Until Google gives us the option to exclude prior buyers on PPC campaigns (are you listening, Google?), I don’t think there is any way to optimize along the new customer / current customer vector. It just is what it is.

That said, you can still use the dis-Engagement idea to predict which campaigns will be the most profitable *in the future* and move spending around before the final results are in. I talk about that in detail here:

http://blog.jimnovo.com/2007/04/07/engagement-campaigns/

Basically, you can predict which campaigns or keywords are generating propects that are more highly engaged than those coming from other campaigns and keywords. This is essentially what Cisco is talking about, looking at levels of engagement based on behavior.

And yes, over the longer term, you optimize for relative LTV or what some are now calling “NTV”, or Near Term Value - which is essentially the Tactical LTV or “breakeven” idea I talk about here:

http://blog.jimnovo.com/2008/01/30/strategic-tactical-lifetime-value/

Multichannel Merchant just published an article with a similar approach:

http://multichannelmerchant.com/crosschannel/lists/article_35/

Hope that helps!

My question on engagement: For PPC optimisation does it make sense to optimise against anything other than reach?
Obviously we can’t simply optimise against conversion as this would probably include existing customers, but what about optimising against new orders?
What about optimising for typical ‘lifetime’ value? For example, if users who visit for the first time from keyword X, normally buys product Y and Z. Using an engagement formula you could be as sophisticated as your formulae and your back end permits.

At the human-driven level (ACT!), you have a salesperson receiving a new lead, contacting the prospect, and then putting in a trigger event themselves, like “Call this person in 1 week”. That results in a ping or message to call this prospect a week later, and the engagement process continues in that was, with the salesperson reacting to the needs of the prospect in a customized way.

At the data-driven level, where you can’t afford a salesperson for every account or customer there is intelligence that runs in the background (or an analyst) that is looking at behavior patterns and says, “If the prospect has not called back within a week, they have an 70% chance of not ever calling back, so trigger an outbound call or mail piece.”

When you can’t afford to have salespeople making these timing decisions, you use the patterns of past behavior. This is what Relationship Marketing looks like in B2C.

By the tone of the piece, I’d guess this effort (and the stats) are heavily influenced by them entering the SMB market. What they are trying to do is replicate the monitoring and interactivity of a salesperson who would normally service large accounts with these smaller accounts where they “couldn’t possibly hire enough sales people to deal with that”.

That’s really the idea behind Relationship Marketing - you are trying to create an “automated salesperson” for each customer based on their behavior. You do this because the customer is not valuable enough to support a live salesperson - as is the case in most of B2C, and the SMB portion of B2B.

The “to close” percentage increase I would question. Not so much in the numbers themselves but what that *means*.
ie. Does it mean they’re selling lots of low-end cheap stuff with corresponding razor thin margins, or more of the high-end scarily expensive kit? Margins to drive a Porsche through?

I ask as my long term experience[1] with Cisco has been of them being a highly responsive organisation - down to pretty much having mobile phone access to their local tech’s/sales support team 24×7. Their sales people were generally only involved at the “name a price” point. It was all techs talking with techs.
But then, we were, regularly, buying their … scarily expensive kit. And lots of it.
:-)

I don’t think too many people would be buying several million worth of high end routers and switches over a website. YMMV naturally. ;-)
Research? Yes.

Cheers!
- Steve
[1] And now dated - haven’t been on that side for oh… 6 years ish.