We're currently using RTP on our website and blog to recommend content. However, I was wondering if anyone knows of a way to leverage this recommendation engine in our email programs as well? If not, has anyone successfully created a scalable content recommendation engine for email automation? The end goal is to have a few engagement streams with a monthly cast of recommended content. Ideally this would be entirely automated, and the only update would be when we add new content to the algorithm.
Thanks in advance!
Rajiv Kapoor recently posted a great article in the Marketo Whisperer's blog that can be used here: Use “Interest Scoring” to target like a bull
I gave a (complicated) talk at Marketo Summit 2014 on exactly how to build a recommendation engine: Marketo Summit: Taking Dynamic Content to the Next Level (slides 8–23).
I've spent an enormous amount of hours building a content recommendation engine like you're talking about. I ended up building a whole system of operational programs in Marketo that dynamically changed lead fields for the next recommended asset based on Persona Profile, and then Buyer Stage, and then finally based on what content has or hasn't been consumed by the prospect.
Operational programs ran it, with lots of logic. We saw an incredible amount of content consumption increase from it (rough numbers from memory: content consumed per session before: 1.1; content consumed per session after: 1.8), BUT it was an enormous amount of time, thinking, and energy in order to build it. The final set of recommendation engine programs to run was unwieldy, even with making it as user-friendly as possible while building. I don't think it is worth the investment, honestly. But it is a greatly fun project if you don't mind the time and energy investment!
Less overwhelming is profiling programs instead of a content-recommendation engine.
The difference here being putting people into email nurture tracks based on their best fit—not based on content-specific interests, but general profiling buckets. It means you're thinking about three dimensions instead of (# of content pieces you have) dimensions. A recommendation engine chooses which specific asset to assign, whereas profiling programs choose which type of content fits and then skips over previously-consumed content.
Basically, the idea is to categorize people's behavioral actions into different profiles (I like Persona Profiling, Buyer Stage Scoring, and Pain Point/Product Interest Profiling), where you use a Traffic Director to route people to the best-fit program based on their computed profile.
Key to building a Traffic Director is an operational program which has tree logic build in, watching for changes and then routes people based on smart list membership to the best-fit engagement program. (Skip to 24:57 in this year's architecture talk to get explanation of this Master Routing this kind of program.)
And here is more info on how it all works!
I don't mean to just puke content at you—I just happen to be a big fan of this topic.
I am actually working on something similar right now and this is very helpful. It looks like the external links to the decks no longer work though. Do you still have this content?
I do! Here are the new links (looks likes I can't edit that 2-year-old post):