Something caught my eye today – along with the practical implications of it – an article in Search Engine Land on Paid Search without Keywords says machine based learning algorithms can replace keywords for Paid Search Advertising.
…. keywords were used as a proxy for relevance. Conceptually, there is no reason an advertiser couldn’t achieve the same results without having to directly manage a keyword list. Down the road, Google wants to state outcomes and have machine-based learning and algorithms come up with the best method of achieving specific outcomes. In the case of no keyword search, an advertiser (like a retailer) would provide information on products, product descriptions, pricing, etc. and Google would use the information to find the most effective way to place ads in front of potential customers.
In a way, it makes sense – though I wonder, just what the role of a Search Marketer is going to be if they don’t come up with keyword lists anymore?
It’s almost as if Google is saying – just give us the information and we’ll optimize everything for you – sure, if your a retailer, like The Gap, you’ll need to specify all the features of your shirts, pants, bras, belts, etc – and that could be a job, in and of itself – but I’m guessing by that time, Google will just ask for an XML feed with a list of “attributes” against a set of “products” or “services” and that’s it – your done – no keyword list – Google will do it all for you.
My feeling, gut wise, this will happen, in about 18 months to 2 years – as a trial, and within 5 years – keyword lists will be entirely eliminated. According to the article on Paid Search without Keywords
….. there are several advantages of no keyword term search such as …
- Efficiency for advertisers—there would be no keyword research component to a PPC advertising campaign.
- There could be better connections between searchers and advertisers on natural language queries.
- It would allow advertisers to better connect with consumers and capitalize on all relevant advertising opportunities.
These are obviously initial ideas and now Google needs to figure how to make no-keyword search work. According to Nick Fox, it will be some time before Google shares specific product details. Sharing the concepts with the advertising community at this stage doubles as a feedback mechanism and a trial balloon: if too many people hate it, Google can modify its approach. We sure look forward to hearing more on this.
Taking a forward looking view, maybe getting rid of “keywords” is good idea – as people often use the same terms to mean different things – even in subtle ways – and let’s face it, people usually don’t say what they actually mean, nor do they write what they actually mean, either, for a variety of reasons.
For one thing, Men and Woman process information differently, and different parts of a human brain process different words while different languages put different significances to the same terms
…. In English, months of the year are treated as containers. People say “in January” or “in February.” Other languages treat months as surfaces. For example, “on January” or “on February.”
Considering that many people’s native language is not English, the brain structure may well have adapted some what differently – overlaying different languages, as we learn them, doesn’t necessarily make the same exact mappings – and everyone has their own way of processing information. When you add all of that together – it’s no wonder the list of keywords for a typical campaign is getting bigger and bigger – and harder to manage – more people are using search – and the ability of keywords to describe intent is actually diminishing (Search Engine Land cites “increased Searcher Sophistication” – but as things move faster and faster, with more people searching and exchanging information, the number of words needed to map concepts to objects is multiplying faster than our ability to describe what we mean - which we don’t know, half the time).
… Increased searcher sophistication. In his keynote, Nick provided an example related to cashmere sweaters that demonstrated this point well. He stated in 2007, people searched for cashmere sweaters 47 different ways. In 2008, people searched for the same keyword phrase 73 different ways. It’s becoming a headache for advertisers to anticipate and react to so many different queries. Should advertiser prowess be measured based on mind-boggling attention to long query keyword detail?
Another point, that wasn’t brought up in Search Engine Land – is Keyword Search is already almost useless for Social Media – no keyword tool that currently exists for Search Engine Optimization or Marketing is of much help for doing Social Media Research. Why? Social Media is about conversations, not keywords – often the words around a key-phrase are as important, or more important, than the key-phrase, itself. I wrote a post about it a while back where I postulated that ….
- Keyword Research Tools like WordTracker and GoogleAdWords Keyword Research Tools (there are a couple, now), along with all the rest out there, are unsuitable for Social Media Research because they don’t capture conversations.
- Existing Keyword Research Tools are built in order to find the most searched for or niche phrases in order to drive visits to a site; Social Media is not about visits to your site, it’s about Conversations and Relationships which may result in visits to your site, or may not – but do provide value, and may lead to conversions, but not necessarily right away.
I used Radian6, I think in a rather crude way, to show that using a keyword list generated to find information in a search engine isn’t going to help me find influencers - the purpose of the current crop of keyword tools was more to identify objects – or to sell things.
In bringing up the limitations of keyword research for Social Media, the same machine based learning algorithms that can replace keywords for Paid Search Advertising can also do it for Social Media, and, in fact, are doing so right now, to some extent, with Crimson Hexagon.
The recent Disney-Marvel merger, written about by Mashable brings up the variety of “opinions” that people have about the merger, highlighted by Crimson Hexagon’s Opinion Monitor, a very select service Crimson Hexagon offers. According to the Mashable post ….
.. The Twitter analysis was drawn from a sample size of over 14,500 tweets, and was able to go beyond the high level summary to break down exactly what Twitterers were saying — positive or negative — about the deal. Bottom line, you were conflicted about how much creative control Disney would have over Marvel characters and story lines.
Since I’ve been consulting at Porter Novelli recently, I’ve had a chance to speak with Melyssa Plunkett Gomez at Crimson Hexagon last week – and found out the details how how their machine learning algorithms bypass traditional search. Melyssa pointed out there are three patents to Crimson Hexagon, a company that emerged from a Harvard Think Tank. While I love Radian6 and Alterian/Techrigy/SM2, they, and most of the rest of the Social Media listening platforms are still working based on keywords – though I’ve had conversations with Marcel Lebrun about moving past that with Radian6.
1. The first Algorithm, patented, is able to determine the main meanings in a large amount of unstructured information (ie: that pie chart, above) with a 3% error rate.
2. The second Algorithm, patented, is able to match the unstructured content with the pie chart with a 20%-25% error rate.
3. The third Algorithm, patented, is for ranking of the content for relevance.
But… get this, Crimson Hexagon doesn’t use Keywords. That’s right. I don’t think Crimson Hexagon is “keyword blind” …. it does consider keywords – but it’s not the basis of that pie chart, above, and there’s a few white papers on the concepts behind how they do it on their site.
So, to sum up this post – Keyword Search for Paid Advertising may be an endangered species in a few years, and while the methods Google uses to figure out what you want and serve it to you may not be the exact same one’s that Crimson Hexagon uses to figure out what people mean …. “the Keyword Phrase” the basis of SEO and SEM ….. is fading.
I’m still refining my thinking on this subject, it continues to evolve – and I’ll have more to say about Crimson Hexagon once I work directly with the Opinion Monitors – hopefully, in a few months.
.. And I only intented to write a short post … and here I go again, two hours later – I can’t help these “insights” … they just “flow into me” and once I see the connections – I need to go with it – and write down what see (I’m still an artist, after all – I see and think in “patterns”).
Now, it’s time to stop, go to bed.