Ellerton Whitney – Search Marketing

I would like to submit my feedback that the new ad rotate setting is going to be the most detrimental “upgrade” to AdWords ever rolled out.  I have never been so mad about anything in my six years as a search marketer.  30 days is not enough time for a large number of ad groups to determine what ad should be a “winner.”  There is a fine line to walk between getting a good click through and a good conversion rate.  None of it is as easy as simply selecting “optimize for clicks” or “optimize for conversions.”  Take, for example this ad group:

Guess which ad is going to win eventually under the new “ad rotate” feature?  The ad that is performing the worst in terms of conversion rate (the first ad).  But, we don’t want to simply optimize for conversions either.  These conversion rates are too low to make sufficient decisions based entirely off of conversion rate — 30 days is not enough time to make good decisions.  A given ad group could have four instances of the same exact ad, and conversion rates over a 30 day period could that look like they are four completely different ads.  With the example above, the third ad is most likely to produce the most gross profit based purely off of these numbers, but these numbers are no good to be making decisions with given the nature of the user experience that leads to conversions.  More time is required to make a good decision, dependent upon the amount of traffic a given ad group gets.

Another example:

Why must advertisers be forced to choose a winner here?  These ads all perform, more or less, commensurately.  Perhaps, over time, a real winner will be revealed, but for the past thirty days, the click through and conversion rates for these ads have been fairly similar.  Why must Google force a winner down every advertiser’s throat?  Why can’t the advertiser be allowed to take the time necessary to make the appropriate decisions.

Having ads that are supposed to “rotate” optimize for click through rate is the opposite of an innovation.  Optimizing for click through rate does not take conversion rate into consideration, and many advertisers purposefully test ads with lower click through rates in an effort to increase conversion rates.  Likewise, simply optimizing ad rotation for conversion rate does not properly take into account a number of factors, including click through rate, and the difference between the click through rate metric (which occurs on the Search Engine Results Page) and the conversion rate metric (website user experience).  Sophisticated advertisers have the ability and the know-how to make these decisions for themselves.

See also these posts from one of the most influential blogs in the PPC industry:

http://www.ppchero.com/googles-new-ad-rotation-settings-are-the-devil/
http://www.ppchero.com/sign-the-petition-to-keep-control-of-your-ad-rotation-settings/
http://www.change.org/petitions/google-adwords-allow-advertisers-the-option-of-continuing-to-rotate-ads-indefinitely

Why would Google intentionally anger so many advertisers?  If Google wants to make this change to the ad rotate feature, add it as a fourth ad rotation option.  There is more to improving user experience than simply looking at either click through rates or conversion rates.  I urge Google to roll back this change, or alter it to become a fourth ad rotation option.

Until yesterday, I was ecstatic and excited about the Google AdWords “Display Network” tab.  With its arrival, Google’s Display Network has become more powerful than ever.  The biggest change here is that advertisers can now see what specific keywords are leading to clicks and conversions.  Before, Contextual ad groups were a black box of either “working” or “not working.”  The only damper on my excitement is Google’s announcement yesterday that the “ad rotate” setting as we know it will be turning into the “optimize for clicks” setting after 30 days of stagnant ad testing in a given ad group.  In other words, blatant Google money grab.  Nonetheless, it is important to understand how the new Display Network tab works so that its power can be properly harnessed for optimal campaign performance!

The Display Network tab brings together keywords, placements, interest categories, and topics in one place.  However, the most interesting and important innovation has to do with keywords, placements, and how they interact with one another.  Without further ado, here are my key takeaways from my time with the Beta Display Network Tab:

With the Display Network tab, the placement bid will always trump the keyword bid. Upon first viewing the information available to you in the Display Network Tab, you might uncover keywords that you find are highly unprofitable, or not nearly as profitable as the other keywords they are grouped with.  The logical step to take here would be to lower the bid on the poorly performing keyword.  However, this action is not going to help your ad group if you are running lots of managed placements with placement level bids.  Consider this example:

The logical thing to do here would be to lower the keyword level bid on “banana pictures”  from $.60 to $.30 or $.40 since the keyword is under-performing the ad group so badly.  However, if this ad group has numerous managed placements with their own placement level bids, cutting the bid on “banana pictures” will do nothing.  See this example with the keyword level bids on the left and the placement level bids on the right:

If the managed placement “ehow.com” has a placement bid of $.75, all of the keywords in the ad group will always be bid at $.75 on that site regardless of what their keyword level bid is.  The only placements where the $.40 bid for “banana pictures” will be applied will be on those managed or automatic placements that are utilizing the ad group level bid.  So, you now might be wondering, of what use is this new Display Network Tab?

