A collaboration company wants every team to analyse data freely, and finance needs numbers that reconcile. How do you satisfy both?

The false choice

Treating this as a single dial from "locked down" to "open" produces a compromise that fails both: analysts are slowed and the numbers still disagree, because the constraint is applied uniformly to data that has very different requirements.

The tiering that resolves it

Separate exploration from assertion.

• Governed and certified: a small set of metrics and datasets with owners, contracts, tests, lineage and change control. These are what appears in a board pack or a regulatory filing. Slow to change by design.
• Curated: modelled, documented, reliable data that teams build on. Governed lightly, changeable.
• Raw and exploratory: open access, no guarantees, explicitly marked as unsuitable for reporting.

Analysts work freely in the lower tiers. Anything that leaves the team as an assertion about the business must come from the certified tier.

What makes the boundary hold

• Visible tier markers on every output, so a chart in a deck carries whether it came from certified data. Without this the tiers exist and nothing enforces them.
• A promotion path from exploratory to certified that is genuinely usable — otherwise good work stays in the wrong tier and gets used anyway.
• The certified tier must be good enough to use. If it lacks the dimension someone needs, they will go around it, and the governance is theatre.
• A defined arbiter for when two numbers disagree, so the dispute resolves rather than recurs.

The cultural point

Governance that slows exploration will be bypassed; governance that only applies at the point of assertion will be accepted. Putting the control at the boundary where a number becomes a claim is what makes it both lighter and more effective.