SMX West 2017 Day 2 - AdWords Attribution, Drafts & Experiments, and More Attribution

March 23, 2017 Andrew Park

Day 2 of #SMX West is all about updates on the much-discussed topic of attribution and new AdWords features and experiences.

SMX West 2017 Day 2 - AdWords Attribution, Drafts & Experiments, and More Attribution(pictured: Bart Heideman, Global Product Lead, Conversions for @google)

SMX West 2017, Day 2
QuanticMind is on the scene at #SMX West for Day 2, including numerous updates from Google on attribution and new and upcoming tools, along with additional attribution insights from industry professionals.

Attribution In AdWords
Featuring: Bart Heideman and @SeanQuadlin

  • Google suggests that metrics for growth be reconsidered - specifically, the common practice of last-click attribution.
  • Last-click may have worked fine 15-20 years ago when all online purchases were made via desktop, but given changes in the customer journey, particularly the rise of mobile as a device for Shopping, last-click's inherent nature of ignoring all previous activity prior to the final click before purchase effectively ignores the majority of the sales funnel - not an effective approach as customer journeys become even longer and more convoluted.
  • AdWords attribution has six primary options at this time, though of these, it most strongly recommends using data-driven attribution (DDA) - a system that automatically tracks and adjusts to campaign changes. The minimum threshold for using DDA is running campaigns that pull in a minimum of 20,000 clicks and 800 conversions within a 30-day time limit.
  • DDA is a machine learning-based attribution method that attempts to properly credit upper-funnel keywords for their contributions to conversions. Consider these two customer journeys as an example:

    - Search keyword "great tech gifts" -> search keyword "best phone" -> search keyword "Google Pixel" gets 3% conversion rate (CVR)
    - Search keyword "best phone" -> search keyword "Google Pixel" gets 2% conversion rate

    DDA attempts to quantify and assign the additional 1% CVR in the above example to the keyword "great tech gifts."

     
  • As a case study, Google highlighted German online pharmacy medpex.de, which switched to DDA and noted a 29% lift in conversions, 37% higher return on ad spend (ROAS), 95% increase in mobile conversions and 148% increase in mobile ROAS. 
  • For those who can't meet the DDA threshold of clicks and conversions, there are five other options for AdWords attribution, arranged here roughly in order of "conservativeness" to "growth orientation":

                         Last-click -- time decay -- linear -- position-based -- first-click
                         <- Conservative                                                      Growth-oriented -> 
     
  • Time decay-based attribution attributes more weight to most recent clicks, then gives a decreasing amount of credit with the passage of time.
  • Linear attribution equally attributes weight across touchpoints in a customer journey - if a standard customer journey takes an average of 4 clicks to convert, each of the four clicks is credited 0.25 clicks.
  • Position-based attribution credits the last click and the first click with 40% of all activity and equally distributes the remainder of the attribution to any clicks in between.
  • First-click attribution attributes all conversions to the very first click in a journey, which presumably leads to a highly noise-filled set of metrics, and is not recommended for any merchants with a customer journey observed to be of any significant length.
  • To test different attribution models, Google recommends first defining the most important metrics against which to test: conversions (at the same cost-per-action or CPA) or revenue (at the same ROAS).
  • Recommended best practices for testing new attribution types include:
  1. Establish a control period - Avoid peak session times and seasonal highs/lows and focus on a consistent period from which a robust, high-volume data sample can be pulled.
  2. Update attribution model and bids - Note that with attribution model changes, CPAs will also change - Keywords that might have been attributed/segmented differently based on last-click attribution and your new model will have a changed CPA, which means you'll need to adjust your bids accordingly.  
  3. Monitor your test - While running tests, Google recommends avoiding major changes and keeping ads, CPA targets and budgets "steady" to observe long-term effects.
  4. Evaluate your performance - Google recommends focusing on the most important indicators to your business (namely conversions or revenue), rather than focusing on "secondary" indicators such as CPC.
  • Once you have determined the right attribution model for your business, Google recommends aligning ads and landing pages with the click path you've determined. On the understanding that your customers may take an extended journey before making a purchase decision, it may make sense to include landing pages that are inspirational rather than entirely transactional - transitional content such as product reviews may help here.

