Insights Into Keyword Research Through My Failures

| July 25, 2012 | 16 Comments

Panda 3.9 just dropped.

Now that that’s over and done with, I wanted to take a look at the way I’ve been doing keyword research and analyzing past failures to see if I can improve my methodology. The way I’ve been doing it thus far is to use Ultimate Niche Finder’s keyword evaluation along with an analysis of statistic data from the top 10 in any given niche (page authority, domain authority, title competition, backlinks, etc.). In addition, I’ve also been taking a look more closely at link profiles and seeing how hard they are to replicate, as well as the types of sites that rank well in the top 10. What I mean by this is simply if the sites at the top are blogs, e-commerce sites, etc.

Typically I don’t dwell on past failures too much as I prefer to move on a build new sites. However, with some new keyword research methods from Hayden over at NohatSEO, I decided to compare how SEOmoz and UNF determine keyword difficulty using a select bunch of keywords:

  1. Keywords that I will putting on sites using a combined SEOmoz/UNF strategy.
  2. Old keywords that were used on sites where I failed to establish any good rankings.

My findings are summarized in this table:

Keyword SEOmoz UNF Niche #1 Site Status
KW #1 43 Moderate Health Forum New
KW #2 38 Very Easy Home Lead Gen New
KW #3 55 Moderate Business Authority Site New
KW #4 33 Easy Home Affiliate/CPA New
KW #5 51 Easy Education Authority Site New
KW #6 39 Very Easy Business Affiliate/CPA New
KW #7 55 Moderate Home Authority Site Failure
KW #8 61 Moderate Education Authority Site Failure
KW #9 49 Moderate Health Authority Site Failure
KW #10 53 Moderate Beauty Authority Site Failure
KW #11 55 Hard Technology News Failure
KW #12 45 Moderate Appliances E-Commerce Failure
KW #13 51 Very Easy Education Affiliate/CPA Failure
KW #14 46 Easy Automotive Web Magazine Failure
KW #15 53 Very Hard Beauty Authority Site Failure
KW #16 59 Moderate Self Help Authority Site Failure
KW #17 33 Very Easy Diet AdSense Failure
KW #18 41 Moderate Weddings E-Commerce Failure
KW #19 48 Hard Technology Authority Site Failure
KW #20 64 Moderate Education Blog Failure

*** Post-Panda 3.9 Rankings, July 25, 2012

Defining Easy and Hard Keywords

Sup? I'm back, did you miss me?

Sup? I’m back, did you miss me?

With Hayden’s method, an SEOmoz score of 43 or lower is roughly the tipping point where a keyword is “good enough to build a site for.” According to him it will generally take the least amount of links to rank well using this criteria. I haven’t been able to prove this yet as my private network is still in the process of construction.

With Ultimate Niche Finder, the program itself evaluates the keyword using amalgamated results from the top 10 that the keyword is found in. It then gives a graphical rating in addition to the difficulty on a scale that goes from:  Very Easy, Easy, Moderate, Hard, and lastly Very Hard.

Comparing the Values

If you look at the table above, you’ll find that SEOmoz and UNF are mostly in line with their evaluations. What’s interesting is where the numbers don’t jive. For instance, keyword #13 where UNF ranks the keyword as being Very Easy whereas SEOmoz assigns the keyword a difficulty rating of 51.

It’s difficult to make a blanket statement as to who is “correct” in their difficulty assessment of the keyword. This analysis is literally the day after the Panda 3.9 release, which skews the results somewhat. If you take a look at keywords #1-6, you’ll see that I am potentially screwed on a couple keywords there where the SEOmoz difficulty is above 43. They were below that number when I did my initial research. It’ll be interesting to see what happens to those sites when I build them out. The content has already been placed on those sites, so I’m past the point of being able to put the brakes on and re-analyze the niche.

Analyzing Failure – What the Hell Went Wrong?

Quite frankly I cannot explain what happened to keyword #17. Both SEOmoz and UNF seem to agree that the niche is very easy to rank in. Also, this is the only niche from my sample set that has an AdSense-related site at the #1 position currently. It’s basically a niche where you get points just for showing up, so to speak. Anyway, I’ve gone over this one quite a few times and can’t put my finger on it, so I’m just chalking it up to random failure on my part. I may try again with a new site and new content just to see what happens.

For the other failures (keywords #7-20), I believe what happened is likely caused by a few factors:

  • Underestimating keyword difficulty. If you recall I didn’t always have UNF and was using SECockpit before. It’s very possible that I misread the data at the time.
  • Being the odd man out. This is one of my new theories. Basically the idea is that I won’t have any success trying to rank a blog in a sea of e-commerce sites, for example. I believe that you have to somewhat match what you can find (and should find) in the top 10, otherwise you’ll have trouble with ranking during the next merry-go-round shuffle.
  • Being greedy. Several of these failures are exact match domains. I chalk up the failure to seeing the domain available, snatching it, and trying to rank the primary keyword. This is why I am not a fan of EMDs per se, since I have been successful without them for the most part. The moral of the story though is not to be greedy; sometimes is better to pass on the EMD as it may not give you the rank boost that you’re looking for.

