We are just so busy, busy, busy as search engine marketers on behalf of our advertisers! There are the mission critical tasks such as managing budgets, getting tracking implemented, hitting CPA goals, and uploading new keywords/ads to the account. Then there are the things we definitely need to get to every week such as pacing, performance reports, testing, managing bids, etc. And then there’s everything else.
All of the other hundreds (if not thousands) of things we could be doing to help our accounts accomplish their goals like match type testing, landing page optimizing, keyword expansion, quality score issues, query mining, yada yada yada…the list goes on and on. Some things just get pushed down due to priority. It’s reality!
Well, I hope that one of the things you’re still paying close attention to is your negative keyword strategy. Just as having relevant keywords in your accounts is the right way to bring in consumer attention, I see negative keywords as the opportunity to optimize your relevance at the other end of the spectrum. When it’s a cost-per-click game, the winners are the ones that maximize their impact every time.
I worked for a client once that sold bulk light bulbs to electricians, maintenance men, contractors, etc. They were getting inundated with people looking for single hard-to-find bulbs from their paid search marketing. These users were wasting their budget and tying up the phone operators. We tried running just on bulk terms such as warehouse pricing or by the case but the search volume wasn’t there. We had to turn our light bulb terms back on but this time added in a ton of negatives. That did the trick.
I asked Larry Kim from Wordstream, to share his thoughts on the importance of utilizing negatives. He made a pretty good case that getting negative keywords right can affect your entire account:
If you only had time for one keyword optimization task for an existing PPC account, I would suggest finding negative keywords – because whenever I look at a client’s search query report, there’s always so many junk search queries that generated clicks for their ads. But you should be even more concerned about the junk searches that didn’t generate clicks – these are likely wasting impressions and lowering your click through rate, which lowers your quality score, which in turn raises the cost-per-click across your entire group/account and even reduces your impression share/ad position – thus negative keyword discovery is a very leveraged keyword discovery task!
One of the best tips I can share about negative keywords was passed to me by Rich Stokes of AdGooroo. I asked him point blank a few years ago what his single best PPC tip was for me — something I ask to many SEM pros I meet. He said negative phrase match. Huh? Most people usually say stuff about bids, landing pages, or quality score. Negative phrase match? Hmm. In the past, I had treated negatives like a black list which is more of a blunt object than a surgical tactic.
So why negative phrase match? Remember, there are three negative match type when combined with broad, phrase, or exact. Rich knew that mastering those are just as important as mastering the standard match types (should we call them positive match type? i.e. Positive Broad vs. Negative Broad). Here’s a good visual breakout by AdWordsPro Sarah from the Google support forums.
Have you using the three negative match types appropriately for your campaigns?
Now that I’ve got you ramped up on some negative keywords appetizers, it’s time to bring in the main course. The following is a Q&A with Ken Jurina from Epiar who is someone I think just might know more about negative keywords than anyone else in the Western Hemisphere. His search firm handles traditional SEM duties but also specializes in negative term list generation via their Epiar® Negative Keyword Lists™ (NKL) product.
Q. So what was the draw for you to negative terms as a specialty?
Ken: Of course the importance of negative keyword lists played a huge role, but we were also already perfectly positioned to develop negatives for two reasons:
First, our specialty has really been developing keyword software capable of efficiently mining and processing tens to hundreds of thousands of phrases per project, primarily for our SEO and market research services. We have been effectively working with the long tail for many years. (Getting PPC impressions in the long tail is the goal of broad and phrase match.) Our keyword software is a very advanced system, and it’s that system that gives us the edge.
Secondly, one of our fundamental strategies in the development of our keyword research software has always been to work with individual words (and two, three and four word occurrences) by applying various semantic theories, databases and filters to those words and use the outcome to understand entire the entire group of phrases. One aspect has been to identify the words in a phrase that indicate that the search is out of scope, out of context, out of market, non-transactional, etc. Since it’s these same individual words and often two word combinations that make up negative keyword lists, we have, like the long tail, been working with the meat and potatoes of negative keyword lists for years.
Q. Most SEMers understand the value of using negatives for PPC campaigns. I’d love to hear what your thoughts are on this being a specialist on the topic.
Ken: There are just so many reasons to using broad and phrase match with an extensive negative keyword list. For a start, companies that aren’t using broad and phrase match in PPC just can’t get a big share of the market. I know impressions, click through rate, conversions etc are the buzz words in PPC, however, one of the most important concepts in marketing communications is an old media term called reach. I would ask the question, “Of all the people that were actually searching for your product or service last month, how many were you able to show your ad to?” Since the long tail comprises around 50% of all searches, the maximum potential reach of an exact match campaign will consequently be less than 50%. With a good broad and phrase match campaign, comparatively, the potential reach is 90%+.
To give you an idea of how long the long tail is, in a study of 326,474 search frequencies relating to cooking, cookbooks and recipes, 157,386 were unusual or longer phrases that were only searched once, almost half. 235,279 phrases, 72%, were searched less than ten times. Then when you consider that the long tail is consistently shown to convert better than short tail phrases and for a lower CPC, what it adds up to is that if you want a lot more reach, a lower CPC, and a better conversion rate you go broad and phrase match.
Then there is the issue of negative keywords. The overview is that, besides many seemingly related searches being out of scope, context, etc, the fact is that a vast majority of searches are made by people not looking to spend money on any product or service. That means that a good negative keyword list is going to have to take out a vast majority of possible impressions, and that is going to take an extensive list.
