Thread by @ToferC

Next up: Katie Malone @multiarmbandit From Presidential campaign trail to Enterprise - Building a #DataScience team @CivisAnalytics

Effective #data-driven teams: different in existing organizations than start-ups. Need to find the best places to build in data value without overturning everything else. Challenges: personal-level analytics, resource allocation, security, privacy. #ODSCWest

Business & data teams will see problems differently. Data teams need to be comfortable with trade-offs in business. Metrics can be ill-defined. Speed can be better than perfection. Important to understand adequate accuracy of models. #ODSCWest Oh yeah, and security.

For business teams: it can take a long time to deal with messy or poor quality data. Algorithms (today) can be easy to interpret or powerfully predictive (#NeuralNetwork). Not both. Your data teams will need help to ensure that models developed are used. #ODSCWest

On that last point: Model deployment is getting really big. Keeping it and maintaining models & data takes time and effort. (See @karpathy notes from earlier today). #AIDevOps #ODSCWest

So think about:
- Who to target
- What to say
- Who to talk to
- Where to say it
- Did it work: what are your outcomes?
- Foundational technology & curated data: lots of resources required for good work

#ODSCWest

Good overview of what a #DataScience team might look like #ODSCWest

And what skills these folks need. #DataScience #Teams #ODSCWest

@threadreaderapp unroll