Next-Generation Personalization

In 2003, several contributors from Amazon wrote a compelling industry report on “ Recommendations, Item to Item Collaborative Filtering.”  While Amazon has long set the standard for product recommendations, personalization and doing it at scale, no one could have imagined that data would outgrow humans’ ability to slice-and-dice and generate insights from it.  

If you haven’t read this report, I’d highly recommend you look through it so you can see how the thinking from 12 years ago differs from today’s.   It used to be about the algorithm and how to scale it.  Now, if you read all the marketing hype, there’s a black box with magical algorithms that will do all the thinking for you.

With that bit of dark humor out of the way, the best advice I can give you is both a practical viewpoint, but also a challenge to dig into this for your business, I’m a firm believer this kind of personalization will be standard operating practice. To be relevant, you better know what you are talking about. 



First, recognize that the first-generation Web personalization and product recommendations engines were like first-generation CRM systems.  These products required incredibly talented quant minds and IT resources to create -- and even then, justifying the return was a difficult proposition. 

Today, while storage, analytics and personalization tools are better, faster and more intuitive, there are many more options to choose from, depending on what issues you're trying to handle.   Do you know what those options are?   Can you use the personalization engine from your content management system?  Do you need a specialized Web personalization engine?  How do you incorporate rules and algorithms with your email platform? How do you deploy to mobile experiences?   Don’t believe the “blackbox” myth -- it will never be distilled to a true/false or multiple-choice question. It’s a crossword puzzle! You just need to decide where to start: vertical or horizontal?

Second, recognize that humans will never be able to mine all the data and generate insight fast enough. You will continually be looking at the past -- and, regardless of how good you are at running predictive analytics on past behaviors, you will never catch up with today.  But technology alone will not make you better than your competition.  How you become smarter, faster and more agile in this process is the real trick to making it work for you.

Thirdly, balance your aspirations with a dose of realism.  This isn’t easy. All the stats you’ll read in marketing materials predict a 60% improvement by deploying personalization. Your boss reads those same stats -- and, like Amazon and Free Shipping, it can quickly turn into a business rule vs. an incentive.  So understand what you can really do today.   Do you need to be real-time?  What can you actually measure? Do you have the ability to produce the content fast enough?  What channels do you need to deploy (email, mobile, site)?

Lastly, analytics are “must-have,”  according to this video.  (Warning! You need a bit of coffee to watch this parody.)

The point I’m trying to make is, if you don’t have a clear path to metrics, analytics and reporting, you will have a problem.   If you think understanding what’s working is hard today, it becomes dimensionally harder as you take this on.  

Whether you are trying to build the most adaptive, fluid site/search and product recommendations capability or simply trying to better personalize email, it starts with understanding what you really want to do, matched with the capabilities and elasticity of your company.

As author Fritz R.S. Dressler said, “Predicting the future is easy. It’s trying to figure out what’s going on now that’s hard.”

Next story loading loading..