HBR article, Beware hyper-personalization

[joe.pistell]

Uncle Joe Rule #118: "Just because it's a good idea, doesn't mean it'll work".


Why Websites Still Can’t Predict Exactly What You Want


Netflix is rolling back it's giant personalization efforts.

 

[yagoparamo]

Uncle Joe Rule #118: "Just because it's a good idea, doesn't mean it'll work".


Why Websites Still Can’t Predict Exactly What You Want


Netflix is rolling back it's giant personalization efforts.

We are more complicated than an algorithm (well... at least you are!)

Netflix didn't work for me because my wife watches movies, my kids watch movies, etc so the suggestions are usually Ben Affleck and Disney movies.

So a question to ask is; are we recognizable enough just by seeing our browsing history?

 

[JessicaRuth]

I'm with you on the Netflix suggestions - Pretty sure "Jessica" doesn't want to watch Star Wars the Clone Wars.

I'm not sure I'd go with browser history - kind of the same scenario for Netflix, if basing on a shared computer.

 

[yagoparamo]

I'm with you on the Netflix suggestions - Pretty sure "Jessica" doesn't want to watch Star Wars the Clone Wars.

I'm not sure I'd go with browser history - kind of the same scenario for Netflix, if basing on a shared computer.

LOL

the must have paid a lot for The Clone Wars because it is all over my account!

 

[Chris Cachor]

I've been working with some personalization software and it's pretty cool stuff. It's a better way, especially for car sites, to determine related vehicles than they do today. More so for used cars than new.

This is an area of software that's just taking off, so I expect things to improve by leaps and bounds over the next few years. It's also worth mentioning that each service has their own algorithms, so if Netflix is toning down their personalization efforts, it doesn't mean Amazon, Beats, or Pandora is.

 

[joe.pistell]

.... so if Netflix is toning down their personalization efforts, it doesn't mean Amazon, Beats, or Pandora is.

Netflix's 100+ algos failed because they were pointed at the wrong target. Chose the wrong target and you'll never win.

 

[yagoparamo]

Netflix's 100+ algos failed because they were pointed at the wrong target. Chose the wrong target and you'll never win.

Joe,

But big part of the argument against it is; how can you chose the right target in this type of scenario?

 

[joe.pistell]

Joe,

But big part of the argument against it is; how can you chose the right target in this type of scenario?

Netflix Algo's failed because leadership failed to understand their users FIRST. In other words, they built a space ship to Mars when their customers really wanted was help to get to grandma's house.

This is why I'm pounding the table on being very very cautious about all of the spins that'll come from "big data". Big data is the ultimate play toy for mega-geeks and marketing loves to spin the crap that circles around it. From a leadership POV, you can't let the promise of big data over shoot the gaps in your customers needs.



Uncle Joe Rule #41: "Just because it's a good idea, doesn't mean it'll work".

 

[PaulS]

Website personalization is an area I believe is significantly underused on dealer websites.


Firstly, I agree with Joe's point that your number 1 design consideration in implementing any personalization in an automotive website (or indeed any ecommerce site), is to understand how it will benefit your online visitors.


I think at the root of a lot of the criticism personalization receives is because of the complex way it's been implemented. If it's trying to do something inherently complex like predicting vehicles/products that a customer might be interested in, then the risk of this backfiring is going to increase. We understand what the input variables are (the current shopper's actions and other shopper's actions) and what the output is (a list of suggested vehicles) but we've no insight into the "algorithm black box" that comes up with this list, and this makes us suspicious.


However, personalization does not need to involve a complex algorithm. There are several "not so big data" ways it can be used to good effect and at lower cost and risk. For example, if a user has arrived from an Adwords campaign "cheap cars in <your area>", then it's a good bet that the shopper is budget constrained and your website can then highlight cheaper inventory to them and emphasize your finance offers. I wrote an article on my blog a few months back with a list of personalization ideas for dealer sites.


As with most website enhancements, implementing personalization should be an iterative process rather than set and forget. The feature should be tracked with your analytics tool of choice and different variations of it could be A/B tested to see which gives the best results.

 

[yagoparamo]

Paul,

Great article,

But all your examples include a person that is already looking for a particular task (buy cheap, finance, research, etc). All those are or can be available in current systems as they have been integrated in most VDPs over the last few years.

It seems that humans are experts at using the web for the most part nowadays (I don't believe anymore in dumb sites like some dealers want to have sometimes) and that people with a task in mind just find it and complete it.

So, wouldn't that customer just find that piece that he is looking for (calculator, incentive, etc) and just use it? Maybe not?