Credit Where Credit is Due: Demystifying Attribution (from SMX East)
Wednesday, October 6th, 2010The following is a recap from a session at Search Marketing Expo (SMX) East. Follow the conference and session on Twitter.
Cross Channel Attribution
- Unified Tracking: track interaction of the user with all media channels at one place.
- Attribute and Analyze
- Model how different factors contribute to conversion
- Model how factors interact with each other
- Optimize: Allocate media spend based on cross-channel models
Unified Tracking
- On average, there are 4-6 touchpoints across before conversion
- Conversion Data (over 10,000) for several clients
- Identify key variables influencing conversions
- The more time it takes to convert, the less likely a conversion will take place.
- Relative importance of conversions
- 70% time to convert
- 20% first touch source
- 7% search engine
- 3% brand term
What is the chief limitation of revenue attribution?
Conclusions
- First prerequisite: unified tracking program
- Attribution is a partial solution
- Attribution + Algorithmic Optimization is the solution for answering the media mix question
Siddharth Shah, Efficient Frontier
Improving ROI through Web Analytics
Success means different things to different people.
Test A/B pages with bounce rate, but be sure to check conversion rates down the funnel!
3 Tips for Better Attribution
- Look at the whole picture
- Use a fair measure of success
- Know your limits
1. Look at the whole picture
- Window shopping
- Entering the store
- Asking questions
- Trying it on
- Buying it
Look at the whole sales cycle
- Awareness
- Interest
- Consideration
- Purchase
- Retention
Solution: create a weighted metric
Assigning value to behaviors for “assists”
Which is more valuable?
- Identify online behaviors that correlate with conversion
- Analyze trends to assign to corresponding value
- Regression Analysis – built into MS Excel
- Optimize to align with weighted metric
When to use weighted metrics
- Long sales cycles
- Multiple “successes”
- Offline sales
- Efficient campaigns
2. Use a fair measure of success
The problem with “last click”
Tools that are starting to get the message
- Omniture Markeitng Channels
- Google Search Trends
- Google Analytics “first click” hack
- Interim solution: combine first + last click
3. Know your limits
- Correlation does not imply causation
- No crystal ball
- Review regularly
Always be testing!
David Sprinkle, ROI Labs
Demystifying Attribution: Two myths
Myth 1: We need to be perfect in order to do it
You don’t need to have a PhD in Math to get attribution right
The perfect is the enemy of the good. -Voltaire
Evolution of Attribution Management
- Last click
- Even
- Exclusions
- Rules based
- Algorithmic
Myth 2: Isn’t attribution just about moving money around?
Just because you have the same amount of money at the end of the month doesn’t mean you should attribute where it comes from so that next month you can optimize accordingly.
Adam Goldberg, ClearSaleing
Demystifying Attribution
Deep down:
- Consumers are constatntly moving around the purchase path
- Multiple touchpoints could all lead to a conversion
Early on looking at attribution:
- Lots of data, no clear takeaways
- Data would only show certain channels based on system redirects
- Or to get the whole picture, you were looking at a hefty price tag
First steps toward attribution:
- Attribution through an ad serving platform
- Able to track 2 channels and 10 touch points
- Found a 15% overlap in revenue attributed
- Proved the need for a deeper level of understanding
What they needed to make some impact:
- A solution built to track all digital activities
- Easy to implement with strong customer support
- Flexible in its attribution models and accommodated multiple views
- Ability to report back at an actual profit level
In action: affiliate valuation
- Ensure affiliates aren’t over inflating credit
- On last click model, affiliate was converting over 10%
- Affiliate was only ‘introducing’ the transaction 2% of the time
- Action: renegotiate pricing structure
In action: network valuation
- Better evaluate individual network performance (consumer path and strength of performance)
- Layer in segmentation (acquisition, LTV)
- Use data to make more targeted and efficient buys
Mixed Media Modeling
- Weight the incrementality of channel buys
- Assign credit in a conversion path against these findings
- Allows for real time ‘exclusion’ set
- Leads to more accurate judgement on each individual channel’s impact
Danielle Smith, Range Online Media
















