• Alan Mitchell

    Dimensions are a fantastic way to fine-tune campaigns, but generally need a large amount of data to make any insights and actions reliable, as you rightly point out.

    Without making a log or note of optimization dates, it can also be relatively easy to accidentally optimize the same data twice, as I point out in point 4 here http://www.calculatemarketing.com/blog/techniques/5-common-ppc-optimisation-mistakes/. This is especially true with dimensions, with it all too temping to select ‘all time’ as the date range for each and every subsequent analysis.

    Making a log of optimization dates can make your dimension analysis more meaningful and powerful.

    • Gabriel Kwakyi

      Thanks for your insight, Alan!
      I definitely concur – as with any optimization, in order to maintain control, it is imperative to keep track of both what changes were made and when they were made, otherwise you can create problems like the re-optimization one you mentioned.
      Also, to your point of requiring enough data, your unit of time studied must conversely not be too long, since then you can encounter performance changes caused by other independent effects, such as seasonality or even other campaign changes you yourself have done, like adding keywords.
      Personally, I like to consider the trailing week or month as efficient, assuming your campaigns have significant traffic.

      • Nick Paolino

        Can you give tips on the best way to keep track of your optimizations? Its a constant struggle for me and I haven’t yet found a system that works for me

        • Gabriel Kwakyi

          Hi Nick,

          Thankfully Bing Ads has a great tool for keeping track of optimizations called the change history tab. It’s a subtab visible right under the performance line graph and shows you, per day, which changes were made to what entities and by whom. By exporting this report into excel, you can easily create a record of the types of changes made on what days and to which keywords, ads, ad groups, campaigns and accounts, and then by combining this data with an appropriate performance report, you can use a pivot table to see which changes occurred over which days alongside performance data.
          One was is to use day/week/month as your row value, change type as your column (make sure to group them by bid increase/decrease, budget increase, new keyword added, etc. so you can make sense of the changes), metric as values, such as impressions, then slice by the appropriate entity level, such as campaign or switch the slicer and column types according to your preference.

  • http://iampacheco.com/ Duke Tanson

    Really useful tips Gabe. As always, optimisation that includes these dimensions yield great results. One frustrating element of the dimension data is the “unspecified” location date which can sometimes account for circa 5% of clicks and cost. Adwords need to sort that one out in 2014.

    • Gabriel Kwakyi

      Thanks Duke for the encouragement and additional consideration – it is definitely a good point and important to be aware of such limitations of your analysis so that you can factor in a margin of error.

  • Erika McCarthy

    How did you create the in table chart in the sum of conversoins column?

    • Gabriel Kwakyi

      Hey Erika,
      I put the each dataset into a pivot table, created a calculated field for the CPA & conversion rate value fields and included the conversions as a default sum value field. Then I added 3-color conditional formatting for CPA (reverse-colors) & conversion rate and a data bar gradient fill for the conversion column; I used a data bar format for conversions to keep the focus on the conversion rate & CPA patterns but still ascertain the absolute conversion pattern at-a-glance.

  • Boyang Sun

    Hey Gabe thanks for the great article! I have some questions

    1. In Part IV you mentioned to use enough data. You mentioned to run campaigns for 1 month, is this the same case for both small accounts and large accounts?
    2. Also in Part IV, Earned and realized. In what situation would you care more about Earned conversions than realized conversions? I know my campaigns in Google assign attribution on a last click basis. So does that mean my campaigns are using realized conversions? This sounds very familiar to the First/Last click distribution models are they about the same? if So where is google adwords or bing ads can I find and make adjustments to these dimensions?


    • Gabriel Kwakyi

      Thanks for the questions, Boyang!

      1. Generally, the smaller the campaign (i.e. less traffic), the longer period of time period of time you want to use, and vice versa – the larger the campaign, the less data you need, since having more data raises your confidence level. However I’d recommend using at the minimum a full week’s data, since performance almost always varies from day to day, and capturing a full week’s data will give you a full view, un-skewed by a single day.

      2. Yes, you are correct; attributing to last click does mean you’re using a realized conversion model. The last vs first click model most commonly refers to more of a multi-channel system, where you are thinking about conversion attribution across ad types (e.g. SEO, display, social media, email SEM). But it does also apply within a single channel, such as when your search ad earns a conversion along one dimension and realizes it along another.

      One situation in which you would care more about earned vs realized could be analyzing the hour of day dimension. If the click that drove the person to your page happened at 8AM, but the conversion happened at 1PM (the customer shopped around a bit, left the window open and came back after lunch to purchase), then you would want the attribution to be 8AM and you want to show your ads in the best position at 8AM, since that’s when your ad actually earned the sale, not 1PM.
      You can make adjustments to dimensions by navigating into either the campaign or ad group settings and selecting a particular dimension (e.g. day of week), and applying a bid multiplier (e.g. 125% baseline bid for Mondays).

  • Szendy Tibor

    On the third table you mention to bid up on tablet, is there a way to adjust bids just for tablets ? Very nice post !

    • Gabriel Kwakyi

      Thank you Szendy!
      Speaking from the Bing Ads point of view, yes you can adjust bids for tablets independently of smartphones & computers. Just click into the campaign or ad group level settings, open up the device settings (inside of advanced targeting settings), and apply a bid multiplier to either increase or decrease your baseline bid.