Case Study: Boosting Display Campaign Traffic Quality

By , Associate Director of Paid Search at Hanapin Marketing


Last week PPC Hero’s own Eric Couch detailed how to make display campaigns with higher quality traffic using category exclusions. On the back of his post I thought I’d set up a case study in one of my accounts and actually try to make higher quality display campaigns. How did I get on you ask? Read on…


What I did:


I won’t rehash old ground with a complete run down of how to set up category exclusions, but the essential premise of my test was to see if I could improve the quality of my display campaigns by appearing only on ‘Ad Planner 1000‘ sites and ‘Above the Fold‘.


Excluded Categories


My hypothesis was that first, ad planner 1000 would represent a higher quality audience (not strictly true), and that above the fold would mean that my impressions were more valuable even if my ads remained un-clicked.


For the record, selecting these two exclusions also forces a number of others on you (because they apply to both):


Excluded Categories2


I also made sure to replicate the settings in my other display campaigns – there was no significant geographical, max CPC, text vs image ad changes in how this campaign was set up. Doing any of those things will mean you aren’t conducting a fair test. As my goal is to generate phone calls, my key metrics for determining success come from the Analytics tab in AdWords: Pages/Visit, Avg. Visit Duration and Bounce Rate. You can read a little more about using these stats to judge your experiments in a previous blog post.


So far this test has been running for two weeks and I thought I’d share a few of the results with you.


Campaign Stats


The takeaway from this so far are that of all my campaigns, the high quality experiment currently has the joint best Pages/Visit (1.43), the third best visit duration (25s) and the third best bounce rate (84.25%). Overall, there doesn’t seem to be a huge shift in quality, but it is noticeable that the traffic is slightly better.


I pulled this data into a quick pivot table to analyze the impact of my HQ display when compared with my regular display and mobile campaigns.


HQ vs Regular vs Mobile

Green = winner, Yellow = 2nd, Red = last


As you can see, the HQ campaign is faring well compared with the average of my other display campaigns – the only metric in which it isn’t best is the Avg. Time on Site and that is mainly because one or two visitors heavily skewed the numbers for the Mobile Display campaigns.


What we learned:


  • High Quality display campaigns are slightly better than the average – although I am a little data limited in this conclusion.
  • If you are severely budget limited with your display, you should definitely try this approach.
  • You should test creating a high quality, category-excluded campaign in each of your accounts that run Display.
  • If you have given up on Display, give this a go to see if it can generate conversions within your tolerance threshold (CPA).


Alternatives to my approach:


I tried looking at Ad Planner 1000 and above the fold placements, however there are plenty of alternatives in the category exclusions list for you to experiment with. A few to try might be:


  • Error Pages
  • Parked Domains
  • Forums
  • Sexually Suggestive Content (could this be damaging your brand?)


If you are able to set up an experiment, please let us know how it goes in the comments below!

Twitter Facebook LinkedIn Google+ Email Print More

6 thoughts on “Case Study: Boosting Display Campaign Traffic Quality

  1. Danny Mac

    Hey Sam,

    I have been running similar tests also.

    I normally do paid search campaigns but I have one client’s business lacking in search volume. As far as I am aware; this company is the only one of its type in the UK – Hence, no one is searching for it.

    The company does restoration of wooden shutters (American, Louvre, Plantation etc), this saves the customer a small fortune in replacing them. The manufacturers know all the big margins are in ripping out and installing from new. They tell / promote to customer’s restoration is not possible for the reasons above.

    When customers are in the buy funnel the conversion rate in pretty much 100% and they cannot believe such a service exists.

    I have been running search campaigns trying to turn buyers into the restoration route by driving the message in the ad copy and landing page. The bounce rates are high for obvious reasons and of course only a percentage of the customers actually have the shutters to restore. So, I expected low yield but was definitely worth testing.

    I am pretty new to the Display Network (AKA the beast, stupidity tax etc etc). I have been running campaigns for a couple of weeks, deploying various tests on all the targeting options, adgroup’s segmented by ad type/ dimension and target method. Experimented with exclusions, with and without below the fold, parked domains etc.

    I still can’t seem to get any clicks of real value; with time on site above three seconds (I have been monitoring page load times etc).

    The questions I have are:

    What would be more effective, bidding by impressions or CPC and why?

    Would you let Google optimise or control manually in terms of bidding / ad optimisation?

    Would you bother with Text ads? I seem to drive a huge amount of impressions from spammy sties.

    What seems to be the most effective display ads (leader, MPU, Sky etc)?

    How much data should be gathered before looking to optimise?

    How much budget should be allocated to get any reasonable results?

    If you are running an Interest Targeted adgroup as an example, would you use exclusions and negative keywords to filter for refined targeting?

    Any help would be amazing 🙂

    1. Sam OwenSam Owen

      Hi Danny, thanks for commenting!

      Generally I run all of my Display as CPC because attributing a value per click is easier for me to do than by impression.

      If you are spending a lot of time reviewing your ads then bid and optimize yourself – however if you don’t have much time to devote, then I’d leave it with Google. I’d say most of the time I do it manually.

      Text ads can definitely work – if you are worried about spammy sites, use the category exclusions I discuss in this post. You’ll want to test them and see what your response is like – they might not be great quality, but they are often very inexpensive.

      Effective display ads = whatever your data tells you is the best one. The last report I pulled had MPU winning in terms of Conv/Rate. Download an ads report, pull a pivot table with the size of your ads as your row label and clicks, impressions, cost and conversions in your columns and you should find all the data you need.

      How much data = when you have enough to be statistically significant. At least a couple of hundred clicks for bids etc – for ads make sure you are statistically significant –

      How much budget? If your campaigns are profitable – as much as you have, if they aren’t – enough to get the few hundred clicks you need to be sure it isn’t going to work.

      I generally don’t use negative keywords on display – I might layer positive keywords & ICM together though.

      Thanks for the questions!


      1. Danny Mac

        Thanks for taking time out Sam and feeding back. I will take onboard what your said and implement your recommendations.

        I will come back and post some results in a few weeks once I have acquired some good data to work with!

  2. Patrice Albertus

    Hi Sam and thanks for the clue. This really confirm many things on online advertising : CPM model and above the fold are like old rugh advertising thought than don’t perform for conversion.
    I’m curious about hos you been able to know that campaigns are “High Quality Display” and others “Regular Display” on your excel ? Cheers


Leave a Reply

Your email address will not be published. Required fields are marked *