homepage Welcome to WebmasterWorld Guest from 54.225.1.70
register, free tools, login, search, pro membership, help, library, announcements, recent posts, open posts,
Become a Pro Member

Visit PubCon.com
Home / Forums Index / Google / Google AdSense
Forum Library, Charter, Moderators: incrediBILL & jatar k & martinibuster

Google AdSense Forum

    
How to identify Advertisers' negative keywords
A practical technique to identify problem words on individual pages
21_blue




msg:1324334
 11:59 am on Jun 3, 2006 (gmt 0)

In his recent thread on keywords [webmasterworld.com] Kaz has highlighted a potential problem for Publishers: that the inadvertent use of certain (negative) keywords on pages can depress earnings. This is because Adwords advertisers can nominate negative keywords whereby, if a page contains one of those words, their ads do not enter the auction.

So, having had the problem brought to our attention, the next question is how us publishers can identify the negative keywords that we have inadvertently included in our pages but our advertisers are using to withhold their ads.

One approach is to block generic words that seem likely to be unwanted by advertisers – for example, the word “free” (which Kaz suggested in his thread). Advertisers may block this word because it suggests that the surfer doesn’t want to pay. We publishers may inadvertently include it in our pages because there are over 20 other meanings/uses for the word which may seem appropriate to our page – eg: “free thinking”.

Blocking generic words may work for many sites, but there aren’t too many words that are such obvious candidates for blocking as “free”. In fact, even “free” may not be an appropriate word to avoid in all circumstances. For example, having done some research on my advertisers, there are some niches where the word “free” is used extensively in both adverts and on advertisers’ landing pages. So, removing the word “free” across all my pages might damage income in some areas.

Another approach to consider

Another approach that I’d like to put on the table is to do word comparisons between relatively high and low EPC pages in the same niche. Words that appear on low EPC pages and not the high ones are then candidates for being negative keywords.

For example, I’ve done this with a couple of popular pages on my site and found the name of a proprietary licence in the low EPC page. Advertisers may be blocking this because, if they are not licensed, they can’t provide products, services or support so the clicks from such pages may convert poorly. From my perspective, the use of the proprietary licence name on this page is incidental – I have other pages that talk about it specifically – so I could easily remove the name without affecting the page content. However, the fact that I have many other pages that talk extensively about aspects of this proprietary product means it is not something that I would normally think of excluding from individual pages' content (that is, until I used this technique).

I’ve had a look round for a software download that compares two pages and produces a word-list comparison between the two. There probably is one, but I haven’t found it yet (please let me know if there is). So, I’ve designed an approach to do this word comparison using Excel and Word that I’ll outline here. It takes about 5 to 10 minutes once you know what you are doing.

Step 1 - Identify relatively high and low EPC pages

This requires you to have URL channels, or custom channels on individual pages. Choose two pages in the same niche that have differing EPCs over a long period (eg: a month). (To do this comparison, download a CSV report of pages in one niche into Excel, calculate EPC for each page as earnings/clicks, sort by EPC, and then pick the highest and lowest page in the sort order).

Step 2 – Create a word list for each page

You can do this with Word, by:

  • Make a backup copy of your html file
  • Open it in Microsoft Word
  • Select all text and copy to the clipboard
  • Open a new Word file
  • ”Paste Special” into the new Word file, selecting the option for “unformatted text”
  • Do a Replace on the whole document, changing “ “ (a space) to “^p” (the new paragraph symbol)
  • Save the file as a text file

Repeat the above process for every file you want to compare.

Step 3 – Create an Excel file and column with the page name

Create a new Excel file with two columns labelled “Word” and “Page”. Save it as, say, “comparison.xls”.