The best use for keyword level performance information provided in the Display Network Tab is to create more intelligent ad groups. If you have five keywords in an ad group and two are consistently subpar, those two subpar keywords now belong in a different ad group.  That different ad group can have a lower ad group level bid in an attempt to lower the acquisition cost on said subpar keywords.  Breaking out low performing keywords from high-performing ad groups will allow the lowering of bids for those keywords while maintaining the ability to adjust managed placement bids up or down.    In the example above, the best action to take would be to move “banana pictures” into its own ad group with a lower ad group level Max CPC (or discontinue that keyword from targeting altogether depending on your CPA goals).

If you really want to make use of keyword level bids in your contextual ad groups, it is best to do so without individual managed placement bids. As mentioned, just because we now have the ability to use keyword level bidding doesn’t mean we have to use it!  The best use of keyword level performance information is for the intelligence it provides.  However, if you decide you really want to use keyword level bidding, keep in mind it will only be applicable to those placements which do not have their own individual placement-level bid set.  The only scenario I can think of where this would be useful is an ad group where placements perform very similarly, but there is a deal of variation between the  performance of the keywords.

With the Display Network Tab’s arrival, it would be best to take a trip back in time. Those contextual ad groups that didn’t work before?  Pull back the date range and look at the keyword-level data, because it’s all there.  Uncover the golden nuggets that were working, if any, and re-launch the ad groups without the bad keywords.  I have seen quite a bit of success pulling  out a handful of high performing keywords from paused ad groups and jacking their bids to the sky.

The best practice of five keywords per contextual ad group is not as relevant as before. With the added visibility into what is going on with each keyword, there is now a bit more leeway with regards to the number of keywords you are grouping together.  While it is still good to try to “theme” your ad groups so that your placements perform similarly across keywords, your number could be fewer or greater than five and not really matter as much because you will know what is going on.  If you are diligent with your optimization, number of keywords per contextual ad group no longer matters.

Don’t Over-Optimize! With even more ways to slice and dice now, it will be important to resist the temptation to over-optimize.  While breaking out your 100 contextual ad groups into 500 might sound like a great way to increase control in theory, keep in mind  that major reorganizations don’t always going to go according to plan.  Make meaningful changes where you think they are necessary, but don’t break everything out just for the sake of it.

Final thoughts. The Display Network Tab user interface might take some getting used to, but for most experienced AdWords users, I think it’s fairly self-explanatory and does not require much of a tutorial.    Who will the Display Network Tab benefit most?  Advertisers that are doing large amounts of manually managed Contextual advertising.  If you are a heavy user of Google’s DCO and have a limited amount of Contextual advertising outside of your DCO Campaigns, the Display Network Tab is probably not going to be incredibly exciting to you.  Likewise, the Display Network Tab is probably not going to be a huge help for most advertisers’ Remarketing Campaigns.  In grand conclusion, the Display Network Tab offers a wealth of information that is a true game changer for heavy Contextual advertisers.  Make good use of the intelligence it provides, and be mindful of its limitations!

Not a terribly insightful post, but Google seems to be adding forward slashes after top level domains in both paid and organic listings.  I didn’t see anything in the blogosphere about it, but I know this is fairly new as I haven’t seen it before this week and it’s not in screenshots I took last week!

Once upon a time, I had a number of AdWords contextual network campaigns humming along quite nicely.  Then along came the Google serpent and offered me a bite of the Display Conversion Optimizer.  It did not come with a guarantee of working, but the promise was that many of my competitors in similar verticals had seen large upticks in volume at good CPA’s.  So, I said sure, let’s give it a go.  I think I wish I hadn’t.  The DCO has made a crack addict out of me, and now I can’t quit it.  I hope to God Google never stops supporting the Display Conversion Optimizer, or I will surely be SOL.  The following are my impressions, backed by both my numbers and my gut.

What Is It?

The Google Display Conversion Optimizer (or DCO for short) is a new option for heavy spending advertisers to start new Contextual Network campaigns.  What it does is attempt to automatically find new websites for your placements based off the historical performance of your Display URLs on the Google AdSense network.  Your Google Representative can probably provide you with a worksheet of eligible Display URLs and recommended CPA’s to try out with the DCO.  Thus, if your Display URL structure is weak, you might have a bad time with the DCO.  Note that not any old person can go and set up a DCO campaign in their account; you have to ask your Google Rep to get opted in.  If you do not have a Google Rep, you probably have a bad chance of getting opted in to the DCO.  However, if you have been opted, in, PPC Hero has a good blog post on what to do for DCO Implementation.

