1. Not sure I fully understand this question. I’m not a GA expert and have not played with the new advanced segmentation, so don’t know the capabilities.

That said, the above example was based on Campaign averages, not individuals, so I’m not sure it matters. If individuals in the aggregate can be tagged with a campaign source, and then the average Recency and Frequency calculated, that’s all you need. Don’t know if GA can currently do this at the Campaign level, but they certainly have the info on Recency and Frequency and could do it in the future. The idea of “Relative Camapign Performance in the Future” is certainly important enough they should consider it, if not available now. Please not the camapigns would have to start at roughly the same time for the Recency to be comparable.

2. Sure, there are people doing this, though it’s not common yet and the level people are taking it to varies widely. Just like offline, you can “reverse engineer” or turn inside out any RFM cell and see what the composition of it is. Drivers you will typically see are:

1. Media / Offer, which usually implies channel though does not have to

2. Category / Product, certain Cats / Products are “stars” with high satisfaction / repeat and others basically create low value customers

3. Channel, though clearly the view from corporate and the channel need to be reconciled to be useful, e.g. if a customer is defecting from one channel to another, is that really a defection? Is it OK for one channel to “steal” another channel’s defecting customers? Does the original channel have a “right” to first try to retain the customer? Messy…

There are other primary drivers of behavior, but they are more applicable to longer cycle (LifeCycle) studies rather than RFM / Campaign oriented work.

Thanks for the interesting series of articles. I particularly engaged with the application of RFM models to online activity.

A couple of questions.

Firstly, for those who are using the free analytic tools like GA where PII restrictions means that segments are aggregates of IP addresses/browsers rather than addressable individuals, how are you tying up the web data to the allocation to one of the four segments ?

Secondly, for those that are able to develop models of customer interaction including online activity, are you seeing RFM models augmented by both Channel and Product/Product Area segments to more finely tune these segments? Or is that way beyond most organisation’s abilities currently ?

Regards

Dan Croxen-John

A lot of them? No, probably not, not yet anyway. But many are working on these concepts at some level. Often they lack the time or the tools to go as far as they would like to, but the turnover rate of good web analysts is quite high right now, so I expect folks will get the time and tools pretty soon! The evangelizing conversation leads to a “mix” of culture, people, tools, pain points, and so forth.

Personally, I think the way to do it is to take a core concept and explain it in as many different ways with as many different examples as possible. This allows people to pluck out different parts that make sense to them and rebuild something customized to their own situation, the “mix” of all the above they face. Not too many “one size fits all” kinds of situations out there, with the possible exception of classic DM operations. They basically get much of this behavioral modeling stuff already, though they might not understand how to cross this knowledge over to the web yet.

The real challenge to all of this is the lack of an “analytical culture“, which is why I talk about that so much. There are plenty of people who are willing and able to go down the analytical road, but for various reasons, just cannot get past the roadblocks.

If not, how do you evangelize these concepts with the folks you work with?

Just curious. Keep up the good work!

Kevin.