Grow your business by advertising on bing, MSN, Xbox, and more!
Funny how the world revolves around keywords and key phrases these days and yet no one understands them!
So what's to understand? First off is getting to grips with the intent of the person behind the keyword or phrase. Search engines have been struggling with this for decades and very little progress has been made. The attempt to understand intent has led to personalisation, search engine logins, search histories and cookies the size of saucepans. Still the user's intent remains elusive, not just to those of us not benefitting from the data search engines are collecting - but also to the search engines themselves .
Why?
Think of the keyword as an abbreviated thought - a thought which has been abbreviated by the user. The term 'keyword' tends to suggest that, because it is 'key' it must be commonly used and we must know what is behind it. But we don't. Nor is the searcher aware of all the wide ranging potential ambiguities which make it tough for the search engine to respond to their search with a relevant list of potential answers to an implied question.
Part of the problem lies in the training that searchers receive from search engines as a result of their search experience. Every searcher's queries change and evolve as a result of the feedback 'food' they receive from search engines who are training people to search in particular ways. Unfortunately that confuses matters further because that training tends to focus the user more on certain keyword structures irregardless of their intent.
But there is a way around this problem that is not new - in fact it has existed since before search marketing, before the internet and even before computers existed - and that is the mathematical approach of probability.
We cannot change the impact of the varied intent of the user confusing the picture. But for each web marketer, every keyword has a certain probability of leading to some sort of action or conversion. It is of course different from product to product and website to website.
So how can you deploy this? You probably already do you just don't consider it as a probability but rather as a percentage (which is, after all, a measurement of probability). In essence, you need to look at things differently.
For instance, let's just consider the concept of 'share of search' or 'share of voice'. Neither of these measurements is valid because that would mean every searcher had exactly the same intent all the time. Yet we know that searchers change their intent - even using the same phrase, over time.
To try and win every 'click' resulting from a search would be a misguided goal - meaning that almost certainly you were generating visits of no value to your business. This is why 'share of voice' or 'share of search' are unwise bedfellows to your campaign.
To be fair, I also have problems with 'click through rate' as a quality measure. If 2% of a particular keyword search are interested in your company, why would you want to increase that rate and pay for people who are not interested and will not convert.
When we select our keywords for our campaign we need to include some form of probability of them converting. WebCertain is introducing a probability index or PI into its keyword research in multiple languages. We're looking at various bits of data - which very significantly by language, to create this index which is intended as a guide to help in keyword selection.
Demographics can also help. If you know the searchers that convert are generally women - then a keyword used by a women and targeted with a relevant sponsored link will have a higher probability of conversion than terms targeted across both men and women searchers.
But as a marketer, the key step is to build your own probability model based on the data you have on your site and use this to help select your keywords for the future.
Andy
Andy is MD of multi-lingual Search Agency - WebCertain
Sign in to adCenter | Need an account? Sign up now
Follow us on Twitter @adCenter & @MSAdvertising | Find us on Facebook and YouTube | Share your thoughts and ask questions in the Forums | Subscribe to the adCenter Blog
You've whetted my appetite there Andy, but what can you reveal further about how probability is measured.
Is it for existing account - based on known conversion to different actions like conversion Goals in Google Analytics or can it be applied to a new accounts before conversion data is available?
Or is it a softer measure such as "Intent index" 0-10 gut feel on likelihood to convert?
Dave
I think the engines are making reasonable strides in this area and I see them doing this probability calculation for you, mostly. Bing in particular (and in the US only in particular) is making great strides in working out the balance of probability in terms of user intent and taking the user further down their most likely search path than ever before. But the other search engines - as well - have started using "personalized search" to help those search results get influenced, with the probability of user intent changing depending on new variables. So you ARE entitled to use click through rate as a KPI in my opinion. What you shouldn't use is "clicks" as a metric I guess... because your KPI is diverging (as opposed to convergent) with the KPI of a good search experience.
