• http://www.rimmkaufman.com George Michie

    Hi Matt,

    I’m going to respectfully disagree with you, here. No question the return on effort of managing the long tail has to make sense, and we have to factor in the opportunity cost associated with tail management potentially taking away time from other higher value activities. I’d make the case that the cost of that management can be greatly reduced with an advanced suite of tools, but I grant that that isn’t available to all.

    Where I disagree is with the methodology. Bucketing by conversions, rather than click volume, creates a self-referencing loop. It is similar to a catalog company only mailing catalogs to the 2% of people who bought from the last catalog. To use another metaphor, it’s like choosing which parts of your fishing net have worked in the past and cutting away those parts that “don’t work” to reduce the drag.

    I did some research years ago showing that for one particular client something like 80% of the long tail KW that got orders in October of some year, had not generated an order the previous October.

    Smart bid management and smart time-management are required to make the long-tail strategy effective. Totally agree with you that if it’s a huge time sink for you, it may not make sense to pursue that strategy in every case.

  • http://www.findmefaster.com Matt Van Wagner

    Hey George,

    As always, your considered comments are welcome and enlightening, thank you!

    The problem for smaller budget, smaller scale campaigns, is that that no tool can automate the low volume sets – they will always remain a fog – because of the sparse data. My example was meant only to demonstrate that the cost of the long tail is a directly observable overhead that needs to be accounted for – as well as the indirect and less explicit value of avoiding other optimization tasks. Non-productive long tail ad spend is very easy to overlook in reporting and analysis.

    I agree with you that bucketing by clicks is good approach, and thank you making that point. I should have fleshed this topic out a bit more. The clicks data, when combined with conversion data, will tell you where hot and cold spots are in the productive part of the long-tail.

  • N.B.

    Some of the highest ROIs we’ve seen have been on Long Tail Keywords that only have had one or two conversions. This is because the product being sold is very high-ticket compared to the cost of the keyword. So I agree what you have to tightly group your long tails into similar themes, brands, product niches and then make the decision based on the overall group performance when lacking enough data. The tools to do that are already in Adwords and it is not a very time consuming exercise.

    Here’s a couple related questions I would be interested in your take on.

    How much data do you deem significant to make a decision on any keyword, ad or ad group as far as conversions? 50 click-thrus? 100 click-thrus? 1000 click-thrus?

    Also, do you advocate choosing your long tail keywords by hand picking them using the keyword tool, or have you used precise broad match keyword at first (using + sign) and then looked at the matches and used a combination of moving the matches with conversions and with no conversions but enough click data into exact match keywords and negative kewords. The former can be very time consuming for larger campaigns and you will miss many long tail keywords as the keyword tool won’t find 100% of the long tails. (Same thing the first commenter was saying about there being keywords you don’t even know about yet that may produce a sale.) Whereas the latter you can waste money discovering keywords, a lot of it if your don’t have a good negative keyword list to start. We’ve got sales though from broad match on very very long tail keywords that we never would have gotten without using the broad match. It really boils down to CPC for the keyword(s) in question. Some campaigns we can afford to do broad match, others, it is too pricey as too much competition even if you have the most expensive exact match keywords in your negative list. Did I just answer my own question?

  • http://www.eBizROI.com Rick

    Great post Matt with some good PPC Management tips and techniques. The one thing I can say about the long tail clicks are cheaper (excluding management overhead) and tend to convert better as generally speaking, the more specific the search query, the further in the purchase funnel the search like is. As you point out, there is a lot of dead wood and not enough data to make statistically information performance decisions.

  • Benjamin Krauss

    How do you delete that mass of longtail-deadwood?
    In your example you have 3557 KW, ready to be deleted – but adwords only gives you the opportunity to delete 100 by one click.
    so – what to do?

  • http://korcule.com Jaromir Scepka

    Hi Benjamin,
    I guess you should try adwords editor.

  • http://www.findmefaster.com Matt Van Wagner

    Benjamin – Jaromir is right (thanks, Jaromir!) Use AdWords Editor. It gives you the ability to filter and select according to various criteria.

    I’d also recommend taking a good look through the triple 0 keywords (that is kws with 0 conv, clicks or impressions) before you begin deleting and see if there is a setting, or negative keyword, or other technical glitch that is preventing impressions – especially when you believe the keyword is a good one that should be getting impressions.

    To N.B. and Rick – I’ll reply to your posts later today. Great questions, not quickly answered- they may actually require an entire post on their own.

  • http://www.ivantemelkov.com Ivan Temelkov


    If this was several years ago I would have disagreed with you respectfully. Although it wasn’t until I encountered and had the opportunity of managing a multi-million dollar budget via AdWords that I found out you’re quite right about your statement. Based on best practices suggested by Google its strongly recommended that you build out lengthy keyword lists within relative adgroups. Long-tailed keyword terms that is and then moving onto the broader side if needed to capture additional traffic. Although as I mentioned above it was quite surprising to me that once I condensed the number of keyword terms inside my adgroups that the ROI increased.

    There was simply more traffic flowing in and individual keyword terms began to perform better. For the most part I agree with the information provided although in some instances you may be forced to build out lengthier lists.

    Thanks for the great information!

  • http://www.findmefaster.com Matt Van Wagner

    Ivan, thank you for that feedback.

    The environment we deal is complex and the interactions between elements in our accounts and the ad serving algorithms are beyond our ability to directly observe and understand them.
    A single bid change to one keyword is capable of dramatically changing which other keywords flow into the auction, too, and impact ROI.

    Automation can help you manage the long tail. So can simplicity in account design, as you point out.

