Thursday, July 20, 2017

Are we in a Big Data hype bubble?

Who are you going to believe? Newsweek or your own lying eyes?

Based on past experience, an industry or economic trend becomes a market bubble when the media offers up credulous stories based on their sources in the stock hyping business. Said stories, to the reader willing to look around, seem to conflict with observed reality and direct experience.

Reminder: Bubbles always depend on ignoring Conquest's Law #1:

Everyone is conservative about what he knows best
I'm beginning to wonder if Big Data hype is now at the bubble stage. This realization came while reading this piece in Newsweek:

The quote that broght me up short:

Data make a company’s machine-learning software get smarter so that the company can better serve customers and vacuum up more market share. Think of Amazon’s recommendation engine.
Now, Keven Maney may be right about many things. Apple may be headed for a fall. Amazon may be a master at using data to drive profits. (Or maybe not.)

But what I know to be categorically false from personal experience is that Amazon, Netflix, etc. are using data to improve their recommendations and "better serve their customers."

As long-time Amazon customer I've found their recommendations to be less useful and less accurate over the past several years.

Netflix is even worse.

So put me in the skeptical camp when it comes to Big Data.

1 comment:

Doubting Richard said...

You are right about Amazon. I love reading, almost never read anything except on Kindle, yet it takes me at least 30 minutes to find a new book I want to read when I have finished my latest book or series. They are not the worst though.

The worst is YouTube, owned by the one company that should know a lot about us. The video recommendations are either simply the latest plus a random few old ones from a subset of my subscriptions or videos I've already watched. No apparent curating whatever. They cannot even use the information they collect to exclude the videos I've already seen, or put them in a separate "watch again" category. I don't know what these companies do with the information, but it is not personalise recommendations.