ABSTRACT: Federating data analysts provides all the benefits of both decentralized and centralized models with few of the downsides.

Most companies don’t think much about how to organize data analysts. The number of data analysts evolves without central coordination or forethought. As a result, many business and data leaders have no idea how many full- or part-time data analysts work for the organization, who manages them, or how much money the organization spends on them annually. These organizations often fail to reap the full value of their investments in these precious human resources.

For companies looking to do better, there are three primary models for organizing data analysts: centralization, decentralization, and federation. Only federation provides exceptional value without downsides. 

Decentralization

In a decentralized approach, department heads hire, pay, and manage their own data analysts. The data analysts reside in a department along with their non-data analyst colleagues and take their daily marching orders from the department head. This is by far the most common approach to organizing data analysts. 

Benefits. There are benefits to this approach. The biggest is that the data analysts gain valuable domain knowledge. By being immersed in a department, they attend departmental meetings, develop relationships with departmental colleagues, and learn the business inside and out. This helps them understand the nature of the challenges and opportunities facing the department first-hand. Armed with deep domain knowledge, they can proactively suggest solutions to the department’s issues and work quickly and effectively to meet their needs. 

The other major benefit is political: department heads maintain control of their data destiny. Department heads who have been frustrated by corporate IT bottlenecks are reluctant to cede control of their data analysts to a corporate IT or data team. 

Downsides. The downsides of a decentralized approach are numerous and costly. First, since department heads don’t have deep data analysis experience, they don’t know how to hire the right type of data analyst to meet their needs. Also, most department don’t provide training or coaching for the data analyst and may have unrealistic expectations for what they can deliver. Finally, a data analyst can become isolated in the department if there are no other data analysts with which to collaborate. As a result, they are easy targets for recruiters who can lure them away with positions that offer higher pay or more responsibility.

Centralized Approach 

Benefits. Recognizing the downsides of the decentralized model, executives naturally gravitate to the centralized model. Here, they pull data analysts out of the departments and put them on a central analytics team where they report to the director of analytics. Executives like this approach because it provides economies of scale and greater control over funding for this resource. 

A centralized approach also makes it easier to hire, train, evaluate, and manage data analysts. The director of analytics is usually an experienced data analyst who understands what it takes to succeed in the role. Good directors provide ongoing training for the team and establish collaboration, peer networking, and mentoring opportunities. 

Downsides. There are many downsides to the centralized approach. By pulling data analysts out of departments, they lose the valuable domain knowledge. Now, when they are assigned a project, they need to start from scratch to build up knowledge of the people, processes, and data in that domain. The result is that it takes much longer to deliver a solution which is often is less optimal than one they would have created if they had been immersed in the business. 

Another major downside is political. Department heads must wait in line to obtain analytical services. This is frustrating beyond belief for hard-charging business leaders who need data to run their business. As a result, these executives often work behind the back of the analytics department and hire data analysts to work directly and exclusively with them. 

A potential downside to the centralized approach is when corporate executives get greedy and either reassign analysts to work on non-departmental projects or get stingy and refuse to grow the analyst team, even if department heads are willing to fund the resources. This only feeds into the distrust and suspicion lurking among department heads and will chill relationships for the foreseeable future.

Federation Approaches

The ideal approach to organizing data analysts is to federate them. That gives organizations the best of both decentralized and centralized models with few of the downsides. Here, data analysts sit centrally where they are managed by the director of analytics. But each data analyst is assigned to an individual department and take their day-to-day work orders from the department head, who pays their salary plus a premium to cover the expenses of the director of analytics who manages them. 

There are two variations of the federated model. In the “decentralized coached” model, the data analyst sits in the department but has dotted line responsibility to the director of analytics, who helps hire, evaluate, and manage the analyst. In this variation, the analyst participates in weekly standup meetings and quarterly retreats with all data analysts throughout the organization. Conversely, in the “centralized aligned” model, the data analysts all sit centrally but are assigned to individual departments where they do all their work. The data analysts may spend 2-3 days a week in the department working side by side with colleagues there and the rest of the time at corporate. 

Benefits. The benefits of this approach are a superset of the benefits of both the centralized and decentralized models. The company achieves economies of scale by centralizing the analysts, but the analysts get deep domain knowledge because they work exclusively with a single department. The federated approach makes it easier to hire, manage, and train data analysts and gives them mentoring and career guidance, all of which help increase their loyalty and retention so the data analyst role is not a revolving door. 

Downsides. One potential downside to the federated approach is political. Some department heads may not want to share data analysts with a director of analytics at corporate. If they’ve been burned in the past, they might not trust the federated approach. However, most department heads will recognize the value and view it as a gift they can’t refuse. The only sticking point might be who pays the analyst salary; in most cases, the department head will still pay for a “dedicated” analyst resource even if the analysts sit centrally. To avoid the problematic optics of this approach, some companies centralize analyst salaries. 

Another potential issue is when a department has a unique requirement or timeframe that the corporate team can’t meet. In these situations, the director of analytics will agree to an embedded approach for the department and then monitor the situation to ensure the business needs are being adequately addressed. 

Conclusion 

Federating data analysts makes common sense. It takes the burden of hiring, evaluating, managing, and retaining highly technical business users from the department heads and shares it with an experienced technical manager at corporate. It provides training, collaboration, and mentoring to data analysts, which makes them more productive and more valuable for department heads. It also improves the job satisfaction of data analysts who are more likely to stay at the company for a longer period, reducing the churn and cost of hiring new analysts.

Most companies already practice some form of federation in the data team or elsewhere in the organization so the concept is not new. Most executives and managers will embrace the concept and others will eventually see the benefit once the approach gets widely adopted inside an organization.