Open each text file in Excel (each word should appear in a new row), create a second column, fill that column with the name of the page from which the words came, then copy the two columns to comparison.xls

Once you have done this for two files, you should have an Excel spreadsheet that looks something like this:

WORD.....PAGE
The......page1
quick....page1
brown....page1
:
:
dog......page1
How......page2
slowly...page2
does.....page2
:
:
big......page2

Step 4 – Run a pivotable analysis

Pivotable is on the Data menu in Excel. After confirming the data to be used, select layout and drag “word” onto the row, “page” onto the column and “word” onto the data (making sure it is set to ‘count’).

When you run the pivotable, this will give you a report showing how many times each word is used in each page.

Step 5 – Sort the table

Sort the pivotable on the high-EPC page column, which will cause all the words in the low-EPC page and not in the high EPC page to group together.

Conclusion

Despite the length of this post, this technique does not take long and it enables you to identify a manageable-sized list of words from which you may be able to identify negative keywords that may be depressing a particular page's performance.

Considering negative keywords on a page-by-page basis can be beneficial because, as can be seen in my example above, in one page a proprietary name may be a big plus (increasing relevance) whereas in another it may be a problem (putting off advertisers).

 

dhiggerdhigger




msg:1324335
 4:11 pm on Jun 3, 2006 (gmt 0)

I'm not sure if what follows is good advice, but it's certainly an option. If you want to include "negative" keywords, you can write them in the source in reverse, and then switch them back using CSS.

For example, the word "free" can be written as "eerf" in your html source, surrounded by <span> tags:

<span class="reverse">eerf</span>

Then a CSS rule (provided for languages which don't read left-to-right) can reverse the contents of the <span></span>:

<style type="text/css">.reverse {unicode-bidi:bidi-override;direction:rtl;}</style>

The reversing of words identified as "negative" could be done automatically on the server before a page is sent.

21_blue




msg:1324336
 9:10 am on Jun 4, 2006 (gmt 0)

Very clever idea, dhigger. However, it has a slightly spammy feel to me.

I don't need to do it anyway. I've been able to just remove the proprietary licence negative keywords from selected pages and already seen an improvement in the quality of ads appearing on those pages.

howiejs




msg:1324337
 3:55 pm on Jun 4, 2006 (gmt 0)

"proprietary licence negative keywords"

What is the above?

kaz




msg:1324338
 7:08 pm on Jun 4, 2006 (gmt 0)

My understanding of that .. an example would be a term like Nike, where some retailers may exclude it because they are not licensed to carry it.

Mentioned was

Advertisers may be blocking this because, if they are not licensed, they can’t provide products, services or support so the clicks from such pages may convert poorly

Interesting approach 21. I'm curious, have you tried to reverse the tactic to try and identify higher EPC terms also? Perhaps those that you consider higher EPC pages would be worth considering broadening your content. I say this because of an old marketing equation i try to follow ... improvements in exposure are not usually found by only eliminating poorer performing marketing tactics, but also by combining with an increase of marketing tactics towards those methods that are working. Basically, never reduce but replace and increase your marketing. Your approach seems a bit focussed on reduction without replacement of more marketing.

21_blue




msg:1324339
 8:17 pm on Jun 4, 2006 (gmt 0)

kaz wrote:
I'm curious, have you tried to reverse the tactic to try and identify higher EPC terms also?

Yes and No. Yes, in that identifying high EPC words is something that we've done for a long time - it's part of the equation in prioritising our work, or to help decide whether it will be worth our effort to develop into a particular niche.

But what I haven't done (until 2 minutes ago, that is) is use the above technique to see if there are any specific trigger words on the high EPC page that could be used on the low EPC page to attract higher paying ads. This was what you were suggesting in the opening post of your original thread, I think.

So, I've just had a look at the spreadsheet and, guess what, I think there is one word that could possibly expand the range of high paying ads (eg: the low EPC page talks about widget engineering, but the high EPC page talks about widget design and engineering).

I agree with you that one doesn't want to focus on negative keywords only. My above approach may seem oriented that way only because it has been something we've neglected to date (and your "free" example woke us up to). And I've talked about it because on some pages the immediate impact of removing negative keywords has been quite dramatic - eg: on one page we had previously removed ads altogether because of poor targetting, because Adsense would only display ads from a low paying niche. Having removed the proprietary name and reintroduced the ads, Adsense is showing ads that are much more appropriate and I recognise as high paying.