Pros of the DCO

• Gain effective volume from sites you wouldn’t have thought of that have next-to-no content on them, or gibberish content.  There are sites that might be a logical fit for you in terms of theme, but Google’s contextual network would never have really picked up on them because of their (lack of) relevant seeming content.  This doesn’t mean the sites are a bad fit, though.  Likewise, you probably never would have picked up on these sites as managed placements alone, either, because often times that would be too broad a targeting method.
• If you aren’t the brainstormiest of search marketers, the DCO goes out and finds places where your ads would not have otherwise been running.  This is not just referring to the aforementioned point.  This is to say it will find other contextual themes that you might not have thought of.  The double edge of this sword, however, is that if you have already done a fantastic and thorough job, the DCO might have a hard time finding more nuggets that make sense, and cannibalize your existing nuggets
• On the whole, you may be able to get a lot of incremental volume from the DCO.  However, this was not exactly my experience.  Ultimately the DCO took over from more ROI positive contextual network campaigns.  Though conversion volume improved greatly, margins shrunk so much that the net effect was more or less a wash.  Google – 1, Advertiser – 0.

Cons of the DCO

• My biggest complaint about the DCO is a lack of control.  You are basically at the mercy of the CPA you set, and CPC’s can vary greatly from day to day.  This can mean a less than desirable CPA for you.
• Cannibalization is an enormous issue with the DCO.  Many of my existing contextual campaigns have been, for all intents and purposes, obliterated.  This is why the DCO is like crack.  One you start, you can’t stop, because you have nothing left to fall back on.  The only control you are left with is negative placement optimization, but no ability to adjust bids at a placement level.  In an attempt to promote the best performing campaigns, I have added negatives to the DCO or my original contextual campaigns on a case-by-case basis, letting each placement reside in the campaign where it has performed best.
• DCO is still in beta.  This means there can be technical issues sometimes, and because Google has a stronghold (dare I say a monopoly) on this contextually-relevant sort of market, advertisers are basically at Google’s mercy to make it right when these occur
• DCO is probably not going to work well for you if you have a weak Display URL structure, unless all of your ads are extremely relevant to one sort of theme.

The Numbers

While I cannot publish exact numbers or where they are from, I can say that my overall contextual network conversion volume by month has decreased since implementing the DCO.  Furthermore, my acquisition cost has increased on a monthly basis.  While my numbers contain other hiccups and nuances to be sure, I have no concrete data thus far which indicates that the implementation of the DCO was a winning decision.  Any gains in certain pockets or places have been offset by increase in acquisition cost and decrease in control (and thus efficacy)

My Conclusion

I still have mixed emotions about the DCO.  It seems plausible to me that my campaigns could one day exceed the gross-profit performance of my “old” campaigns, but that has not happened yet.  Furthermore, as a performance marketer, I detest the hands-off approach that has to be taken with regards to the DCO.  If you do decide to try the DCO, I would recommend not to be too stingy with your starting CPA bids.  Furthermore, when you do make changes to those bids, make them meaningful ones.  If you do not give the automated tool enough wiggle room to play with, things may never get off the ground.  Watch your placements closely, and add negatives on a daily basis as needed (as soon as data is available for viewing) if you are spending in the thousands a day.

I forgot to post this here: My guest post at PPC Hero about the 10 things you shouldn’t neglect when performing Search Engine Marketing!  Here it is: http://www.ppchero.com/the-10-cardinal-sins-of-paid-search-marketing/

Google recently gave search marketers another nugget to let them know how well they are doing — Relative CTR.  While it’s no help for your search campaigns, think of it as a quality score of sorts for your contextual targeting.  Relative CTR tells you how your click through rate stacks up against your competitors on the websites where you ads are showing.  It does this with an order of magnitude of 1, with 1 meaning your click through rates are just as good as your competitors.  If your relative CTR is .3X (as the terminology goes), yours are only doing 30% as well as your competitors.

Now you ask, so what?  I’m already doing everything I can with my creatives to maximize my click through rate.  Well, maybe you are, and maybe you aren’t.  If you are targeting multiple ad groups that are only a few degrees of separation away from other one another in terms of theme, Relative CTR might prod you to take top performing ad creatives from one ad group and put them in another, or vice versa.  Relative CTR is also a helpful metric to let you know how your brand-new ad group is doing.  If you’ve recently started an ad group, and your relative CTR is .3X, you know things are going to be an uphill battle.

In sum, make sure you have relative CTR check on your list of columns.  It is sure to be a somewhat helpful metric when you are optimizing your content campaigns.

