categories are no longer broad enough for the modern search landscape. However, I believe that, even with the addition of “Know Simple,” the model could be improved upon.
SEOs have traditionally used the navigational / informational / transactional framework to classify the searcher’s intent. We believe that a large part of this (maybe not all) can be captured by considering how user intents may change and be better served when they’ve been conducted via an Intelligent Personal Assistant app.
The last couple of years has seen an explosion in the functionality and usage of an increasing number of personal assistants:
- Apple Siri – Built into every iPhone. Now also “Proactive” in a fashion similar to Google Now.
- Google Now –Technically, the name refers to a certain set of functionality, but people are increasingly referring to the “assistant” app and its functionality with this name.
- Microsoft Cortana – Originally just in Windows phones, but now available for Android as well. Microsoft has a great research team and are working hard on Cortana.
There are also a couple of newcomers that look very promising, but which are not fully released yet:
- Facebook M – An interesting cross between Artificial Intelligence and human workers who help with certain tasks, Facebook is looking to really take things to the next level with an assistant that can do things for you.
- SoundHound Hound – A preview video the company put out last year got a lot of attention for Hound’s speed of understanding and flexibility with complex chained queries. I’ve tried it myself and it’s pretty awesome.
Upgrading Do-Know-Go
At Distilled, we’ve been thinking about this a bit, and have gone through many iterations trying to work out how a new model should look. We still aren’t sure we’ve cracked it, but wanted to share what we have so far.
We propose adding a second axis or a second row to the model, such that we have something like this:
Notice we’ve moved “Go” to the side, because we don’t think it changes in a very meaningful way (you’re just looking to go to a certain destination, be it app or web or whatever) — but it’s also the least interesting category!
The logic of this model is that previously, the navigational / informational / transactional categorization allowed you to two things:
- Understand the user’s intent for a query
- Frame how an appropriate landing page (or set of landing pages) should be formed for this category of queries on your site
None of that has changed when we’re still in the the “web model,” but we now need to extend that model so we can do those same two things above when considering IPAs.
In our model, we can see that “Informational/IPA” is where a huge amount of Know Simple queries exist. Then we have a box for “Transactional/IPA,” which is also very interesting.
What’s fascinating is that, in the case of both “Informational/IPA” and “Transactional/IPA” instances, SEOs don’t yet have a good understanding of how to do any sort of optimization of improvements (which is to be expected as they are still developing). Let’s discuss each.
Informational/IPA
Google puts their Know Simple queries into this cell but, as you’ll see, Informational/IPA also covers slightly more ground.
Know Simple queries are for short, relatively factual searches; a good example is [how old is barack obama], which Google can answer from their own knowledge as it’s something that changes infrequently.
However, there are other searches that fall into the Informational/IPA category, such as our example of [are the trains to London on time]. Google says that “queries where different users may want different types of information” are not considered Know Simple, but such queries clearly fit into our structure there.
These latter queries are the interesting ones, as they’re the types of queries that search engines will be looking to fill by connecting to APIs. These are very related to “data-driven search.” Also worth noting is that if 50–80% of queries are Know queries, we can imagine that this cell could serve a huge number of queries.
An Informational/IPA query via Google, served via the facts in their Knowledge Graph. |
An Informational/IPA query via Hound, pulling live and context-based data via an API. |
A particularly interesting variant of Informational/IPA queries are those which involve some sort of computation. I don’t mean requests such as [the square root of 1764], as they have no “search” aspect. However, a query such as [100 US dollars in British pounds] requires the IPA to fetch the exchange rate for you, and then do something with that answer before giving you a response.
Currently, the capabilities here are still hit and miss; only Amazon’s Evi IPA successfully answered my query [is barack obama older than his wife], whilst Google, Siri, Hound, and Cortana all sent me to a web search.
As the capabilities of IPAs improve, you could imagine some complex compound queries that could fit into this category.
Transactional/IPA
Transactional/IPA queries, which are the transactional relative to Informational/IPA, are queries where there’s an intent to do something further after getting the query response. The initial response may be served within the IPA where you can do some additional filtering before you move to an app to complete the purchase (though, in the future, that part may be unnecessary), or the initial response may be opening an app right away.
|
|
|
|
A Transactional/IPA query via Siri; you’re able to go several steps into the funnel before finally the “Buy Tickets” button takes you to an app. |
||
This category feels like it’s going to become absolutely huge in the near future; currently on iOS, full integration with Siri is only available to selected partners, but it seems likely that wider integration is going to become available. With Google Now and other platforms, we’re likely to see similar patterns. Once apps can integrate right into the personal assistant apps and add functionality in terms of being able to do things, then the range of available queries available will explode. Being able to “rank” is going to be very heavily based upon being a flexible and comprehensive service for that niche.
Future and wrap up
It seems inevitable that search is moving away from just being about web search and 10 blue links, and so it seems inevitable that we’ll need to update our models to keep up. The original navigational / informational / transactional categories were designed for the web, and the model proposed above was built to allow us to extend that model into this new world.
The original framework was used by SEOs to help us understand and categorize queries and to help educate and persuade clients. This new framework allows us to do these same things, but expand them to cover searches done via Intelligent Personal Assistants.
I would love to hear from everyone on in the comments on whether you think there are more opportunities to improve this model. Distilled’s R&D team is working on better understanding how to do SEO in a world of Intelligent Personal Assistants, so be sure to watch this space!
