Breaking The Generational Curse: Structure Your Data Team for Success
At this point, it is a generally well-accepted fact that building an in-house data team is a good business move. As more leaders are starting to realize every day, investing in data & analytics teams can not only help to measure a business’ progress and success, it can more broadly harness data to help leadership navigate rapid change and identify new opportunities for growth and operational efficiency.
And yet, many data professionals are hamstrung by the structure of their organizations. Analysts who want to focus on meaningful data work find themselves stuck in the daily grind of always having to build the next dashboard. Brilliant data scientists that should be focusing on the core work of building better products often get stuck in a continuous cycle of responding to ad hoc business requests. These problems, which center around how to effectively navigate and support the larger organization, are more likely to undermine a data team’s potential than a lack of infrastructure and tooling.
So, what’s the best way for data leaders to ensure that their teams are set up for success?
How to find your seat at the table
When building your data team, one of the first questions that arises is where the team should sit within the organization. I typically see a few standard models of structure and reporting, each of which have their own advantages and disadvantages:
- Centralized. In traditional businesses, in-house data teams are usually centralized, reporting up to an executive leader like the Chief Financial Officer, the Chief Data Officer, or even the Chief Executive Officer. In a centralized team, the Head of Data typically owns all of the organization’s data and underlying technologies (e.g. AI/ML engineering,data engineering, analytics, etc.) This model has some advantages in that the data team lead can serve as a right hand to executive leadership, closely connected to the vision and strategy of the business. However, there are some drawbacks to this model as well. Centralized data teams are typically seen as “service teams,” functioning like in-house consultants for the rest of the organization. This exacerbates “service desk” dynamics, causing analysts to constantly react to ad hoc requests from business stakeholders, rather than engage in more meaningful and strategic data work. The physical and logistical separation of the data teams in this model lends itself to transactional relationships, and if the data team is perceived to be unapproachable, the rest of the business may not be inclined to utilize their products and services. And without “rich support” from the executive team, data teams working in the centralized model may find themselves with aspirational buy-in but no ability to execute.
- Embedded. In this model, various business departments hire their own front-line data specialists to support their domains. For example, the sales team might have an analyst tracking their pipeline, and Customer Success might have someone else who specifically works on user behavior and retention rates. This embedded approach can offer some advantages in terms of operational efficiency and allows the data professionals to gain a deep understanding of the needs of business stakeholders. On the flip side, it can also cause analysts to be myopic about their own specific business unit, without cultivating a deeper understanding of the bigger picture, as it relates to the state of the business. It also tends to result in a highly fragmented data stack, since each team is pulling in their own tools and is responsible for maintaining their own assets, and normally necessitates a separate data engineering team to manage data pipelines and infrastructure.
- Hybrid. Many organizations have embraced a hybrid hub-and-spoke structure. In this approach, a centralized team, sometimes called the “Center of Excellence,” is responsible for building tools and establishing analytics and data science processes for the entire organization, but the data organization simultaneously maintains pods to support specific business units. Importantly, these all report up into the same manager or executive. The functional pods provide a data compass for those business teams, and help them adhere to the organization’s larger roadmap. With pods reporting up to the data leader, and all team members functioning as part of the same organization, employees have opportunities for training, mentoring and clear paths to advancement. But they also have dotted line reporting to the leaders of the business units they serve. The hybrid structure is the most innovative, and likely the most effective for future-proofing, since it offers a mix of close collaboration and visibility, along with the high level of control associated with the centralized model. But it is also the most complicated to pull off organizationally. The data team views the various business units as customers, and the pods are structured to provide data products, services and ongoing support at appropriate levels. Pods can consist of any title or role (e.g., data product managers, analysts, data engineers, machine learning and AI specialists, etc.), depending on the businesses’ needs. For companies that have a high level of sophistication in their data stack, and view data as a strategic asset, this model has proven to be the most successful.
Although the orientation and delivery of each of these models differs, for most organizations, the most suitable structure will reflect the way your data team was founded and its current stage of maturity. In businesses that take the centralized approach, there was often a forward-thinking leader who decided at some point to invest in a data function and proactively built a team from the ground up to fulfill it. The embedded approach typically emerges organically out of acute needs as a company grows, rather than as a result of an intentional data strategy. The recommended hybrid approach often evolves from experience and frustration with the limitations of the other two structures — it rarely is the form factor chosen at the outset.
Regardless of which model you choose, it is critical that there is a data leader spearheading the company’s data initiatives.In the hybrid approach, it’s also important for data leaders within pods to have a leadership voice within each functional area.
3 Tips for Data Leaders
For data leaders, it is critical to establish a prominent voice (even if that voice comes through an executive champion). Here are 4 tips to ensure success, once they’ve found their place at the executive table:
- Be intentional about the type of structure you implement, since they all come with tradeoffs. To determine the best course forward, assess what kind of talent you already have internally and build from there. Start planning now, so that you can continue to evolve your organization’s structure in the right direction.
- Think longitudinally about how you should evolve your team structure. When beginning your data journey, it may seem sufficient to just hire and expand in order to meet your immediate needs, but it’s important to also develop a long-term view about the complexities of your business as it evolves.
- Secure support for your data objectives. Data leaders need to not only have a seat at the executive table, they also need to secure support from executive management for the data team’s objectives. Without organizational support in place, the data team won’t be well-positioned for success, no matter how well conceived the structure and plan may be. Identify internal champions who are inherently analytical and might be able to advocate for your team and for the organization’s use of data.
3 Tips for Data Professionals
Looking to the future, all business leaders will need to develop a solid working knowledge of how to use data in order to build consensus with their peers. And in order to execute effectively, organizations will also need to build effective data teams to support the organization’s understanding and use of that data.
For younger data professionals who are attempting to navigate organizational dynamics while developing their own careers, there can be enormous benefits from understanding the nuances of each of these models. Digging into the current structure of your data team can help you to answer vital questions such as:
- How are you perceived and valued within your organization? Does your team’s structure allow data professionals to develop true partnerships with other departments, or are you siloed and perceived as a separate entity from the business teams?
- Do you have the time and support you need to produce broad and meaningful analysis, or are you functioning in a support capacity, being asked to fulfill random requests for numbers?
- How can you encourage your organization to evolve its thinking and processes around data? What would you suggest, in terms of the 1-year, 3-year and 5 year plan?
