Most mid-sized companies do not have a data problem. They have an ownership problem that shows up as a data problem. Sales and finance report different revenue numbers. Operations cannot say which system holds the real inventory count. Nobody can say who approved access to a customer file.
The usual response is to shop for a platform. That is the expensive way to start. Catalogues, lineage tools and analytics workspaces are useful, but they record decisions that people have to make first. If those people are not in place, the tool becomes one more system nobody maintains.
Start with the questions, not the software
Before any purchase, write down what your leaders cannot answer today. Typical examples:
- Where does this number come from, and who last changed it?
- Who is allowed to see employee, customer or supplier records?
- Which report is the one we use when the board or a lender asks?
- What happens when two systems disagree?
If you can answer those in a week with a named person, you are further along than you think. If the answers depend on who you ask, you have found the work.
The roles that make governance real
Governance is not a single hire. It is a small set of responsibilities that have to land on actual people.
| Role | What it covers | Where it often sits |
|---|---|---|
| Data owner | Accountable for a data set and who may use it | A business leader such as the controller or VP Sales |
| Data steward | Keeps definitions, quality and fixes current | Analyst or operations lead |
| Data engineer or architect | Builds and maintains the platform and pipelines | IT or an external team |
In a company of 200 to 2,000 people, owners are almost always existing leaders given clear accountability. Stewards are the hardest to fill, because the work needs someone who understands both the business process and the data. A strong analyst with plant, finance or sales operations experience often beats a pure technologist here.
Where the hiring pressure actually lands
Three gaps come up repeatedly when employers modernize their data.
The builder. A data engineer or analytics engineer who can connect source systems, document them and keep them running. Demand for this skill is steady in both countries, and a candidate who has worked inside your core platform family is worth paying for.
The translator. Someone who sits between finance, operations and IT and turns a vague complaint into a defined data requirement. Controllers and plant managers often say this is the missing person.
The sustainer. After go-live, somebody has to run access reviews, handle classification, and answer questions. Projects fail quietly here when the consultants leave and no one has the job.
Contract or permanent
The split is not complicated if you are honest about what is one-time and what is ongoing.
- Contract suits the initial design, platform build, migration and cleanup. It is bounded work with a finish line, and specialist skills are needed for months, not years.
- Permanent suits stewardship, ownership support and ongoing engineering. This work does not end, and institutional knowledge matters.
A common mistake is hiring contractors for the build and never planning the permanent hand-off. Set the hand-off role, and ideally hire it, before the contract starts. Then the contractor can train the person who will keep the system alive.
Compliance changes the stakes
Employers in Canada work under PIPEDA and, in some provinces, their own privacy laws such as Quebec's Law 25. US employers face a patchwork of state privacy rules plus sector requirements. In both countries, you need to know where personal and sensitive data lives and who can reach it. That is a staffing question as much as a legal one. Someone must own the answer, and it should be written into a job description, not left to goodwill.
A practical sequence
- Name an executive sponsor, usually the CFO, COO or CIO.
- Assign owners to your five most important data sets.
- Hire or designate a steward for each major function.
- Bring in contract engineering for the build, with a defined hand-off date.
- Choose the platform with those people in the room, not before them.
This order feels slower at first. It is usually faster by the end, because you avoid rebuilding what was configured around the wrong assumptions.
What to do next
List the three data questions that cause the most arguments in your management meetings, and write a name beside each one. Where the name is missing, you have a role to define or a hire to make. Decide which parts are project work and which are permanent, then staff each one accordingly.
A Traverse recruiter can help you scope those roles and find data, AI and operations people in Canada and the US who have done this work before.