One of my colleagues sent me over this short report “”3 Things Are Holding Back Your Analytics, and Technology Isn’t One of Them.
They think the following elements should be considered:
How the analysis teams are structured - they need to report in a way that is understandable (i.e. not be too separate from the business), but at the same time be independent enough to provide unbiassed views.
You need to get the culture right, but to be honest the example given is almost information free.
Applicability of the models - the analytics team needs to create artefacts that can be used by the business.
They go on to recommend building an “”analysis nerve centre” with representation at C-Suite level - I guess where data can speak truth to power, or something. I do like their idea of creating “MVP” insight products, but this is advice is not much different from how one might approach other kinds of business improvements.
I think their advice on how to structure teams is something you can take or leave. I’m reminded of a talk I saw last year at an O’Reilly conference on AI about how to build data science teams. The gist from that was that you can go for a centralised or cross-functional approach.
The part though about jumping from a model to a useful model is critical, and that totally depends on the current capabilities of your organisation, the state of your data, and the lead time to be able to implement a decision.
On reflection my advise would be to think about how quickly you can take any data insight and validate whether it has helped the business. Anything at all, at even a small scale, will help you understand the context that you are working in.