In response to Dave, it's not a 'soft' measure as in we anticipate from "experience" that these terms would convert - but it is following the various patterns which are associated with terms which typcically convert. So it it is to guide the selection of terms before conversion data is available. The index is assumed to be 100 for terms which convert - not on every occasion - but some of the time. (100% conversion would be nice - but not realistic!). The important thing for us is that it has to be applicable in any language.
Dixon - don't agree with you on click through rate. What happens if you put your telephone number in the creative for instance? Can be good for conversion but reduces click through rate. I can also quote an example where the click through rate of a particular campaign is 15% - but it almost never converts, very very low percentage of conversions, and in the same campaign a term with a ridiculously low click through rate of 0.11% - almost a 100% conversion of the subsequent clicks. In this latter case, the creative is the best performing one of 10 - so it's not that the creative is poor. The reason is that not all of the users of the term have the 'intention' to buy that particular service - but clearly the ones who do have a have been 'filtered' by the creative being very specific preventing wasted and expensive clicks.
Frankly, I think margin a better measure of success than conversion rate. Margin is to conversion what conversion is to CTR; you need your prospect to take one step after another but the measure of success is profitability. Obviously, not all sales are equally profitable.
And I agree it would be preferable to qualify your prospects based upon probability of success (preferably, the probability of higher margin transactions), it's the "various bits of data" used to qualify your prospect that are the critical, missing piece. I'd be interested in hearing more about that.
Are you talking about probability or significant confidence?
For example, "Dear client, we are 95>% confident that searches of these keywords are interested in your brand/products before going live"?
Would be good if this can be determined pre-campaign to a level of stat confidence.
Very thought provoking Andy. Thanks for this.
What struck a chord was the suggestion:
"We cannot change the impact of the varied intent of the user confusing the picture. But for each web marketer, every keyword has a certain probability of leading to some sort of action or conversion. It is of course different from product to product and website to website."
...having worked for a number of sites in the same sector, I would never recycle the same keyword research...and nor would my conversion assumptions transcend from one site to another. I'd love to hear more about how you're going to develop this Probability Index, but I'm sure we'll hear about it more in your newsletter!
Cheers Andy,
Ben
I like this idea of probability for the searchers intent. It would greatly improve the comparison of impressions, CTR, and conversions in a search campaign - if the keyword had a low probability of intent then I would expect low CTR. All in all it would be very useful metric to consider.
Also, I hate to nitpick but 'irregardless' isn't a word.
This is the kind of subject that really matters these days. Keyword selection is a highly refined process. It's been long understood that for a new site breaking into an established market, you shouldn't just go for the phrases that have been searched the most. Yet there's very little out there on how to refine the selection after that. High volume search, low advertiser competition those are all good and fine.
Yet what you're talking about in the article is really more about grinding down into the mind-set of the most likely prospective buyer, and that's not something that is so easily boiled down into an algorithm or formula. It takes the ability to really get into that mind-set.
For an SEO who has a plethora of clients across a great number of markets, it's even more challenging because we have that many more markets we need to understand, live and breathe.
This is one of the most enjoyable aspects of the work I do. It's as challenging as SEO work can get.
Synonyms of key words is often the extent a person has to go with both the single and plural tense key words all embedded into unique, well written, copyright. There's a reason why it's considered both an art, and a science. The one thing one can appreciate from all this is that it's always evolving, and as a result of this so does the content improve as well all surge towards perfection.
I recently contacted Google about the problem of most learning stopping at the border due to the geo-ip-locating of hosting, and the limitations of their webmaster tools interface. We should not have to build redundant sites just to rank in various regions.. It's definitely time for change.. we can only hope the search engines are listening.
I think you're right in saying that search engines are teaching users to search in certain ways, particularly with their suggest facilities. One example is when I was doing some keyword research for a client recently, it became apparent that there were double the searches for the singular version of a phrase, as a pose to the plural, even though when written singularly it made very little sense. This was definitely not something that I feel a user what type in, but as soon as you started typing anything similar, it was the first item on the suggest list! Interesting stuff...