Amazon Driven by Data

Possibly this belongs on my datablog and I might write it up there later (or I might not).

The New York Times has published a piece on working conditions in Amazon. It does not make for comfortable reading and certainly does not paint Amazon as a company which I would want to work for. But I did want to pick up on one comment in it.

“Amazon is driven by data,” said Ms. Pearce, who now runs her own Seattle software company, which is well stocked with ex-Amazonians.

That comment about driven by data is something that worries me. Sure, data is very sexy at the moment in the tech but data in itself can be meaningless and what matters is the information you can derive from it. But the information you look for is generally skewed by humans who decide what questions they want to ask of that data. Data isn’t the answer and how it is queried is not benign or completely rationally independent.

The average tenure in Amazon, apparently, is one year. You can argue that this might be as a result of a toxic culture. You can argue that Jeff Bezos is a genius. But to my uncertain knowledge, Amazon does not turn a profit and I think, has never turned a profit. We have an ongoing assumption that this is somehow okay, that it’s a new paradigm, and times have changed. With an average tenure of a year, you have a company which cannot possibly have stability in output quality, and you have a company whose knowledge base never develops.

Amazon’s search is atrocious and its recommender system has deteriorated badly in my experience. A hypothesis for why that might be is that they appear to have rapid staff turnover and by definition, limited continuity.

Many of the datapoints in Amazon’s evaluation system appear not to be datapoints at all. They are entries in the AnyTime Feedback Tool:

Ms. Willet’s co-workers strafed her through the Anytime Feedback Tool, the widget in the company directory that allows employees to send praise or criticism about colleagues to management.

However, many workers called it a river of intrigue and scheming. They described making quiet pacts with colleagues to bury the same person at once, or to praise one another lavishly.

Anyone who implements a tool like this either a) expects it to get gamed and considers that a value or b) doesn’t expect it to get gamed in a high octane organisation is naive at best and making decisions on potentially faulty data. Either way, they are unlikely to wind up with the best staff. Such organisations, however, can only operate on the comfort blanket of never admitting this.

Amazon can call itself a data driven organisation, and if it wants to measure everything down to the nth degree, they should be bright enough to know the limitations of what they are doing. Going by the content of the NYT’s article, they probably aren’t.