I’ll try powering through this one. I just returned to work from a 2 week vacation and my eyes feel like I’ve been rubbing tabasco sauce in them all day. You should try it (the vacation, not tabasco).

It’s funny how being away from work for an extended period can motivate you to actually think a little more clearly. I spent time in the Pacific northwest (San Fran, Seattle, Vancouver) and had a really nice trip. I didn’t do anything outside of the ordinary (for me), and thinking about how predictable I can be it only made sense that I’d start thinking about using data to better predict customer or potential customer behavior to influence (or guide) them. Retargeting messages have proven to be very successful if you do it the right way.

So, what’s the right way? There are a few basic rules (like don’t be intrusive, do lots of testing, know your content, know your customer), but it’s pretty much different for everyone. The basics, but good starting points, are geo-targeted (based on their geographical location), demographic and past engagement with your company (do they open, do they click, do they buy). Treat them all differently. Tone is the key to predictive targeting. If someone shows mild interest (open or click, but no purchase), don’t swoop in for the hard sell. You’ll likely lose them forever. Instead, don’t be afraid to ask them what they were looking for or try finding out by dangling some other options. Think Amazon, not used car salesman.

Again, this all boils down to how good your data is. You’d be surprised at the amount of information someone will give you if you use it responsibly and relevantly. But, don’t forget about the data that you probably already have. Those engagement metrics mean a lot. Think of it as a gauge to how they interact (or don’t interact) with you and message them differently based on past behavior and where you’d like them to go. If you segment the data based on these basic rules it makes it interesting seeing people move from bucket to bucket and how much power you have to actually influence behavior if it’s done correctly.

Please post some comments about predictive messaging or modeling. I’m curious to know what others have done.

so will your program.

I’ve been very fortunate to have worked for a few companies that had their data figured out. We’ve either had too much (never a bad thing) or just enough. With too few facts about your customers it’s nearly impossible to strike up a meaningful, relevant conversation. The same applies to your email program. And with email it’s all about EOI (Expressions of Interest) these days. Positive ROI has been and always will be part of email, but to genuinely strike interest with your customers you have to be relevant… and timely.

Neither are easy to accomplish, but if you have a well thought-out, scalable data set, you’re half way there.

Over the next weeks we’ll walk through some best practices as far as what types of data are key to creating a successful email program. I’m interested to hear what challenges you’re facing with your current program and any steps you’ve taken to fix or enhance your current system.