AdWords Power Tools
Featuring:  @SeanQuadlin, Tris Southey (Product Manager for Bid Optimization), Karen Yao (Product Manager for Campaign Management)

- Drafts & Experiments:

  • Google considers Drafts and Experiments to be a “tool built for practitioners” - presumably meaning day-to-day SEM professionals who regularly run and manage campaigns.
  • Drafts and Experiments begins by creating a draft campaign, then as many experiments as needed. Google recommends focusing on "high-value levers" - testing changes that may lead to high-value outcomes, such as changing landing pages, rather than low-value outcome changes, such as punctuation or capitalization in ad copy.
  • For testing methodology, Google recommends setting up experiments that focus on testing one variable at a time, such as target CPA versus manual CPA. Updating multiple elements over the course of an experiment will make it impossible to tell what difference a single factor makes.
  • For best results, Google recommends attempting to "reach statistical significance as quickly as possible" by using a high-volume campaign (which will yield more testing data and quicker understanding) and a balanced campaign split (50-50, for instance).


- DoubleClick Search and Smart Bidding

  • Google explains the Smart Bidding algorithm as being able to decipher the value in longtail terms by way of machine learning.
  • In order to best use an algorithm like Smart Bidding, Google recommends using as large a keyword portfolio as possible to manage risk across different keywords, as well as to optimize toward your "main goal," keeping in mind that algorithms, being mathematical constructs and not human beings, will be unable to ascertain any subtleties that your or your colleagues may look for.
  • Implement controls and constraints when needed - be aware that certain traditional guidelines may not make practical sense. Examples: Setting minimum bids is "somewhat toxic," since it essentially orders the system to purposely buy poor-quality keywords. Likewise, setting a maximum bid is problematic, since it orders the system to avoid purchasing potentially the best (but most expensive) keywords.

 

- New AdWords Experience

  • The recently-unveiled new AdWords experience is still under construction but was built with two goals: quick assessments of campaign performance, and empowering SEM professionals to quickly take action on those insights.
  • Google assures users that the new experience lets users do the same things they've always done - no campaign changes have been made. Some features that appear to be missing in the early beta are only experiments and have not been omitted intentionally.
  • New enhancements include using the left nav to drill into campaigns, access keywords and locations/ad schedule/device information. Ad schedules have also had a horizontal bar chart + heatmap interface added for quicker visualizations.
  • The Dimensions tab has been replaced with Pre-Defined Reports, which let you generate charts to more-quickly visualize results.
  • New features include Showcase Shopping Ads, which engage users earlier in their purchase journey while they’re discovering what to buy and where to buy it.
  • There are also numerous enhancements added for speed-of-use, including creating campaigns faster, managing extensions more quickly, a streamlined interface that limits the amount of data onscreen and lets users use the "?" key to quickly pull up help, or the "G+T" keys ("Go to") to navigate anywhere in AdWords "within two clicks."

 

For more in-depth session coverage of SMX West 2017, please check our previous coverage:

  •  

About the Author

Andrew Park

Andrew Park is a content marketing manager at QuanticMind. A UC Berkeley graduate and lifelong Bay Area resident, Andrew has done tours of duty in editorial, PR and marketing, and now works with the QuanticMind team to communicate the importance of data science and machine learning in digital advertising.

More Content by Andrew Park
Previous Article
SMX West 2017 Day 3 - Audience Targeting, PPC for Brands, Testing and Optimization
SMX West 2017 Day 3 - Audience Targeting, PPC for Brands, Testing and Optimization

Day 3 of SMX West brings new insights on paid search audience targeting, PPC for brands and SEM testing tac...

Next Article
Winning at Local Search: Using all Available Tools to Beat the Competition
Winning at Local Search: Using all Available Tools to Beat the Competition

Winning at local search means using all available tools to beat the competition and stay ahead of the curve...

Heading to SMX Advanced Seattle?

Get a Starbucks giftcard when you stop by our booth #46.

SAVE YOUR SPOT