The title says it all on this one; Microsoft has finally added conversions to its search query reporting.  This was one of the very glaring features absent from Microsoft adCenter (ahem still no CPA column for Campaigns).  The feature appears to have been added August 2nd mid day, so use August 3rd, 2011 going forward for search query reports from Microsoft.  Unlike intelligent Google technology, which adjusts its CPA columns in search query reports for historical data that did not have conversion tracking installed, Microsoft just rolled this out very quietly and with no fanfare.  So, if you run an all time search query report hoping to see the conversion data, it isn’t going to be there!  So, don’t try that.

If you are a beginning search marketer and are wondering why this is significant, it is very convenient to be able to see which of your actual search queries (the actual phrases people searched on) are or are not leading to your conversions.  Thus, you can optimize your negative keyword lists, and add new keywords for increased efficiency.  If you wanted to track your adCenter search queries and conversions before, you needed a 3rd party tracking system, or at the very least, some Google Analytics hacking.  I personally find the latter to be a cumbersome approach when search marketing, so I am very pleased Microsoft has rolled this new feature out.

This week, Google finally started letting advertisers see Impression Share metric for their Content Network campaigns.  If you are wondering what Impression Share is, this is the number that tells you what percentage of the time you showed up for eligible impressions — in other words, how many times your ad showed up vs. the number of times it could have showed up.  Now, what does one do with these metrics?  Depending on the number of ad groups you have in your Content Network Campaign, Content Network Impression Share may or may not be useful to you.  Like Search Impression Share, if you are maxing our your opportunity at a given lead cost, Impression Share metrics may be a moot point for you.  However, they can be a good indicator of missed opportunity, especially Exact Match Impression Share for Search.

Presently, I have a number of Content Network campaigns, and many of them show a much lower Impression Share than I would imagine for them — either “under 10%” or just above that amount.  However, others show in the 20% to 60% range, which seems more acceptable.  The numbers do not make total sense to me given campaigns that show high positions and seemingly have excellent coverage are saying they have low Impression Share.  This may have to do with ad groups that show up in poor positions muddying the entire campaign.  Ultimately, Content Network Impression Share is just one more metric to let you know how you are doing with maximizing your opportunity.  So, if you really want to try to maximize this metric, perhaps consider splitting up ad groups into different campaigns — hopefully your ad group reorganization doesn’t do more harm than good!

At long last, Google has finally given users access to hourly conversion data through the AdWords interface.  This is extremely valuable information, particularly to search marketers performing lead generation, as they typically have little to no conversion latency to deal with.  With this information, you can choose to bid more in high converting hours of the day, and bid less in lower converting hours of the day.  But first, let’s identify some key concepts and themes to grasp before you start monkeying with your strategy!

Conversion Latency
If you are going to daypart, it is important to understand “Conversion latency.”  This is the concept of how long it takes a user to go from their first click to conversion.  If you are selling fancy toilets for $2,500, chances are you are going to have a lot of conversion latency while the person makes up their mind.  It might take one day, a week, or more for the person to convert, and they might click your ads multiple times.  However, if you are dealing with lead gen, there is typically little conversion latency, and as much as 70 to 80 percent of your conversions could be happening right there on the spot.  So, before you daypart, identify how much conversion latency you have.

In order to identify your conversion latency, go to “Conversions” under AdWords’ “Reporting and Tools” menu and then click “Search Funnels.”  By the way, if you do not have Google’s Conversion Tracking set up, I suggest you do so post-haste.  Once you are in the Search Funnels menu, you can look at path length and time lag. See menus at left here:

If you are getting a large number of conversions occurring with just one click, I think it is safe to say you are okay to daypart.  However, if you are getting two, three, or four clicks to a conversion, and it is showing latency of two to three days, you might want to think twice before you daypart up or down.  This is because your “Assisting” clicks may be happening during hours of the day that don’t show any recorded conversions, however, these clicks are critical to your closing the sale.  Thus, you do not want to bid lower during these time periods simple because you could potentially be cannibalizing all of your efforts.

Identifying Your Dayparts
Once you have identified that you want to daypart, then what do you do?  Return to your Campaigns tab in Adwords, and select the campaign you want to daypart.  If you want to daypart everything together, do not select a campaign — depending on your account structure, this may or may not be advisable.  Next, click the Dimensions tab.  If this is not selected, you will have to select it from the little arrow dropdown next to Campaigns, Ad Groups, Keywords, etc.  Here it is:

Once you are on the Dimensions tab, select Hour of Day, and then choose your date range.  I suggest at least three months of data to get a good sampling, but if you are working with a large spending account, less time might be acceptable.  Once you have your report, toss it into Excel, and identify groups of hours that are working well and not working well.  Here is the key: You can only identify six dayparts.  Thus, try to find six groups of hours that go together (sequentially) so that you can decide to bid them up, down, or leave them alone.  Also, you may want to consider bidding differently for different days of the week (particularly the weekend).  This might require some work to get the data you want in a useable form.

