So your ad is sitting on position 1 on Google, and you get a reasonable amount of clicks a day. Your ROI is excellent and you think you have optimized your campaign to it’s full potential… You may also be mistaken.

One of the metrics in the Google Adwords reports that is often overlooked by search engine marketers is impression share. Impression share is basically your campaign’s market share of all the impressions it is eligible for. If you are bidding on the word ‘widget’ you would ideally like your ad to be shown every time a user searches Google for this term. Many people will be surprised to hear that this is not the case, not even if your ad is normally in the top position.

Here’s why: every time a user enters a search queries, Google executes an auction to determine which ads should be shown. As your competitors come and go, and alter their bid prices, your ad does not always win this auction. Sure, on average you might be doing OK but most likely you don’t win every auction.

According to the Google reports, there are two reason why you might not reach your maximum impression share. Firstly, if your daily budget is too low, Google will exclude you from certain auctions in order not to exceed your budget. In the other case, you lose the auction because your ad rank is too low.

Google defines ad rank as the product of Quality Score (QS) and your max CPC so increasing either of these two factors will increase your ad rank. Unfortunately, Google only reports on impression share on a campaign level, so you need to structure your campaigns cleverly to make optimal use of this data. For instance putting all your trademarked keywords in a seperate campaign will help you to determine how often your ads are shown when someone searches for your brand (ideally, this should be 100% of the time).

The easiest way to increase your impression share is to increase your max CPC. Increasing your bid prices doesn’t mean that your effective CPCs will increase, especially not if you’re already occupying the number 1 position. It does definitely increase your risk though, as a determined competitor could send your CPCs soaring. And it definitely will increase your total costs as you will gain more impressions and more clicks. But as long as your are generating profit from your increased traffic, it makes sense to try and grow your impression share.

Obviously obtaining a 100 percent impression share is unlikely to be a viable goal. But looking at this metric will give you a good idea of the size of your potential market and your position in it.

The whole idea behind a negative match is to cut out irrelevant impressions, and possible clicks, on your advertisements. Irrelevant clicks are similar to unwanted customers, or even shoplifters coming to your store. Of course, if you’re running a large campaign thats generating a lot of traffic, it becomes difficult to monitor your traffic.

This brings me to my principle argument. Negative matching is not always a helpful procedure, and in fact, if done wrong will ultimately be damaging to your campaign bottom line. For instance, if I’m selling goods for a merchant who sells organic products, but does not stock pumpkin seeds, I will most likely have to use negative matching for certain keywords. I’ve structured my account in such a way that I have made one campaign for organic skin products, and another for various organic food products. Now to the negative matching. Instead of running negative matches on [pumpkin seeds], [cheap pumpkin seeds], and [what are pumpkin seeds], I simply need to broad match the negative keyword ‘pumpkin’ on a campaign level.

The more complex negative matching happens when you have conflicting keywords in similar campaigns. For instance, your food campaign contains different brands of poppy seeds, as well as various brands of sunflower seeds. ‘Organix’ has both types of seeds, while ‘Natures Way’ has only sunflower seeds. Broad negative matching on a campaign level, for either the brand, or product in this case creates a problem. In this situation negative matching should ideally be done on an ad group level, using broad match types.

For a campaign with 100 000 keywords though, where the merchant has a vast product line, this process is very time consuming. The ideal is that you want to cover the long tail of negative matches, giving you a high CTR, with relevant copy, and a great landing page for every keyword you’re running, but more often than not, the time requirements of these bigger campaigns do not allow for this. If you don’t have the time for group level negative matching, then the second best thing is to look at your high density and high traffic keywords…either before or after the campaign has launched.

If it’s before the campaign launch, you can use Google’s traffic estimator tool (allows for bulk operations), and then run your high traffic keywords through the keyword tool. In this case you can do some head negative matching pretty quickly for all your high traffic terms.

If you want to do negative matching afterwards, take your existing information on your high impression keywords as a pointer of where the work needs to be done.

Lastly, with regards to negative matching for newly generated keywords. Personally, I feel that instead of exact negative matching all keyword terms that are irrelevant, I would firstly look for patterns in your irrelevant keywords that you could generate strong broad negative matches for. This will get to the root of the word, and avoid any possible misspellings/variations that could come up the next time you do new keyword research. By simply running exact negative matches, you are not accounting for the sheer amount of semantic variation that Google has to deal with on a day to day basis.

Any other thoughts on improving your CTR through negative matching?

Over the past few weeks we have been spending extra attention to the evil beast of PPC advertising: click fraud. During the holiday rush it is easy to miss suspicious traffic since the ‘normal’ campaign metrics no longer apply. Traffic soars, adcopy needs to be updated to reflect the latest offers, items go in or out of stock, seasonal and holiday related keywords come and go. All this within the space of a few weeks.

There were two main reasons to really scrutinize our traffic for fraudulent clicks at this time of the year:

1. To offset the increased risk of click fraud during the peak retail season.
2. To see if Google’s assertion that they catch most invalid clicks holds true when traffic volumes spike.

We looked at mountains of data from our campaigns and created experiments to test the robustness of our own monitoring systems and Google’s filters. In general, we found that both performed really well although one can never assert with 100% probably whether or not a click is fraudulent (the best you’d come to certainty is when it converts into a sale).

We learned a lot from this exercise and decided to update our click fraud risk calculator based on our findings. This calculator does not tell you what percentage of your traffic is fraudulent. Google already does an excellent job at that if you’re not using your own (or third-party) monitoring systems. What it does do is assess the risk profile of your PPC campaign so you can take appropriate measures to mitigate this risk.

Hopefully you will find our little tool very useful and if you are interested in a more in-depth analysis of the intricacies of click fraud, have a look at this Google click fraud presentation, or read more about our findings here.