Jul 06 2009
Walking a Fine Line With Web Analytics Data
Sending email based on web analytics data is a topic we’ve already covered on this site in some detail. As most email marketers know by now, web analytics data can be used in abandoned shopping cart campaigns (ReMarketing), browser behavior campaigns, and other targeted campaigns. But to me, the most important aspect of these campaigns is how the data is being used and what the net effect is on the subscriber base.
Here’s my real-life example that illustrates the issues surrounding web analytics/email integration:
I am a frequent shopper of the online retailer Amazon.com, where I have purchased books, electronics, and baby items exclusively. Last week, however, I received an extremely targeted and somewhat perplexing email promoting Amazon.com’s variety of gourmet and specialty meat selections. Wild Boar 10 Rib Rack anyone?, asked the email. The first line of the message was “As someone who has shown an interest in gourmet meat…”
What? Gourmet meat? Me? From Amazon.com? Huh?
Then I realized what had happened. About three weeks ago I was doing a search on how to best cook beer can chicken. After re-tracing my steps, I recalled landing on an Amazon.com meat page, clicking around about five times, and leaving. I do not think this makes me a lover of online gourmet meat…but maybe it does?
Items to consider:
- Personalization was not used in the subject line. If Amazon knew I clicked on a handful of pages, surely the company knows my name, right? Would name personalization make this message more or less intrusive given that it is based on browse behavior?
- How do we differentiate a user’s definite “interest” in a particular subject area versus just casual browsing?
- The timing of the message was smart and relevant – this is prime BBQ season, after all. Had it been sent in the dead of Winter, I would have found the message less useful.
Takeaways:
- Browse behavior messages should be targeted, but not so targeted that users are looking over their shoulders to see who is watching them.
- Marketers need to be selective in choosing which pieces of behavioral data are used in an email.
- Be sure to carefully define what “showing an interest” means. Long browse durations, cart abandonments, repeat visits, etc. all seem to be strong indications of user interest rather than casual browsing.
I’d like to hear from you. Can an email be overly targeted? Where should email marketers draw the line?
Very interesting post. Email marketers should target their customers but not to the extent that some do. There should be a fine line drawn between showing an interest and coming off as a stalker.
I think the stalker concern is a valid one, but what ruffles people’s feathers the most is when they feel a label has been misapplied to them. It’s always jarring to see how other people perceive us, no matter what the context.
There’s also the risk that the friend of a PETA activist could have borrowed their computer and triggered the e-mail, unwittingly precipitating a cyber-jihad against amazon.com. Assuming there isn’t one already ongoing.
I think you make a good point that products browsed on a site are often better used as content that compliments a more generic main message, such as cooking in your example. Whereas the main message in the email you received was about a specific type of cookbook. However, in testing many types of programs using web behavior, typically the larger portion of targeted customers were browsing items that they were interested in versus products they reach by accident.
I don’t agree with your first takeaway. I think that companies should go deeper into browser behaviors to get an even more accurate segment to which to market. In your example, the email list Amazon used was obviously not targeted enough. They probably just looked at visitors who browsed any meat category page just one time. However, if they would have gone deeper into the data to segment visitors who looked at multiple meat category pages, did so multiple times in the past 30 days, and added a meat item to their cart, then the relevance to the recipients would have been much higher (and conversions would have been much higher), and you wouldn’t have made the list at all.
Deep behavioral analysis is the best way to differentiate “interested” from “casual browser.” Web analytics isn’t enough, and usually results in your scenario. Since there’s limited data and limited ability to slice and dice it, companies are forced into taking the shotgun approach and hoping that their 10,000 emails hit a few hundred actual targets.
Jason Rushin
Quantivo
Thanks for the reply Jason. Yes, it does seem that the shotgun approach was utilized in this case. A more targeted segment would have been better if that data was available.