Q. What common mistakes do you find with search marketers that have built their own negative lists?
Ken: Of course any negative keywords help, but we find the lists tend to be far from extensive enough – typically less than five hundred terms terms long. As in everything, you need the right tools for the job, and developing a good negative keyword list is a huge job.
Consider the cooking study mentioned earlier. Those 326,474 phrases totaled 1.37 million words! Who is going to tackle reading them all, or even a good number? Even when we broke it down to studying individual words – there were 40,196 different words. That’s where applying various semantic theories, databases and filters to those words are necessary.
Without the use of the proprietary keyword tools Epiar has developed, doing a good job is almost impossible.
Q. What are some things you can tell me about your term database?
Ken: For a start, we have extensive databases of languages and geographic places numbering in the millions of words. We also have over one hundred and fifty thousand words in over seventy categories. We also have categories like company names, drugs, celebrities, health, education. We even have verbs, adjectives, adverbs, superlatives, etc.
When you consider the average college graduate knows about twenty five thousand words and uses around fifteen thousand, you can see that just our categories alone have ten times as many words as most people even know.
Q. How do you mine these terms?
Ken: We’ve been developing these databases for over nine years, in some cases painstakingly one at the time. But the semantic capabilities of our keyword software have also been great at developing these lists.
For example, in a keyword study of 18,000+ phrases containing the word service or services, our software has about 3 ways of automatically determining that station is not a service but landscaping is. From there we very quickly put together a category of 550 business services.
When doing a client’s project our databases handle about 30% of the problem and the semantic capabilities of the software handles about 40%. The rest is just sweat.
Q. What kinds of ROI impact have you seen with campaigns that have used your negative terms service?
Ken: For almost 2 years we’ve measured a typical 5% to 40% savings on PPC spend for clients using Epiar negative keyword lists; depending on which of the three types of negative keyword lists they bought from us. While I know that type of ROI is probably shocking to some, it’s true and we have the testimonials to prove it.
Due to the competitive advantage we provide, clients don’t often let us mention their company name publicly, however Allan Dick from www.vintagetub.com has allowed us to do so:
Vintage Tub, who sells a great deal bathtubs and bathroom/kitchen accessories online, saw a Google AdWords savings of almost 20% on their PPC spend while maintaining their same sales volume. For another client who manages the PPC spend for a series of online dating sites, the client was spending approximately $1.2 million per month across Google, Yahoo! and Bing. Depending on the month, we were able to save him between $200,000 to $400,000 per month on his PPC spend.
For those interested, we sell three types of Epiar negative keyword lists for which the details and typical ROI for two of them (Epiar Online NKW and Epiar Premium NKL) are outlined at www.negative-keywords.com. While those two lists are typically suitable for people spending from hundreds up to $25,000/month on their PPC spend, we offer a very custom 10,000 term long negative keyword list for companies spending $25,000/month or more on PPC.
Q. How would a company measure the impact of the negative list once implemented?
Ken: The results are instant and very apparent via comparing past and current PPC reports. For the campaigns we have monitored the typical results we have seen as are follows:
- CTR up 20-100%
- Average time on site up 10-50%
- Page views per visit up 10-50%
- Conversion rate up 20-40%
- Profit up 10-40%
- Wasted impressions down 30-50%
- Bounce rate down 10-50%
- Cost per sale down 15-25%
- Wasted ad spend down 10-40%
As you can see the impact is positive in all the right places, and a very nice side affect is of implementing a proper negative keyword lists is a much improved Google Quality Score as well.
Q: Do you see any new ways we will be using negative lists in the future?
Ken: In both PPC and organic ranking, search engine indexes are so huge that businesses are necessarily at the mercy of probabilistic information retrieval. Developing an extensive library of both positive and negative words that define a business, as well as a few other variations, is itself probabilistic and a necessary step to deal with the massive datasets search marketing research yields.
For example, we have been collecting the complete set of search results for clients, usually numbering in the hundreds of thousands of listings, for the purposes of link development campaigns. Basically, in our opinion, Google views links from sites within the search results as relevant. Based in that premise we have been experimenting with negative key words to filter out sites that are inappropriate link providers.
For example, in the search results for “signs”, for a sign manufacturers link campaigns, we are able to filter out sites that mention “astrological”, “Gemini”, etc as well as “signs of pregnancy”, “addiction” etc. It’s interesting that the same themes that are negative in keyword research consequently, perhaps necessarily, turn up in clients’ search results also. There are incredible analytic synergies between keyword research, search results and web analytics, and the best way to capitalize on them is the development of extensive negative and positive word lists. In the probabilistic world of the Internet, you need to fight fire with fire.
Q. Any tips for the do-it-yourself negative term builders out there?
Ken: I’ve got a great “how to” presentation which I gave at Shop.org on the tools and techniques to create a PPC negative keyword list yourself.
But in the end, if a person wants to try and create a good negative keyword list themselves they need to be persistent and have patience. Developing an exhaustive negative keyword list is hard work even with the tools available, which is why most are 200 terms long or less.
One word of caution is that many people will look to their Google Search Query Performance Report on a monthly basis, sort by the highest and lowest CTR to identify negative keywords. While that report will likely provide some good negative keywords, people need to be prepared to do some good old keyword research to find negative keywords as well. Many people don’t seem to know that the Google Search Query Performance Report only shows you the phrases that caused your ads to appear and be clicked on, not the phrases that caused your ads to appear that were NOT clicked on. If you’re willing to do some keyword research to identify the negative keywords that cause your ads to appear in the first place, that’s often where you’ll find the real gold!
Opinions expressed in the article are those of the guest author and not necessarily Search Engine Land.