Sunday, May 17, 2015

Multi-Attribution Explained: The Full Buy Cycle Rewarded

Does Brand seem to get most of the conversions? Can't figure out why Display performs so poorly? Does your client want to move around budgets based on Google Adwords attribution? Does your client wonder why conversions yesterday look so much lower YoY? Each PPC platform has a different attribution model.

Google Adwords Conversions Attribution
Google Adwords has a "last click" model conversion attribution. This means some conversions can be retro-attributed. For example, someone clicks on an ad 5 days ago and converts today the conversion is attributed to the click 5 days ago. Google Adwords also has a 30 day default conversion window so a click happening up to 30 days ago can be credited with a conversion. This means you have to wait 30 days before all the conversions will be allocated to yesterday's total.

Google Adwords conversion window can now be increased to 90 days (The max length for view-through conversion is still 30 days):

Google Analytics Conversion Attribution
Google Adwords default has a "last non-direct" model conversion attribution.

Here is the default model for conversion attribution for both Google Adwords & Google Analytics:

Google Analytics New Attribution Model
Google Analytics now allows you to compare different attribution models to see how credit would be allocated. You can select up to three attribution models at a time and compare the results from each model in the table. In addition to the default models, you can use the Model Comparison Tool to create, save, and apply a custom model that uses the rules you specify. This allows you to tailor models specifically to the set of assumptions you wish to evaluate in your conversion path data.

To create a custom attribution model:
  1. Click on the model drop-down selector, and choose Create new custom model.
  2. Enter a name for your model.
  3. Use the Baseline Model drop-down menu to select the default model you want to use as a starting point for your custom model. The baseline model defines how credit is distributed to touchpoints in the path before the custom credit rules are applied. You can choose LinearFirst InteractionLast InteractionTime Decay, and Position Based as baseline models.
  4. (Optional) Set Lookback Window to On to specify a Lookback Window of 1-90 days.
  5. (Optional) Set Adjust credit for impressions to On to customize how impressions are valued.
  6. (Optional) Set Adjust credit based on user engagement to On to distribute credit proportionally based on engagement metrics.
  7. (Optional) Set Apply custom credit rules to On to define conditions that identify touchpoints in the conversion path according to characteristics such as position (firstlastmiddleassist) and campaign or traffic source type (Campaign,Keyword, and other dimensions). After defining the touchpoints you wish to identify, specify how these touchpoints will be distributed conversion credit, relative to other touchpoints. See the next section for examples of custom credit rules.
  8. Click the Save and Apply button to start using your custom attribution model.
Note that the rules all specify relative credit distribution. 

Kenshoo Conversion Attribution
First click, last click, even distribution between clicks, and U-shaped conversion attribution are available. The main difference is each click in the buy cycle gets a portion of the credit for a conversion. I prefer U-Shaped attribution that gives 40% of the conversion value to the first & last clicks, leaving 20% for the assisting clicks.

Convertro Conversion Attribution
On top of PPC & SEO conversion attribution, Convertro brings in offline conversions including TV, catalog, direct mail and phone calls to see how they influence digital campaigns. You can attribute conversion across multiple devices or multiple channels. Learn what is really driving your conversions and where you should be putting your budget.

The Bottom Line
Regardless of what attribution model you use understand their limitations. Don't compare attribution across multiple platforms because they will never match up. Lastly, stick to one platform to measure success so the delta stays the same as you compare performance over time.

No comments: