[parpfish]

I think the problem is that people read the analyses that companies share publicly (e.g., in a press release or earnings call), and they assume that that's all they did.

You should assume that any numbers shared publicly are just the tip of a giant-ass iceberg and that there were probably 10x more analyses going on internally that weren't shared.

The thing that ultimately gets shared publicly is whatever avoids using advanced stats or internal jargon; They want a single soundbite, not a scientific paper with a full methods section.

[coredog64]

Story time: Worked at a large multinational, and 6-7 years ago they decided they had to ape into the whole “data science” gold rush. They spent millions of dollars on hardware, software, salaries, and consulting.

After a year with not much to show for it, the VP for the silo starts to put out kudos for the team for break-even revenue impact. However, behind the scenes, the insight they were taking credit for was a common sense idea that had already been in the e-commerce team’s backlog.

Nothing surprises me when a company says that they’ve run the numbers.

[parpfish]

i think that many business get lured into doing data-science because they think that math/stats powers grant borderline-mystical powers and that they can be used to peer into a data warehouse and create ideas/business strategies that no mere mortal (i.e., people with domain expertise in business) could have dreamt up.

But... most businesses aren't that complex, and people can usually come up with really good common sense ideas for how to make improvements.

Data-science is often most effective when it serves less as a visionary idea-maker and more as translator that helps common-sense ideas become real (optimizing values, figuring out the best roll-out strategies, building forecasts).