The Data Science Maturity Curve is a tool developed by our partners at Bardess Group. I recently interviewed Daniel Parton, Lead Data Scientist with Bardess, for more insight about each stage of the curve.
Data Science Maturity Curve – Phase Three – Functional
After organizations move through ‘learning’ about data science (phase one), through the experimentation of the ‘emerging’ phase (two) – they’ll land in phase three, the ‘functional’ stage. At this point, data science is typically more visible across an organization, with more resources and staff.
This is where companies have a more mature data science practice, with a dedicated team or data scientists embedded into business units, explained Daniel Parton, lead data scientist with Bardess Group. But while staffing might be more clear in this phase, technology is often nascent.
“At this stage, although you have dedicated data scientists, there may still not be dedicated IT or data architecture available to those data scientists. So you may still have the data scientists mostly doing their work on laptops…and if they’re using a cloud cluster they might be doing the admin themselves. That becomes increasingly difficult to scale.”
This is also frustrating to the team, delaying impact and leading them to make their own homegrown, not-necessarily-secure, solutions. This is the stage where “do it yourself” tech tends to become problematic, Parton said.
“One of the risks here is that the data scientists end up spending half their time just managing data architecture, managing a compute cluster and managing their software environment on their laptop.”
This can mean that valuable data experts are spending their time on lower-value activities, and can’t focus strategically. While data science on a laptop may be feasible for some ad hoc discovery work, most organizations will need more infrastructure to scale and successfully mature their programs.
That’s because the next transition — from “functional” (phase 3) to “integrated” (phase 4) — is critical and risky, Parton said. Many companies will get a team in place, make some progress and then stagnate after a couple of years. They may question the value of these often-expensive teams or wonder if it’s worth continuing the program.
“In a lot of cases, that’s due to either failure to integrate the team from a technical perspective, failure to provide them with good data architecture, or not providing good ways to productionize their results and their insights. And the other side, of course, is the cultural and management aspect.”
Watch the video below for additional insight and then read more about this transition from Phase 3 to Phase 4 in the next section of this series.
Watch this Data Science Maturity Curve interview excerpt to hear Daniel share more about “Phase Three – Functional” – and hear more about the biggest risks when building a new data science team.