Delayed Conversion - Allocating Returns
In an earlier post, I looked at the problems of tracking delayed conversions due to complex offerings, long sales cycles, and offline conversion.
Microsoft’s Ian Thomas has an excellent April 13, 2007 post on methods for allocating return when multiple referrers exist. His offering goes a long way towards helping to understand return with complex offerings and long sales cycles.
He identifies 4 revenue allocation strategies that are currently being used and 2 others that are more complicated but might do a better job of modeling reality:
Current Strategies
‘In visit’ allocation
Last marketing source
First marketing source
Simple shared allocationPossible Strategies
Age-based shared allocation
Age and channel-based shared allocation
The current strategies are fairly self-explanatory. The simple shared allocation justs gives each referral source an average credit. For example, if a $100 sale had four referrals (such as a email campaign, a paid directory, and two pay-per-clicks), each would be credited with producing $25 of revenue.
Where the post really kicks into high gear is when Ian beguns to discuss what is possible by allocating based upon age and/or channel. These strategies involve the estimation of influence curves to model how marketing effect decays over time and varies across channel. For example, you might feel that with your product any impression over 30 days old is worthless, so the curve would be asymptotic at 30 days (approaching zero influence).
While the math is better left for computers, the main notion is that a marketing source gets less credit the further back in time it is. Adding in channel factors would affect the maximum influence each source could have. For example, you might feel a paid directory listing with copious amounts of information has more engagement than a pay-per-click ad. Consequently, you might set immediate influence of the directory higher, or alternatively, have the influence decay slower.
Explore posts in the same categories: web analytics