Most mid-sized companies do not have a data team problem. They have a few analysts, one or two engineers, and a long list of requests from finance, operations, sales and the plant floor. Where those people sit decides how fast answers arrive and whether two managers walk into a meeting with different numbers.
The structure question comes before the hiring question. If you hire first, you tend to hire for the wrong shape of team.
Three ways to organize a small data group
Central. Everyone reports into one group that serves the whole company. Definitions, security and quality are controlled in one place. The cost is a queue. Every dashboard request, integration and AI pilot competes for the same few people.
Embedded. Analysts sit inside finance, operations, sales or production. They learn the business quickly and answer fast. The cost is drift. Each function builds its own version of revenue, margin, scrap rate or headcount, and nobody can reconcile them.
Hub and spoke. A small core owns data platforms, definitions, access and quality. Analysts sit in the functions but report, at least in part, to the core. It asks more of your managers, but it is the shape most growing employers end up in.
Signs your current setup is the problem
Before posting a requisition, check for these:
- A request for a simple report takes weeks, and the business has started building spreadsheets on the side.
- Two departments report different numbers for the same measure.
- Analysts spend more time clarifying what was asked than building it.
- Your best analyst is a single point of failure, and nobody else can maintain their work.
- AI or automation pilots stall because nobody owns data quality.
The first two usually point to the wrong structure. The last three usually point to a missing role or a missing owner.
What to hire for each model
The structure tells you which profile to look for.
Central team. You need people who can say no, set priorities and write down definitions. A data lead with real stakeholder skills matters more than another technical specialist. Look for someone who has run an intake process and can explain a trade-off to a controller or a plant manager.
Embedded analysts. Domain knowledge carries more weight than tool depth. A finance analyst who understands month-end close, or an operations analyst who understands throughput, downtime and labour hours, will be useful sooner than a generalist with a longer tool list. In manufacturing, someone who has stood on the floor and knows how the numbers are captured is worth a great deal.
Hub and spoke. You need two kinds of people, and they are different. The core wants a data engineer or architect who cares about standards. The spokes want business-minded analysts who will follow those standards without being policed. Hire the core owner first, even if it is one person, or the spokes will set the rules by default.
Contract or permanent
Many employers get stuck here. A few rules of thumb:
- Use contract talent to clear a backlog, build a first set of shared definitions, or stand up a platform migration. The work has an end date.
- Hire permanently for the person who owns definitions and quality. That knowledge needs to stay in the building.
- Be cautious about embedding a contractor in a function that has no one to hand the work to. When they leave, so does the context.
Reporting lines matter more than org charts
In practice, an embedded analyst answers to whoever sets their priorities and reviews their pay. If the business manager does both, the analyst serves the department and the standards slip. If the central lead does both, the analyst becomes a ticket handler again.
A workable split is day-to-day priorities from the business, standards and career path from the data lead. Say that plainly in the job posting and in the interview. Candidates with experience in split reporting will recognize it, and those who want one clear boss will self-select out.
Retention follows structure
Data people leave when they are isolated or when they are buried in requests. A lone embedded analyst has no peer to learn from. A central team in permanent backlog burns out. Hub and spoke helps on both fronts if the core runs regular working sessions, shared code reviews and a clear path from analyst to lead.
That is a cost-of-turnover argument as much as a design one. Replacing an analyst who holds two years of context is slow, whatever the market looks like.
What to do next
Before you open a requisition, write down who needs which numbers, who owns each definition, and who decides priority when two departments ask for the same person. If you cannot answer those three questions, fix the structure first, then define the role. If you already know the shape and need to fill the core owner, the embedded analyst or a contract engineer to clear the backlog, a Traverse recruiter can help you scope the role and find the right person in Canada or the US.