Thread by @jessitron

85% of data projects fail.
Yikes! How can we do better?

@jessetanderson at #YowLondon

“AI is not a strategy.” @jessetanderson #YowLondon

You need about ten data engineers and one data scientist.
also an operations team.

@jessetanderson #YowLondon

That vendor may have told you “Use my product and you’ll have AI in a day”
No. You need a realistic timeframe that is NOT “when it’s ready.”

@jessetanderson #YowLondon

Focus! Fight a few battles well, not many battles poorly.

10 teams doing 20 things at once => “Nothing. Nothing happens.”

@jessetanderson #YowLondon

Find the overlap between what is technically possible (in the timeframe)
and what has the highest business value.

Look for a 10x return (because sometimes it’ll be negative).

@jessetanderson #YowLondon

Installing the tech won’t get you the skills. Hiring people with the skills won’t change your organization to work with them.

@jessetanderson #YowLondon

“Data Mesh is a sociotechnical strategy” @jessetanderson #YowLondon

“Put a hard thing in people’s hands without skills, sets them up to fail”

Data Mesh represents a significant amount of organizational change.
What type of company is it applicable for?

- large orgs with complexity.

Data Mesh for startups? Don’t.

@jessetanderson #YowLondon

“Who of you have a notebook in production?”

… some hands …

“Yeah. That’s terrible. I’m sorry.”

@jessetanderson #YowLondon

Data scientists learn just enough programming — a single-player game.

Software engineering is multiplayer.
The tactics are different.

eg: in math, you don’t care what a variable is named. “a” is fine.

@jessetanderson #YowLondon

A data team needs a relationship with their customer.

“The data warehouse team didn’t work with their customer. They just said no”

@jessetanderson #YowLondon

(I’m throwing him off by laughing really loud)