Where Will the Next Trillion Searches Come From?

I want to talk today about a shift in searcher behaviour as big as the growth of mobile. That’s quite a claim, so let’s back up and get some context:

We don’t have perfect information about how many searches are performed each day, but we can piece together some clues:

  • Around 2009-2012, Google was seeing “billions” of searches / day - this was largely desktop search

  • Adding up what we know about the growth of desktop search since then, and Google’s market share, that means there are at least 3 billion desktop searches / day today - or roughly 1 trillion searches / year

  • We know that mobile has grown astronomically fast, with Google telling us that mobile search volume outstripped desktop in western markets in 2015 - with additional growth since then, and the even heavier skew to mobile in developing countries, that’s probably another trillion searches / year

So where will the next trillion / year come from?

If desktop search is the first trillion, and mobile search is the second trillion, where will we find the third trillion?

Well, for a start, there is some growth still to come even in searches like those that people are already doing even in the most mature markets - the best argument I have for that is this chart which shows the growth of e-commerce, and how all of Amazon’s growth fits into around 8% of the total retail market. Just as e-commerce has further to go, so there is still untapped search demand:

But that’s going to be slow going - it’s not going to add a trillion / year any time soon.

So what might?

Well, my theory is that it’s going to come from searches people aren’t doing yet.

Here’s some data and evidence:

Daily information needs studies show unmet needs

Google runs studies they call “daily information needs” studies - which take groups of volunteers and ask them to answer a few simple questions multiple times per day - the most pertinent of which is “what did you want to know recently?”. This is followed up with questions about the importance of the information need, and the degree to which the participant satisfied that need.

Doing this, they’ve identified over 25,000 needs across 21 broad categories - and this research has already led to the introduction of new Google Now “cards” including weather and transit information as queryless searches.

Of all the information needs participants identified, 39% went unmet - i.e. the participants didn’t or couldn’t find the information they were looking for.

Now - clearly many of these were just idle musings, unimportant things extracted from the participants’ minds by Google proactively asking the equivalent of “what are you thinking about?”. Luckily the methodology identified these - it turned out to be 15% of the unmet needs that went unmet because they were unimportant. The remaining 85% broke down as:

Why needs were unaddressed


Did not know how to find the answer


Insufficient time


No internet access


Other e.g. incapable of searching while driving etc


Now - remember that respondents reported 39% of their needs went unmet - when we remove the 15% of those that were unimportant, we are left with 33% (or a third) of important information needs going unmet:

What technologies could meet these remaining needs?

Ability to answer more questions

Technologies like personal search, data-driven search, natural language processing and conversational search enable computers to answer questions they were previously unable to answer. This will eventually mean there are fewer times when the respondent “did not know how to find the answer”.

Convenience and availability

The most obvious form of increasing convenience is voice search - opening up times when hands and / or eyes are occupied - e.g. while driving or cooking. I think that passive search could be a big deal here as well - with always-listening background devices like the Amazon Echo or Google Home. It’s amazing how much habits change with even small variations in convenience. For more context on this, I really enjoyed Dr. Pete’s recent post entitled How to Rank on Google Home.

Contextual search

Not only is Google applying more and more contextual information to search, but there are entirely contextual search capabilities being built into more and more devices - for example:

  • X-ray in Amazon Kindle that lets you search for a character in a book by clicking on their name

  • Force click in OS X allows you to look up a word

  • Long click in Chrome in Android looks up entities and words

Not only does this attack times when people don’t know how to find the answer, but it also massively increases convenience when you don’t have to switch contexts in order to find the information you wanted.

More pervasive connectivity and better offline functionality

Facebook and Google are both investing in expanding internet access to more people and more places. And they are both (FB, Google) investing in technologies to enable their products to work better in situations of low connectivity or even (in some cases) offline. While this hasn’t reached search yet, it may yet do so.

A lot of this comes together in intelligent personal assistants

This explains the land-grab that is going on as every major technology company attempts to outrun the others to develop truly intelligent personal assistants.

Even before you get to the developments in machine learning and natural language processing, Google’s ability (in particular) to put building blocks in place has been impressive:

  • They now have a system named Taba that can identify a new “context” (i.e. a set of user actions corresponding to an overarching information need) in real-time - within 10 minutes of the first action (a brief overview of Taba can also be found on slide 26 of my slide deck at the bottom of this post).

  • More than half of all search queries are now categorised as being part of long-running contexts lasting more than a month

In other words, Google might anticipate your searches weeks or months before you do them.

How far can queryless search go?

When Google Now first came out, I thought that the end-game was just a 5-10% slice of “extra” queries that could potentially be pushed to you based on your context without you performing an explicit search. I now think the opportunity could be many times bigger than that. I had thought that the majority of my searches were unpredictable and just products of my crazy imagination, but think about the predictive power of the long-running contexts I just mentioned and then consider:

  • The technology we have already seen introduced, including:

    • Now on tap (which searches for any entities currently on your screen)

    • Image recognition

    • The improvements in voice recognition we have seen from  “OK Google”

  • Now imagine how many of your searches could be predicted if Google had full access to:

    • Everything on your screen, and everything you typed

    • Everything you could see and everything around you

    • Everything you said and everything you heard

Most of those esoteric searches - that you might imagine were totally unpredictable - are actually sparked by something you read (or wrote) or something you talked about.

Now identify which of those passing thoughts are the important ones, and you get to another trillion searches / year pretty easily.

See you there.

If you would like to chat more about this, you should come along to our SearchLove conference in San Diego which is coming up in February 2017. This post is based on a presentation I gave at SearchLove in Boston earlier in the year:

Get blog posts via email

About the author
Will Critchlow

Will Critchlow

Will founded Distilled with Duncan in 2005. Since then, he has consulted with some of the world’s largest organisations and most famous websites, spoken at most major industry events and regularly appeared in local and national press. For the...   read more