Last-click attribution leaves a lot of opportunity to grow your return on ad spend (ROAS) on the table with respect to investing in top of the funnel awareness which can drive productivity in the bottom of the funnel in terms of both volume and activity. Growing top of the funnel means more people in the purchase funnel from more branded (38% lift over 4 week period according to comScore) and unbranded searches (47% lift over 4 week period according to comScore), as well as higher quality activity in the form of higher click through rate (CTR) on search results and higher open rates on emails.
http://blog.trueffect.com/are-you-stuck-in-a-last-click-attribution-model
There are definite flaws in a fully last click attribution model. In our affiliate program we are still tracking fully on last click but there are more and more tools out there to help figure out where in the affiliate channel the value is. How can I reward the affiliate who first introduced the customer to us and how can I reward the affiliate who closed the sale? And what is the value of each? And what if our paid search or email or SEO was the introducer and the affiliate was just the closer? How do I place value on that affiliate?
Those are questions we are wrestling with here. We are in the process of gathering a lot more data and information around how the different marketing channels interact and hopefully in the next 6 months we will be able to get a better picture of what is going on.
Quotereward the affiliate who closed the sale
Sometimes what looks like "closing the sale" actually means "intercepting the commission".
Many merchants have promo code boxes in their checkouts that effectively encourage their shoppers to leave and go hunting for a discount ... shoppers who otherwise would have been ready to pay.
The clickstream is then rewritten by a site whose only "contribution" is to rank well for [merchant name + coupon] types of searches.
Not only does the -real- referring affiliate get squeezed out when that happens, or the merchant's own promotions, profit on the sale is reduced because of the coupon.
Clueless couponing is a curse.
Quote from: buckworks on November 13, 2014, 06:11:12 PM
Quotereward the affiliate who closed the sale
Sometimes what looks like "closing the sale" actually means "intercepting the commission".
Many merchants have promo code boxes in their checkouts that effectively encourage their shoppers to leave and go hunting for a discount ... shoppers who otherwise would have been ready to pay.
The clickstream is then rewritten by a site whose only "contribution" is to rank well for [merchant name + coupon] types of searches.
Not only does the -real- referring affiliate get squeezed out when that happens, or the merchant's own promotions, profit on the sale is reduced because of the coupon.
Clueless couponing is a curse.
Amen. But after looking at thousands of transactions worth of clickstream data the instances of a "real" affiliate getting overwritten by a coupon affiliate are surprisingly low. Coupon sites overwriting coupon sites? Happens all the time. Coupon sites stepping on other marketing channels? Happens all the time. But a coupon site overwriting a content based site? Very, very rarely.
if anyone wants an intro we'vw been working with a great guy to develop an attribuution dashboard which does deal with many of these issues, as well as handle ROPO issues in a lot of cases.
I don't make commission from it, or get your data, but it would still be great for me to have other good people working with him because good ideas will filter back into the system and benefit us all eventually.
I would say though that with the best dashboard in the world, you still need to agree a model - we have layers and layers of logic (position, recency, branded/longtail etc) but ultimately modelling is still starting from incomplete assumptions, so its a load better than a really basic version, and it does let you reward first touch affilates at a higher rate than last click ones etc etc, but even on clients with a good 18 months data its still only just starting to get to the point where you can think about using it for econometrics rather than just some assumptive attribution.