Hi, thanks for sharing. Quite interesting. You emphasize analysts autonomy. But don’t you think there’s a similar risk as with the self-service BI? At some point all those actions without governance, standard practices (like CI/CD) knowledge sharing might backfire (and be dumped to the “not needed” engineers?)?
We had the same concern when going self service at Nickel, Michal.
After 6 years, 200 bigquery users and 25 data analysts who put algorithms in production I can tell you it’s the other way around.
The more people try to get some figures, the more challenged they are and the most confident you get.
Same for data analysts with algorithms.
What was essential not to get data chaos (that we also feared) was to have a good governance. A central team animating the community to define rules. And most importantly tools that by themselves ensure these rules.
If to deploy the use of a function you need to have a review in a pull request workflow you ensure this confidence.
Hi, thanks for sharing. Quite interesting. You emphasize analysts autonomy. But don’t you think there’s a similar risk as with the self-service BI? At some point all those actions without governance, standard practices (like CI/CD) knowledge sharing might backfire (and be dumped to the “not needed” engineers?)?
We had the same concern when going self service at Nickel, Michal.
After 6 years, 200 bigquery users and 25 data analysts who put algorithms in production I can tell you it’s the other way around.
The more people try to get some figures, the more challenged they are and the most confident you get.
Same for data analysts with algorithms.
What was essential not to get data chaos (that we also feared) was to have a good governance. A central team animating the community to define rules. And most importantly tools that by themselves ensure these rules.
If to deploy the use of a function you need to have a review in a pull request workflow you ensure this confidence.