Analytics staffing: central team, embedded analysts, or contracted

Reference article · 6 min read

When a new analyst position is funded, or an existing one comes open, someone has to decide where it sits. The choice is usually between three arrangements: a central analytics unit inside IT, the budget office, or a performance office; analysts who report inside program departments; and capacity bought through a contract or borrowed from a shared service. Most agencies decide this one position at a time, under time pressure, which is how they end up with an arrangement nobody designed and nobody can explain.

The three models differ mainly on four things: how fast a question gets answered, whether numbers mean the same thing across the organization, whether good people stay, and what the capacity actually costs once you count supervision, tooling, and vacancy time. What follows compares them on those terms and describes the hybrid most agencies settle into, including the classification problems that tend to undo it.

What a central unit does well, and where it stalls

A central team is the only structure that reliably produces consistent definitions. When the same group builds the headcount measure, the caseload measure, and the response time measure, those measures agree. A central team also makes peer review possible, keeps tooling from sprawling, absorbs the loss of any one person, and gives analysts someone to learn from.

The failure mode is distance. Requests enter a queue, priorities get set by whoever has executive attention, and analysts who do not sit in program meetings miss the context that separates a correct query from a useful answer. Central teams also tend to become report factories if no one protects their time for platform and methods work.

What embedded analysts do well, and where they drift

An analyst who sits in the department hears the problem before it becomes a ticket, knows which fields staff actually fill in, and can say no to a bad question. Responsiveness is the clear advantage, and so is trust.

The costs show up slowly:

• Definition drift. Each department builds its own version of the same measure, and the differences surface in public only when two numbers conflict.
• Isolation. A lone analyst has no one to review their work and no one to cover an absence.
• Absorption. Embedded analysts get pulled into scheduling, procurement paperwork, and recurring reports until little analysis is left.
• No ladder. Departmental positions are often classified as whatever was available, which caps advancement and drives turnover.

What contracted and shared-service capacity is good for

Contracted capacity starts faster than a recruitment, covers peaks, and brings specialized methods such as forecasting, geospatial, or evaluation work that an agency does not need year round. Independence can also be a feature when a program's own results are being evaluated. A shared service from a county, a council of governments, or a state agency sits between the two: cheaper than commercial rates and usually familiar with public sector data, but with its own queue and its own priorities.

The limits are predictable. Knowledge leaves when the contract ends unless you require it not to, hourly cost is higher than salary even before you account for anything else, and procurement lead time can exceed the urgency that prompted the request. Contracted staff also cannot own definitions, because owning a definition means defending it for years.

When you compare cost, compare fully loaded figures on both sides: salary plus benefits, recruitment and vacancy time, supervision, licenses, and training for employees, against rates plus your own contract administration and onboarding for vendors.

The hybrid most agencies land on

A workable split assigns the central team the platform, shared definitions, access management, tooling standards, onboarding, and the hardest methods work. Departments hold embedded analysts who sit in operational meetings. Contracts cover surge, one-time projects, specialized methods, and independent evaluation.

The dotted line between central and embedded is what makes this work, and it has to be real:

• The central team participates in hiring panels for departmental analyst positions and sets minimum qualifications.
• Embedded analysts share a code repository, a data dictionary, and a regular working session.
• Any measure that appears in a public report or a budget document is defined once, centrally, and reused.
• Central staff review departmental work products on a set cadence, not on request.

Classification and career ladders

This is where most hybrids fail. Data work is frequently shoehorned into older programmer, administrative, or management analyst specifications, salary bands compress at the top, and the only way to advance is to become a supervisor. Titles vary by department, which hides true headcount and blocks lateral movement.

Practical steps: build a data analyst series with distinct levels and a senior individual contributor level that does not require supervising staff; write minimum qualifications around demonstrated skills with an equivalency path in place of a specific degree; use one title across departments; and start with HR and civil service early, since reclassification usually takes a full cycle and may have bargaining unit implications. Classification rules, examination requirements, and compensation processes differ by jurisdiction, so confirm the specifics with your HR department and counsel.

Deciding a single position, and buying capacity well

For each new position, ask whether the work is recurring and domain heavy (embed it), cross cutting or platform oriented (centralize it), or time boxed and specialized (contract it). Also ask who will evaluate the person's analytic work, and whether they would be the only analyst in their unit.

When you do buy capacity, require in the contract that work products live in your environment, that code and queries are delivered to your repository, that the agency owns the resulting code and documentation, and that knowledge transfer and exit deliverables are specified rather than promised. Ask who is actually assigned, how replacements are approved, and how the vendor will handle a definition that conflicts with yours. Address applicable data handling requirements, which may include criminal justice, health, education, or tax information rules depending on the data, and confirm the current requirements with the relevant authority. Set a term and a not-to-exceed value so staff augmentation does not quietly become permanent.

HQT works with state and local agencies on data and analytics, including platform work and analytic support arrangements.

Frequently asked questions