Keyword research for SEO can be tough- don’t let anyone tell you different. Like most things though it can be broken down into manageable pieces, and with the right tools it can get you on the right track.

Historically I’ve worked with tools like KeywordDiscovery and Wordtracker, but the data they divulge isn’t what I’d call client friendly. For example, while KeywordDiscovery might show you it’s guestimate for search volumes I’ve found them to be sometimes unbelievable. As I show my research to clients it’s embarrassing if you get called on the validity of the data you’re working with. There is a good argument that as a comparative analysis it carries some weight, but if someone can spot potentially flawed data the rest of your findings will lose their credibility.

Ultimately there is no perfect way to discover a set of keywords for your site. One extreme train of thought is that the site itself should be the proving ground, and that it should be well written, have oodles of theme and topic relevant content. Over time, maybe 1-3 months your Analytics and Webmaster Tools data will show you what people are finding the site with. Once you’ve collated the best performing keywords you can then optimise the key pages and get going with link building.

I can see some merit in this approach, but it’s hardly what you’d call pro-active is it? When people want results it is going to be pretty hard justifying that kind of waiting game.

As soon as you can understand that focus keywords are not set in stone, and that they can be revised, tweaked, dropped, rationalised, expanded etc. then choosing starter ones is less fraught. My approach is to ask the client what they think people will use to get the site, as well as asking what they want the site to rank for. Sometimes the answers can be profoundly different, and this is often a good thing- it shows the importance of search intent. Another good question to ask is “where is the content that relates to these keywords?” If there is no content, no page, no directory, then it needs creating- otherwise trying to rank for it will be nigh on impossible.

The tools I’m using at the moment are all free and are within the reach and skill of pretty much everyone. There is one caveat though, that there is good historic Analytics keyword data. Armed with a suggested keyword list you can use Google’s AdWords Keyword Tool to score how much attention those keywords get in paid search. This is a useful indicator but shouldn’t be solely relied on- because there isn’t an equal spread between paid and organic searching. The Google heat map highlights this point nicely, with estimates that only around one quarter of searches are paid. You can also use Google’s Wonder Wheel to find suggestions, and don’t forget to ask people, people are often overlooked when you’re up to your eyes in online research- they’ll challenge assumptions, and give you some pragmatic ideas on search intent.

Once you’ve whittled your list down a little you can then try to find evidence of existing rankings for those keywords by delving into Google Webmaster Tools query data. You can then run it alongside the Google Analytics keyword data, to see which keywords already rank somewhere, and how many visits they might have been responsible for. This can be a pain, but if you know your way around Excel’s VLOOKUP it will save a heap of time.

Another element to factor in is the number of competing pages for searches for a given keyword- using the “ “ operator for each one. This brings back results based on closer relevance, so it’s a bit more accurate and useful. At the moment I haven’t found an automated way of getting this data, but hopefully Keyword Magic will provide the solution when they’ve finished their tool.

So, at the end of all this you can end up with a nice spreadsheet, showing the Local Paid Search Volume, the Global Monthly Search Volume, the number of visits in the last month, any current rankings, and the number of competing pages. Having this information in one place, in a congruent format will allow you start short listing. On it’s own it won’t give you the magic list, but by playing with different sort criteria you can begin to prioritise the likely best starters.