Setting Up Your Dayparts
Now that you have identified your six (or fewer) dayparts, you need to implement them on your campaign(s).  You will have to go to the settings tab for this, and then click on the ad schedule for your desired campaign.  Then you will have to change the Mode from Basic to Bid Adjustment.  Now you will have the option to set your Ad Scheduling.  Start with Monday, add your time periods (up to six), and select what you want your bid to be.  To leave it alone, set it at 100%, and to bid up put 110%, or down 90% — or more or less depending on how much you want the bid to change.  The main thing to keep in mind when setting up dayparting is to keep your changes minimal for a start.  For example, if you set your bid to 25% of its original bid, it’s likely the keyword will not show at all for that given time period, in which case you might as well shut it off (which is also an option within dayparting).  Keep to small changes at first, but not so small they won’t make a difference. In this example, the ads are shut off two hours of the day, but are running at various percentages of bid for the rest of the day.

Final Notes
A couple of final tidbits to go over.  One should know that dayparting doesn’t work with Google’s Conversion Optimizer.  So if you’re using this feature, you are out of luck.  Secondly, if you want to daypart in Microsoft adCenter, you have to select one of their six mediocre dayparts.  Sometimes these may align with your data, other times they may not.  Furthermore, adCenter does not make conversion data by hour available through their User Interface.  Thus, you will have to join your backend conversion data to your hourly spend from adCenter, and this does not properly mitigate for conversion latency.  Or, you can try to get fancy with your Google Analytics, but this will not have the spend data from adCenter incorporated.  In any event, this tutorial is best applied to Google because adCenter’s dayparting options are, as mentioned, weak at best.

In my own campaigns, I’ve seen dayparting boost ROI by 10 to 20% or more depending on the vertical.  So, if you aren’t using Conversion Optimizer, I highly suggest seeing what sort of efficiencies can be gained through dayparting!

Google gives you the option to have up to 100 campaigns in either a standard or an expanded AdWords account.  These are the limits for a standard/expanded Google Adwords account:

• Standard: 100 campaigns / Expanded: 100
• Standard: 2000 ad groups per campaign / Expanded: 10,000
• Standard: 2000 keywords per ad group / Expanded: 5,000
• Standard: 300 display ads per ad group (including image ads) / Expanded: 300
• Standard: 50 text ads per ad group / Expanded: 50
• Standard: About 50,000 keywords per account / Expanded: 1 million

That said, what do you do with a campaign?  There are many different ways to skin this cat, but unless you really need help with organization, campaigns exist only for one reason: to change settings at the campaign level that cannot be changed at a lower level.  The question that begs is, what are those settings?  Here are the important ones:

• Search targeting vs. Content targeting (as well as Google’s search network)
• Day parting: bidding more or less on certain days of week and/or hours of day (now easier than ever using Google AdWords’ hourly conversion data)
• Device type: computer vs. mobile
• Standard or Accelerated ad deliver: how fast you want your budget to deplete throughout the day
• Geo-targeting: by city, state, country, radius of a point, or custom geographic shape
• Bid type: manual bids, or Google’s conversion optimizer (that will adjust your bids based off conversion data)
• Budget: unless you are a brand advertiser or have other constraints, I presume you want to spend as much as possible if you are converting clicks to leads or sales at a certain rate

Thus, if you have a group of keywords that needs to have anything changed together for one of these settings, that means they belong together in a campaign.  Do you have a business unit that needs to have higher bids during the day because you have a call center that is open?  You will need to day part those keywords, and that means it goes in a campaign.  Do you want to target mobile devices separately from computers (which you damn well should be doing)?  You will need separate campaigns for that as well.  Want to target different states for different business units or goals (i.e. if you are targeting Chrysler car dealers in certain states or cities and Nissan in others)?  Those will need different campaigns as well.

As you can see from the plethora of options available in the campaign settings, these are valuable marketing tools!  You only get 100 campaigns, so use them wisely.  If you want to group certain themes of keywords together, that is what ad groups are for.  Don’t waste your campaigns on grouping for reasons extraneous to the need to change settings.  Otherwise, you might be out of campaigns before you know it!