  • Bradley Zeller

    I think that this was a great article when you generalized stand point. I agree that it is hard to figure out the value and worth of long term keywords, its especially hard when you go after keywords that have an average cost per click of $50. I can never gather enough clicks on long term keywords that cost that much. I just use short term keywords and have ad groups broken out with 3-5 keywords by themes.

  • http://mklofurno mklofurno

    Hi Matt – Nice article. Thank you. Good points.

    The only thing I have to add is that what hit me from your article and the comments is whether you are actually using a statistical significance calculator for your data. I think its too easy to trust our “eyeballing” of the data.

    That is my two cents.

    Mary Kay Lofurno

  • http://www.findmefaster.com Matt Van Wagner

    Hi Mary Kay,

    Thank you Yes, you make an excellent point. The example I used to illustrate above could be easily eye-balled based on the size of the sampling, and difference in perforamce so I didn’t add any stats to support it my example.

    Eye-balling is especially tempting for small data sets, when you really want to make decisions about ads. It’s also tempting when you have enough data, but too close a result to authortatively choose the higher nomimal results. I’ve done it, and I suppose we all have – but your point is well taken.

  • http://www.findmefaster.com Matt Van Wagner

    Thank you for your observations, Bradley. Solving the tough problems is what keeps us all employed, I guess. The machines can deal with the easy problems, and us humans have to figure out how to deal with the tough ones….

  • http://www.linkedin.com/in/jasmineaye Jasmine Aye

    Hi Matt,

    Thanks for writing on this. I agree with another comment: What type of time frame are you looking at to gauge whether these keywords are converting or not? Are you associating this information to an ecommerce client or a lead gen client? Ecommerce clients *typically* have a seasonal range to them and traffic can rise or fall depending on the time period, which I’m sure you know, but I was just curious.

    While I was more than happy to read your article and hear a new point of view about how to make accounts simpler, I did not appreciate the fact that you changed the meaning of “long tail” to mean “anything that doesn’t convert more than 10 times in X amount of time.” It completely confused me as to what we were discussing.

    I was a little disappointed that long tail keywords, by either definition, weren’t noted on a QS level and how keeping the good ones as part of the account would actually increase your account QS and allow for lower cost bidding. In addition to the QS metric, I would have liked to hear more about what match type your high converting terms ended up being, as match type normally has something to do with conversion rates.

  • http://www.findmefaster.com Matt Van Wagner

    Hi Jasmin,
    Thanks for reading so closely and for commenting.

    The time frame you choose for your analysis is really up to you and your click volume. This analysis is essentially the same for any type of account, lead gen, e-ecommerce, etc.

    Your point about using conversions in the analysis is similar to the point George Michie wrote, and I agree that I could have made this point better.

    Regarding Quality Score, you raise an interesting point. It is an unfortunate fact of life that AdWords QS does influence the effective CPC prices you will pay, but I can’t imagine that keeping non-productive keywords around just for a QS bump is worth considering.

    In the interest of keeping the column reasonable sized, I didn’t talk about match-types, but you are right. match-types make managing and analyzing the long tail even more difficult. In the case of the higher numbers of conversions – exact and phrase – as you’d expect, accounted for most of the conversions.

  • http://mklofurno mklofurno

    That being said..it would great if…everyone say…**feature request** adwords itself had a statistical significance calculator built into its ads tab so you could run the data right there.

    They do it with GWO. I am suprised it has never come up as a bell and whistle for adwords, especially when it comes to adword creative testing…

    Well, you know, a girl can dream…Have a good day. Mary Kay

  • FernandoGarces

    Thank you for the post, Matt. I agree with you on most everything. Considering the barrage of updates and enhancements that Google keeps putting out, it is very difficult to justify the amount of time and effort required in properly managing a good long-tail strategy.

    What caught my eye, though, was your idea of getting rid of what you call “completely useless keywords that are only taking up space in your account.” If nothing else because, just a few paragraphs earlier, you had said: “you are going to use a long tail strategy, you have to embrace the uncertainty that lies within the fog of long tail keyword inventory and get used to the reality that there is a direct overhead cost associated with maintaining a long tail inventory. An acceptable overhead is one that keeps your overall campaigns profitable, while extending your reach with your search keyword.” As George so adequately explained, your most productive keywords today can be your deadwood of tomorrow. If you continue deleting triple-zero keywords every, say, couple of months, either your keywords are not truly long tail or after a year you will end up with no long tails at all. I understand that keywords that show no activity can bog down your account and create inefficiencies but that’s part of embracing the long tail. Why would you delete a keyword that, so far, has cost you nothing?


  • http://www.findmefaster.com Matt Van Wagner

    Fernando – thank you for commenting. If you have keywords with 0 impr, 0 clicks, and 0 conversions over some period of time, what’s the sense of keeping them?

    George’s point, if I understood it correctly, was that one should not cut long-tail keywords based only on conversion data alone – (which is what my example, unfortunately, seemed to imply).

    Triple zero keywords are more likely to be true deadwood – and never will see the light of day

    True, they don’t cost you anything, and are not part of the direct overhead expense I was describing. However, there are indirect costs in management time and ineffeciency in filtering reports to get at your actionable data. So, personal preference for me is to not deal with the triple zeros.

  • http://www.findmefaster.com Matt Van Wagner

    Hi Mary Kay,

    Google is always very receptive to new feature requests, so a girl should dream!

    The inherent challenge in managing within the long tail fog is you never will have enough data to make a individual kw decision, or else you have to invent ways to aggregate data within the fog – looking for hot spots and similarilities that you decide can be aggregated so you can get to a point where you can make a statistical calculation. Otherwise, you have to just accept that you will always have inefficiencies and move on to other PPC challenges.