How I Define An Authority Site

On a side note, you might observe that authority sites are featured prominently in the table above. Firstly, let me cut you off at the pass and define authority sites as those who:

  • Have had their domains aged for quite some time.
  • Have a large search and social following.
  • Are generally large and provide great value to the reader in terms of content.
  • Have many ways to interact with the visitor.
  • Are generally not monetized through AdSense or affiliate offers.

Authority sites are NOT:

  • Made-for-AdSense/Amazon niche sites with 50-100 pages of content. The more the better.

I see this being touted around under the “authority site” moniker on many IM guru-style blogs. Sorry, but simply having a lot of content doesn’t make you an authority. It ain’t the years, nor the mileage. It’s the years AND the mileage. 6-month old sites with tons of content is all well and good, maybe even preferred. But that doesn’t make you an authority. You’ve got to earn it.

Rant over.

Where to Go From Here

So what types of sites should you build? What types of keywords should you go for? Do the niches matter? Does the site type matter?

I think that the true answer is probably one that you don’t want to hear:  It depends. I think if I continually analyze my past failures, I will end up finding patterns where there should be none, and no patterns where there should be some.  In reality I would probably wasting too much time.  So my direction is just to keep doing more keyword research and building more sites rather than to dwell on unsuccessful ventures.  You get better over time so more experience helps.

I will probably avoid the niches where I have had minimal success, like the technology niche. Instead, I’ll concentrate more on niches where I have found repeatable success. In the past I preferred to do everything in broad strokes, but I suppose that it might be more beneficial to specialize in at lost some small degree or so.

Neither keyword difficulty assessment tool, SEOmoz or Ultimate Niche Finder, will be perfect.  Proper keyword research requires a deeper understanding beyond what mere tools can tell you.  To a certain extent, I think it requires a lot of confidence and a lot of faith in your gut feeling.  Some things that you can’t really quantify (or even qualify) unless you’ve done a lot of keyword research and have gone through hundreds or even thousands of keyword evaluations. The tools do give you a good idea as to what you’re looking at, but ultimately the go or no-go decision is yours.

Andre

If you enjoyed this post, please consider leaving a comment or subscribing to the RSS feed to have future articles delivered to your feed reader.

Tags:

Category: Case Studies

Comments (16)

Trackback URL | Comments RSS Feed

  1. Matt Hagens says:

    Well said. I use a combo of SEOmoz and UNF to identify KW’s. Probably very similar to the way you are doing it. But above all else, a green or red light on a keyword from a tool does not mean you should build out a site for it. You need to manually check every keyword and not completely rely on tools. That ability only come from experience and many, many mistakes 🙂

    • Marc says:

      You’re right that you have to manually check a target keyword. I use Market Samurai(SEO competition only) and SEOmoz. What SEOmoz tells me is the kw difficulty. Then I go to MS to check top 10 backlinks to page, PR, and their on page optimization.

      I only for now go after kw that are less than 43 in SEOmoz then only after it jives with MS then it’s a go. As Andre said if top sites are Authority sites like Amazon etc although they meet the criteria I’ll pass. I’m very nit picky.

      I’m not building Adsense sites at the moment and not in a huge rush to build a network of sites. My goal is to build site a month based on Amazon affiliate. That’s 12 sites a year. That’s on the low end $700/month passive income. After 2 years that $1400/month passive income. I got the rest of my life to build my business and I like the slow yet consistent approach.

      I look for niche’s that in theory don’t need backlinks or a very light backlink campaign. When doing research I don’t necessarily put those kw that don’t in theory meet the above and throw them away. I put them in an excel sheet and/they are save in Market Samurai or SEOmoz for the future.

      Reason I do this is because I’m cautious about back linking methods. Until I find the right method that I’m comfortable with then I’ll continue to do this and stock pile kw that are more difficult.

      I like Hayden’s method of your own network and still contemplating on it.

  2. Marc says:

    Hey Andre where I have success is what Spencer has said regarding analyzing the competition. Competition research is part of keyword research. What I’ve been doing is finding keywords that meet Spencer’s criteria or as close as possible. It’s much harder finding niche’s like this but it’s worth it.