So, I think you've flagged up an important issue. Hopefully, the above technique provides a simple way of identifying important keywords for individual pages (whether they are to be avoided or included).

kaz




msg:1324340
 8:46 pm on Jun 4, 2006 (gmt 0)

Thanks for sharing that example of yours, in your scenario I would make sure that 'widget design' is an emphasis on my pages. I would want to make sure that of all the people that are reviewing my widget information see the 'design' segment of my site and make sure that those advertisers targeting design are receiving exposure to the 'design' visitor subsegment of my traffic. A high quality match for advertiser, google and publisher ... and I believe by placing some focus on this sort of approach over time that smart pricing does improve because conversions improve because the traffic is more qualified based on what the advertiser is offering and communicates converts/pays based on the higher epc/ppc.

dhiggerdhigger




msg:1324341
 9:33 pm on Jun 4, 2006 (gmt 0)

21_blue - yeah I agree with you. I've used it for obfuscating email addresses.

21_blue




msg:1324342
 10:29 pm on Jun 4, 2006 (gmt 0)

dhiggerdhigger wrote:
I've used it for obfuscating email addresses

Thanks, I was wondering how that feature could be used constructively. But in your email application of the reverse feature, does it actually work? It seems a trick that email harvest programs would probably recognise and therefore deconstruct the email address (we hide our emails completely).

kaz wrote:
I believe by placing some focus on this sort of approach over time that smart pricing does improve

I'm not sure I agree with your smartprice assertion, depending on whether you are talking site-wide or on a page-by-page basis.

Our main approach for existing pages tends to be to look for the main keywords that people use to find each page and then keep the page optimised for that keyword. Eg: if most surfers visit a particular page using the search expression "pink fluffy widgets", then we'll optimise that page for 'pink fluffy widgets'. However, I guess that is more an SEO issue, not a smartpricing one - we are concentrating giving surfers what they want, not something that is slightly different but potentially higher earning.

But if you focus on high-earning expressions like "widget design" site-wide then it could depress earnings. For example, 'design' may be used as a negative keyword by maintenance companies who are only looking to establish support contracts, not find design engineers. Using this word site-wide could cause our pages about maintenance to perform poorly.

In fact, I take this view because of the idea you planted in my mind that we may inadvertently use certain words on our pages that depress earnings. Hence, if I talk about "free thinking widget design" or "Nike sportswear" I may actually be using words (ie: 'free' or 'Nike') that some widget designers or sportswear suppliers use as a negative keyword in their campaign, but others use as a prime search term. In future, on our pages that are primarily about Nike sportswear I'll leave the expression there, but I will remove the word 'Nike' from pages that are mainly about generic sportswear.

That is, it has to be done on a page-by-page basis, imho.

kaz




msg:1324343
 11:11 pm on Jun 4, 2006 (gmt 0)

I can agree with the page by page approach, not much other way to handle it, but let me clarify my communication


I'm not sure I agree with your smartprice assertion, depending on whether you are talking site-wide or on a page-by-page basis.

Our main approach for existing pages tends to be to look for the main keywords that people use to find each page and then keep the page optimised for that keyword. Eg: if most surfers visit a particular page using the search expression "pink fluffy widgets", then we'll optimise that page for 'pink fluffy widgets'. However, I guess that is more an SEO issue, not a smartpricing one - we are concentrating giving surfers what they want, not something that is slightly different but potentially higher earning.

