DISQUS

rimmkaufman · 7 months ago

Interesting!

You've discussed different methods for allocating credit, but above don't mention the economic impact of this choice.

We've looked at this issue across our client base of over 100 online retailers, mostly B2C, who in nearly every case take the search-as-direct-marketing vs. the search-as-branding perspective.

Within the population of advertisers we serve, we find that varying the revenue allocation scheme for keywords in multi-step click streams is a second- or third-order effect, because the preponderance of paid click streams we track as are exactly one click long.

Here's a post George Michie did on this last month:

rkgblog: Paid Search Buying Cycle: More thoughts.

A more significant decision is choosing how revenue is allocated across channels, not keywords. This is the case when the advertiser has to decide how how to divvy up credit when, close in time before an order, they have marketing interactions via search, email, print catalog, affiliate, etc with the same consumer. Again, in the studies we've conducted for clients, today a substantial number of online orders are single channel, making the this a smaller impact problem. Likely the multi-channel rates will continue to rise into 2010, which will increase the importance of getting it right.

Two questions:

* Can you give some description of the population of advertisers on which your conclusions are based?

* Can you some sense of the economic impact (sales and profits) of the impact of getting the revenue allocation scheme "right" vs. doing it "wrong?" Would you say this is 2%, 5%, 10%, or what-sized benefit/detriment?

It is great to see folks in the industry thinking carefully about how advertisers should count revenue.

These sorts of discussions raise the bar for the industry on how paid search should be run. Good stuff!

Cheers --

Alan Rimm-Kaufman
rkgblog

Craig Danuloff · 7 months ago

Alan - Thanks for visiting and for your post. I read George's post and commented on it back then - thought provoking and was run just when our dev system was able to allow me to build the first ClickEquations Analyst templates comparing performance between attribution models. There wasn't much data then, and while that very preliminary look showed some of what George and you mention - keywords where there is little impact, it also showed some keywords where there was dramatic impact.

At this point, we have a little more data on those accounts that were running in our dev environment, but the majority of our clients didn't get this upgrade until last week, and we didn't rerun all attibution models historically, so we're just building up a data history to do more comprehensive studies. I promise we will and post real data as soon as it's practical.

My general comments ("last click must die") are admittedly partially philosophical, but even looking at the data from those 'dev' clients now we do see pockets of keywords where the impact is +100% in terms of the revenue allocated to a keyword linear vs last. So even if it turns out to be 5% of the keywords, turning those off without realizing their impact would in my view be unfortunate.

But mainly, the post and opinions were backdrop to our new ability to get and present rich flexible data to learn the truth - any CQ client can now see four different attribution views of their own results, and compare and decide for themselves. So let the analysis begin!

On your cross-channel point, 100% agreement. One step at a time, but stay tuned!

Dave · 7 months ago

Interesting read. But when a campaign is losing money, it's usually those underperforming keywords that get cut first.

The conversion of a keyword is not always constant. It could be seasonal or just a million other things out of your control that affect why people are searching for it.

maybe a TV ad is getting users to search for a specific term that converts well, then that ad is pulled after several months and your conversions drop with just the regular seraches, who knows ...

Craig Danuloff · 7 months ago

Yes, it's a very complicated subject. There's not an easy answer for the reasons in the post, the ones you mention, and those to be covered in the follow-on post. Like many other aspects of online marketing, a very simplistic view has taken hold and there are a lot of decisions being made without any clear understanding that they're being made based on very bad data. Awareness of the issue, and it's complexity, is the first step...