23 May 2026 · 4 min read
Test the data, not just the code
Freshness, row count, uniqueness, and referential integrity checks on every run catch the majority of quality incidents before a human sees a wrong number.
A failing test should stop the pipeline, not warn quietly in a log nobody reads.
Define metrics once
When two dashboards disagree, the cause is usually two definitions of the same metric. A semantic layer with one definition per metric removes the argument permanently.
Document the definition next to the metric, in language a non-analyst can check.
Make lineage visible
When a number looks wrong, the first question is where it came from. Column-level lineage turns a half-day investigation into a two-minute one.
Ownership matters as much as lineage. Every dataset needs a named owner.
Where to go from here
If you are weighing data platform work for an upcoming project, the useful next step is usually a short scoping conversation rather than a longer article. Bring your constraints — budget, timeline, existing systems, and the team you already have — and the decision usually resolves itself in an hour.
We run these as a fixed-scope discovery session with a written recommendation at the end, including the cases where our advice is that you do not need us.
Working on something like this?
Send us the brief. We will tell you what we would build, what we would cut, and roughly what it costs.
Talk to us