I think we agree on some points, but disagree on another. Let me give an example to explain. Say I have a section of content with a focus on widgets. For example, my pages are

Widgets
Blue Widgets
Green Widgets
Orange Widgets
Custom Widgets

I provide good content/value on each and they all attrack traffic based on the approach you discuss. Over time and thru observation I notice that custom widgets seems to be a big draw for my advertisers although the traffic is less than the others. A common scenario with adsense based on my experiences, simply it makes more with less. With the custom situation, it makes sense based on further research because there is a larger minimum order. Now I'm not suggesting the extreme of someone tearing down their blue green orange pages and focusing on the traffic on custom. What I am suggesting, is that I want to respect the people who have found the orange widgets page and suggest to them an alternative - the custom option. This is presented near the adsense advertising, as I want those that might have an interest in custom to view those ads (~adlinks). It filters the traffic so that those with an interest in my highest profit margin page, are exposed to that without devaluing the traffic for the advertiser. Custom is then exposed for each, so essence I'm targeting custom widgets, blue widgets and now custom blue widgets. Getting into that long tail the advertiser loves because they qualify/convert more. I look at as though i'm filtering traffic/visitors for the advertisers in a similar way that adlinks does - but to my own site with my highest value pages.

wyweb




msg:1324344
 11:23 pm on Jun 4, 2006 (gmt 0)

It seems a trick that email harvest programs would probably recognise and therefore deconstruct the email address (we hide our emails completely).

If they had encountered it enough, probably. I doubt they have.

fredw




msg:1324345
 4:06 am on Jun 5, 2006 (gmt 0)

If the style that did the work was in an external style sheet, and everyone who did this used a different style name, how could they ever prgram their harvester bots to recognize this?

ronburk




msg:1324346
 4:26 am on Jun 5, 2006 (gmt 0)

Another approach that I’d like to put on the table is to do word comparisons

Or, if you've cultivated a relationship with your advertisers as publishers have from time immemorial, you could just shoot'em an email and ask what sort of things they do and don't like to see in content that relates to their product. A lot less effort, a lot more information, and another reason for them to stick with you when they're cutting back advertising elsewhere.

(Or, if you believe the oft-repeated myth that the Google T&C forbids you from talking to advertisers, you can go through incredible machinations to make guesses about simple info like this. :-)

dhiggerdhigger




msg:1324347
 8:39 am on Jun 5, 2006 (gmt 0)

I don't want to get too off-topic, but to answer questions posed:

21_blue: seems to work. I convert the addresses to a series of html entities before reversing them.

fredw: the @ sign, probably, entity or not (but yes - I don't use "reverse" as the span class name!)

Dave

21_blue




msg:1324348
 10:00 am on Jun 5, 2006 (gmt 0)

kaz wrote:
I look at as though i'm filtering traffic/visitors for the advertisers in a similar way that adlinks does - but to my own site with my highest value pages.

OK, I understand what you are suggesting. I agree with you, in fact we do use this approach and liken it to offering a smorgasbord that contains what they were looking for and higher paying options.

However, we do also include some lower-paying options on the smorgasbord as well, for smartprice reasons.

To use your example, suppose a surfer is looking for cheap mini-widgets. If we only offer more expensive options - everything but the cheaper mini-widgets - they may click on the ads but not convert. This yields a short term rise in income but ultimately hits smartprice.

21_blue




msg:1324349
 1:05 pm on Jun 5, 2006 (gmt 0)

ronburk wrote:
the oft-repeated myth that the Google T&C forbids you from talking to advertisers

It's a good idea, Ron, but is it really within the terms and conditions?

"3 Communications Solely With Google. You agree to direct to Google, and not to any advertiser, any communication regarding any Ad(s) or Link(s) displayed in connection with Your Site(s)."

Is not asking about the negative keywords 'communication regarding ads displayed at your site"?

Global Options:
 top home search open messages active posts  
 

Home / Forums Index / Google / Google AdSense
rss feed

All trademarks and copyrights held by respective owners. Member comments are owned by the poster.
Home ¦ Free Tools ¦ Terms of Service ¦ Privacy Policy ¦ Report Problem ¦ About ¦ Library ¦ Newsletter
WebmasterWorld is a Developer Shed Community owned by Jim Boykin.
© Webmaster World 1996-2014 all rights reserved