JohnTL;DR: Factually, I have no idea. In my humble opinion, it depends. At times, there may be a need for both.
Full transparency, most of my data career has been in a centralized team or a group that manages most data-related requests in a specific way. Having said that, I have some assumptions of pros and cons of working in decentralized teams.
Pros and Cons of a Centralized Team
As stated previously, most of my career was (and still) in a centralized team of some fashion. The types of teams I’m thinking are ones that are developing data products for the entire company and teams that handle data requests from all areas of the company.
Pro: Control of Data Ecosystem
When your team manages most of the data going in and out of a data product, there’s a high probability that you’ll have the upper hand on most decisions for the ecosystem. By having control, you’ll have transparency on any issue or proposal that can dramatically alter the ecosystem. You can also make decisions that is best from a data perspective.
Pro: Less Likelihood of Duplicated Work
With a centralized team, tasks should be vetted before being approved for work. In the event a new task comes down the pipeline that has already been done or has some similarity to previous work, it can be flagged as a duplicate and tossed out. Having a pinch-point for tasks can make it easier to avoid duplicate work.
Pro: Have People to Share Work With
Having a team or a colleague that has some familiarity with your work is a huge benefit. In the event that you need to take a vacation because of burnout, have too much work, or just don’t want to work on something; it can be easier to pass on the work to someone else. The opposite is possible too – maybe you take on all of part of the work from your team member.
Con: Many Different Types of Contexts
When your team is the only data team in the company or the only team doing a particular ‘thing’, there’s a high chance that you’ll be working on tasks from many different domains. Context changing across multiple departments can be challenging and honestly inefficient when it comes to completing work. The mini-ramp ups that you do to remind yourself of what you were doing may seem to take a little time, but multiply that across the number of subjects you have – it can be a decent amount of time used for the workday.
Con: Less Time to Learn About New Things
Again, with the amount of workload coming down the pipeline, there may not be as much time for professional development or time to “think outside the box”. Often times, if an architecture pattern solution works, it’s implemented over and over in order to accomplish tasks. Sometimes it’s hard to find that time to try and do something new that can provide more value or efficiency.
Con: Slower to Complete Tasks
When managing a big data landscape, there tends to be vague requirements from internal customers or unforeseen obstacles when trying to finish tasks. You’ll run into issues such as customers not knowing what they really want until they see something they don’t want. The other thing could be when you’re working and you find something can’t be done that was already promised. In any event, there tends to be more hurdles that need to be cleared prior to finishing the work.
Pros and Cons of Decentralized Teams
Stated before, I don’t have much experience in this realm, but I can imagine about some of the good and bad.
Pro: More Focus on Less Topics
In contrast to being on a team that “does everything”, being on team that has a luxury of pure focus on one topic or a couple of topics makes it easy stay focused on the work. This also allows one to build that muscle memory for the topic which makes it quicker to recollect most things that are related to the subject. Unless you have an incredible functioning brain, it can be difficult to perform well when you need to context switch between an infinite amount of subjects.
Pro: Easier to Expand on Current or Previous Work
Extending off the previous point, because you build that muscle memory, it can be easier to make changes to the work you’re currently working on or any previous work you’ve done. The ramp up time won’t be as much if you need a little reminder on how things work.
Pro: Complete Things Faster
Again, extending on the first point, more focus and less topic changing can make you more efficient. This should lead to faster task completion.
Con: High-Likelihood of Duplicated Work
Because you’re focused on what you’re doing, you may be doing the same thing as someone else on a different team. In actuality, it may be that you have different reasons for the outcome, but the work itself may be similar.
Con: Lack of Standards
Standards, love ’em and hate ’em; but they’re necessary. Smaller teams may be in the mindset of “getting things done” than trying to stick to best practices. Why would I need standards and worry about how all of this may fit together some day if it’s just me?
Con: Getting Help
Though you may have the benefit of focus, there will be times that there can be a heavy workload or you just need time of. As it is with having more people on a centralized team that has familiarity with your work, you may not have that benefit on a smaller decentralized team. Often times, that work will have to be on pause until you get back or have the time to work on it again. By some miracle that you find someone that has free time for you to explain all the intricacies of your data, would it still really be worth it? I often find that it’s best to hold off.
Conclusion
Taking from the title, there are pros and cons to centralized vs decentralized teams. Ultimately, I think that it depends on the work. There may be times where a centralized team makes sense because there’s a need for standards and management. There may be times when a decentralized team makes sense because there’s no need for a lot of people and focus is important. There may be times when both are needed because there is a need for structure and also a need to have a team to get things done now.
