Successful data/ analytics team structure - Centralized, Distributed or Hybrid?

Recently, I have enjoyed many of the discussions with peers and leaders in the data/ analytics space. One of the topics that I have been asked to share my thoughts and experiences on often was the structure of successful data and analytics teams. This is a sizable topic but, I hope to touch on a few of the key aspects here. So how should one structure a successful data/ analytics team - Centralized, Distributed and Hybrid? Jumping a bit ahead I recommend the hybrid structure. Read on to see why.

Before I share the details of why I made that recommendation let me recap some of the groundwork of the terms that I will be using. When I talk about a successful data/ analytics team I am referring to the inclusion of all the folks focused on Data, Analytics AND Strategy (https://www.dhirubhai.net/pulse/tripod-structure-successful-data-team-amit-shivpuja/?trackingId=jcY7WmncRDSE5h%2FR%2BQJ%2FMg%3D%3D) and I will be touching on the ?ingredients from my sandwich framework too (https://www.dhirubhai.net/pulse/need-structure-data-decisions-use-my-sandwich-amit-shivpuja/?published=t), specifically with a little more focus on the people aspect. So, if you haven’t read these yet, I recommend a quick read through (each is less than 5 minutes long).

Now, diving into the three different team structures. Starting with a centralized structure, where all the team members focused on Data, Analytics and Strategy come under the same hierarchy. Stakeholders need to approach only one organization. This “centralized” approach enables more control/ standardization of policies and leveraging common best practices. It provides clearer visibility of career paths/ growth for team members in the hierarchy and allows easier movement of team members between projects with minimal impact on stakeholders.

The flexibility offered by the ability to plug and play resources also aligns nicely in organizations that are culturally very hierarchy heavy. There is also deeper collaboration (flow of information) between team members from data, analytics, and strategy as they are part of the same hierarchy. The major disadvantage is that a “distance” gets established between the stakeholders and the data, analytics, and strategy team. They are perceived often as “outsiders” to the stakeholder teams. Next, let’s consider the distributed data, analytics, and strategy team structure.

It is in many ways the inverse of the above stated centralized structure. In this structure team members of all their focus areas data, analytics and strategy are embedded directly with stakeholder teams. For example, a marketing team would have its own data, analytics, and strategy folks. They report into the stakeholder team hierarchy. This puts these team members literally in the “business”. They have deeper knowledge, understand the ins, outs, focuses and stakeholder team direction first-hand. This approach feels natural in more flatter organizations that we see today in tech but, also in organizations that are more consultative.

While there is almost no “distance” here with stakeholders it creates almost a hyper-specialization effect where it costs a lot more to move team members to other groups given their deep knowledge and can feel limiting on their career pathways. There is also a risk that this can lead to redundant silos of approaches, technology stacks, best practices, datastores etc. These silos, though tactically sound from the perspective of stakeholder org, are, from the enterprise perspective, detrimental in the long term as they limit the flow of information across stakeholder orgs.

So, we land on the hybrid approach. In this approach the aim is to select the best of each of the above models and overcome the challenges as well. The approach I have leveraged and seen successfully executed is to centralize some of the functions and distribute the rest through a well-defined process of formal handovers, documentation, and governance. So, in the hybrid approach it is recommended that the foundational components like data, tech stack, and strategy be a centralized set of teams. These in turn report to the same hierarchy. The analytics (analytics, data science, AI/ML) teams can be distributed i.e., part of the stakeholder teams. Here we have an interesting choice based on org culture. If the culture is more hierarchical then the analytics teams though embedded (dotted line) in stakeholder orgs can hierarchically report to central data/ analytics leadership. However, if the org culture is more flat then they can report into the stakeholder org leadership.

This approach reduces the distance to stakeholders, avoids redundancies, leverages the same tech stack, and enables a more strategic thought process around data. However, for the effective flow of information and collaboration between the centralized teams and the distributed resources, a clear way of working must be established around mature data governance principles. An enterprise needs only one governance framework and would gain the benefits of a common data layer/ eco system. Distance between stakeholders and analytics is minimized. While there are quite a few aspects to the hybrid structure one of the most under-managed aspects that enables the hybrid approach to succeed is documentation.

Documentation is non-negotiable, with requests between say analytics and data being made in writing with work starting only after all the needed info is collated and shared (formal handover). This means that not only the stakeholders’ needs, data verification/ business rules but also the business context must be captured and passed to data teams. In turn the data team should share detailed documentation of the data built, validations performed, and meta data generated. This along with documented way of working, governance documentation and strategy roadmaps would enable the movement of team members, quick ramp up of new members and a more consistent ecosystem where work gets done. Focus can now be on enabling the insights needed.

Hope this helps introduce and answer some of the questions around how to structure data/ analytics teams. Reach out if you have any questions and look forward to your experiences.