Google turns 15 this week. Which makes it pubescent, at best. That may explain their recent decision to take away our keyword data – unless we pay for Adwords, in which case, we’re more than welcome to see what keywords people are typing in.

The fact that I concluded that sentence with ‘typing in’ indicates that I am perhaps slightly behind the curve here. Google foresees a future where typing in search queries is going to fall behind ‘speaking’ search queries. And that requires a fresh approach.

This, I believe, is why they have unveiled Google Hummingbird, an algorithm update that rivals Caffeine for its (alleged) vastness. It will affect around 90% of search queries, and Hummingbird aims to do the following:

  • better understand co-occurrence (i.e. the match between words commonly found together, yet not always technically hyperlinked to a website)
  • better understand synonyms (e.g. restaurant / place), especially when spoken
  • respond better to questions and queries
  • provide comparison data (e.g. compare chocolate with beef)

Search Engine Land provided a good analogy. Hummingbird is a new “engine”. Panda and Penguin were just parts of the car – patches, if you want, that are applied to how the engine works. Hummingbird is more than that. It’s all about upgrading Google for modern technologies, and the way those modern technologies are modifying the way we search.

Getting Into Our Brains

We’re used to thinking of SEO as being all about keywords. Because Google has always been all about keywords. The reason we often get frustrated with search results is because Google has not fully understood the way we speak – we have had to adapt the way we search to Google. We are used to refining and refining our search to get the results we want.

Equally, Google’s model has always been about picking out the ‘obvious’ signs. For example, it is heavily reliant on page titles and headers, and while it has moved away from its reliance on keywords in anchor text, it’s still a significant sign.

Co-occurrence means that Google can now better understand the context between brands and words that are not always technically hyperlinked.

For instance, if you own a seafood restaurant and you’re famous for your lobster – Google would be able to pick out the co-occurrence between the name of your restaurant and the constant (un-linked) mentions to lobster. It should, therefore, be able to raise your profile for searches that involve the word lobster.


If You are Speaking, You are not Using Keywords

If Google is going to really do voice-search, then it’s going to have to recognise natural speech patterns, and synonyms. Bill Slawski’s example of the way restaurant and place are synonyms in one sense is a good one – to understand that they’re synonyms, Google has to understand the rest of the sentence. This is complex stuff.

Hummingbird is all about getting to grip with complex queries, and quickly providing the right information that matches the voice searcher’s intent.

And maybe, then, all this fuss about (not provided), and the outright theft of our keyword data (unless we pay for Adwords) does not really matter. The landscape is shifting and very soon, it won’t be about keywords, it will be about intent. Everything will be long tail.

And maybe, just maybe, this is a good thing. Instead of focusing on what people type (e.g. creating page titles that say ‘divorce lawyers dallas, texas) – we can focus on being more natural. We can stop worrying about getting exact-match keywords on each page, and we can relax a little, safe in the knowledge that Google is going to match up the right people with the right ‘entities’ – and that’s not always going to be a web page as we know it.

So What Does This Mean for us?

I hate to be inconclusive – but who knows?

Google Hummingbird has been around for a month, and nobody seems to have noticed. Search rankings, at least as far as I have seen them, haven’t fluctuated massively, and with the loss of keyword data, it is hard to notice trends in the long tail.

However, there is one chink of light – a specific client for whom we have been building links with co-occurrence (i.e. brand name is hyperlinked, but surrounded by money keywords), has experienced a nice uplift in searches for those terms that co-occurred with the brand.

Is this Hummingbird in action? Does this mean, conclusively, that co-occurrence is A. Real. Thing and that we should do more of it? Possibly.

However, in a world of increasing data loss, and fragmented search possibilities, we now have to review how we collate, analyse and interpret our existing data sets. Hummingbird reflects the way people now search, and will be searching for the next five to ten years – our challenge is to move away from our old ‘keyword’-based mindset, and find a new one.