    Target keywords where not all the results are using the exact keyword in the title.
    At least 2 of the top 10 results have less than 10 links to the resulting page (0 links preferred)
    At least 2 of the results with less than 10 backlinks also have a PageRank of 0.  Meaning 2 results need to have a PR of zero AND have less than 10 links.
    Target results with weak types of sites (article directories, forums, etc.).
    Target results where ranking pages are NOT root domains.

    I look at the top 3 in this case and if the top 3 don’t meet the criteria then ‘next’. I’ve learned to be very nit picky.

    • Andre Garde says:

      Hi Marc,

      This is actually the exact method that I learned when I first started out late last year. Over time I modified my process and refined it. It now includes more criteria as I mentioned above. Spencer’s strategy is good but it is weak against sites with high PA/DA. A lot of high value authority sites can still outrank you even if their PR is 0. It just goes to show how meaningless it is as a measuring stick.

      • Marc says:

        Of course you have to look at whether or not serps are authority sites or not.

        With SEO moz you’re just going by keyword difficulty?

        • Andre Garde says:

          Not quite. I take a look at the link profile to see if it’s something that I can easily replicate. Also I use their onpage tool. Sometimes they miss the meta description, which I think is important. That makes them a little easier to out-do.

          • Marc says:

            Right as Matt said and I agree you have to manually check the SERP’s.

            Keyword research is one thing but not enough people emphasize competition research. Both go hand in hand but lots of times competition research is more if a passing note.

            Actually there are schools of thought of beating an authority site where the PAGE has PR 0 and not many back links and not optimized for the target keyword.

            There are many searches where a PAGE to Amazon has a PR 0 practically no backlinks and are not optimized.

            And there are many searches where a PAGE to Amazon has a PR 3 and has many backlinks and somewhat optimized.

            It depends on the target keyword.

            You can find many SERP’s where a site beats Amazon and isn’t an authority like Amazon.

  3. Marc says:

    What about #’s 6, 17 and 18? How were they failures? #6 is new.

    • Andre Garde says:

      I touched on #17 above and I’m not sure what happened. I think I might just rebuild this site.

      #6 is a new site and I haven’t started linking to it yet. It’s built though.

      #18 I think I underestimated the true difficulty.

  4. Marc says:

    “With Hayden’s method, an SEOmoz score of 43 or lower is roughly the tipping point where a keyword is “good enough to build a site for.” According to him it will generally take the least amount of links to rank well using this criteria. I haven’t been able to prove this yet as my private network is still in the process of construction”

    That’s not necessarily true. You have to check APA, I pointed this out to Hayden and he agreed. You can’t assume just because a kw difficulty is under 43 you’re set UNTIL you get APA of SERP’s.

    • Andre Garde says:

      I don’t bother calculating APA because I’m still in manual mode yet. But here’s KW #18 for results #1-10 in the SERPs. One small caveat: the difficulty changed from when I posted this blog post. It’s now 44 (not 41):

      41
      34.69
      36.89
      43
      43
      35
      24.29
      27.48
      40.31
      51

      This should have been easily within my reach. When I look across my portfolio I’ve ranked profitable sites with APAs more difficult than that. I can’t take any metric as black and white, since it’s too open to interpretation. I’d spend more time analyzing sites than actually building. It’s been far more profitable to just do the latter.

      • Marc says:

        I agree you can’t take metric as black and white. As I said keyword/competition research is the most important element.

        That’s why you have to get the APA even thigh kw difficulty would show under 43 because once you get the APA then you get the true result/s. I’m in manual mode as well as it doesn’t take long at all.

        Looks like most of your sites were in more competive markets.

        I bet #2 and 4 you’ll rank. Why? Because you’re competing with an affiliate and lead generation. That’s if you can work it better than them.

        A school of thought is to seek niche’s where there’s affiliate competition. A friend of mine won’t go after a niche UNLESS there’s at least 3 other affiliates in the SERP’s.

        • Andre Garde says:

          APA is basically putting a number into understanding domain/title competition. I suppose that I should do the calculations more, but honestly I’ve compared against the niches where I’ve done well and I have been every bit successful just by eyeballing it.

          I suppose that when you’re at Hayden’s size in terms of portfolio, APA is more meaningful as there’s no way you can spend this much time doing manual analysis as it’s programmatically derived.

          Funny that you mention not going into niches where there isn’t affiliate competition. Sometimes there’s no choice (payday loans, diet, finance, insurance, etc.). Other niches are quite dumb; #17 for example. The top 10 are MFA/CPA, and 11th and 12th spot are Facebook and Amazon. Quite possibly the dumbest niche I have ever seen, but then again, it is in the weight loss vertical…

  5. Andre Garde says:

    No idea. Must be linkbuilding or site design. It was one of my first sites and it has multiple pages targeting the same keyword. /shrug

Leave a Reply to Marc Cancel reply

Your email address will not be published. Required fields are marked *