From TheFutureOf (9 Nov 08): Responding to Steve Jackson's 1 Sept 08 7:56am comment
Steve Jackson's Comment
Jim's post closely correlates to one of the points I made on Eric's forthcoming white paper on engagement. I wanted to know how the engagement formula/model would measure dis-engagement, which is where I and I'm sure Jim believes where the real value of the RF measurement model is.
I may have mentioned before that we were using segments of “engaged” visits by our definition and an RF comparison to determine likelihood to purchase in one particular case in which we saved a client in excess of $2Million.
We defined engagement as x amount of clicks and duration on the website. We explained the notion to the client that the more “recent and frequent” a visitor was the more likely they were to buy. By showing that our engaged segment was more recent and frequent (as well as purchasing slightly more) than global traffic we proved to them that certain traffic driving elements were poorer than others.
This meant the client made direct savings (as I said in the 7 figure bracket) that they wouldn't have ordinarily considered in terms of where (what media) they spent their money.
Since seeing this in practice I've been exploring how to use recency and frequency to predict when a potential customer will “dis-engage” because as I mentioned the recency and frequency of the engaged segment was much higher than global traffic.
What I need to understand in more depth is where and how to define a segment that will predict when a visitor is becoming disengaged so I can then advise clients to do something about it.
At least more than the way I do now which is by simply comparing the differences in RF metrics.
I think the formula you're producing and working on might go some way toward that and combines a lot of the current thinking on the subject I'm wondering if it can be done.
What would a dis-engagement formula look like?
(finally, some time to catch up on things)
Funny that I'm getting to this now. The answer (if I didn't explain before) is encapsulated in an email response I gave to Debbie Pascoe's “…when you talk about ďengagementĒ and measuring it, what is it that should be measured? I donít mean the variables, etc to collect specific data points. Rather, it has to do with motivation. Someone comes to the site and we want to whether/know they are engagedÖÖengaged to do what? What are the variables measuring? How will we know if/ when weíve answered the question?
Weíre not all the same, so my engagement is your time-waster. I donít know if Iíve made the question clear enough for your response. If not, let me know and I can provide more clarity.”
I should point out that I'm waiting for some other NextStagers to approve what I wrote in response to Debbie before posting it here.
Let me see if I can abridge somewhat…
If the demonstration(s) of (whatever is being defined as) engagement is known then any diminutions in and/or the absence of these demonstrations is an indication of dis-engagement (by the definition of engagement).
IE, the equations being used to determine whatever is being defined as “engagement” must also determine “dis-engagement” by the same definition. If not, then either the definition of engagement is in error or the equations of the determination are in error.
I described to Debbie Pascoe that there exists a set of universals that demonstrate engagement regardless of individual and regardless of situation. In other words, “Are these things happening?” Answer yes, the person's engaged. Answer no, the person's not engaged. Answer yes and here are the values of these universals, know how engaged the person is in the situation they're in. Doesn't matter if the person's on the web, in their car, flying a kite, watching tv, playing music, pick your situation, doesn't matter.
The next time we're all at a conference somewhere ask me to demonstrate “engagement” and I'll be happy to do so. It's a demonstration I've been doing since the late 1980s and it never fails to impress. It's also a great way to make money if people are willing to bet.
You also write “I think the formula you're producing and working on might go some way toward that and combines a lot of the current thinking on the subject I'm wondering if it can be done.
What would a dis-engagement formula look like?”
The formula I provided Eric in the whitepaper can determine dis-engagement in any number of ways. The simplest versions are max-min problems and step-wise analytics (something I noted in my edits to the paper). Slightly more interesting versions (to me) involve partial differentials and some resorting to geometric solutions, and all still using the formula in the whitepaper. In all cases these methods would allow for increasingly refined measurements of dis-engagement (as I believe you're defining it).
Cuius rei demonstrationem mirabilem sane detexi. Hanc marginis exiguitas non caperet. (Something else to talk about at a conference)
(please, somebody laugh. please)