The real causes, not the generic list. Each one with the specific thing we do differently, so you can hold any vendor to it, including us.
01 · Built to a vendor template that never fit the sources
02 · No schema validation, so errors corrupt reports silently
03 · Nobody owns data integrity after launch
04 · Scoped like an enterprise project the business cannot sustain
05 · The manual-grind trap, and why it persists
Everybody wants the answers. What stops mid-market companies is not usually distrust of a lake they already have. It is that they do not have one, the one they have is broken, or getting anything out of it means one person dumping data by hand.
Five causes, five counters, in the engineer's words.
The headings are above. The diagnosis under each one, and the specific thing we do differently, open here as soon as you tell us where to send the guide.
Built to a vendor template that never fit the sources.
The pattern arrives before anyone has read the schemas. It fits the vendor's last three projects and not this one, and the mismatch surfaces months later as reports nobody can reconcile.
counter · [ Jason to supply ]No schema validation, so errors corrupt reports silently.
Nothing fails loudly. A column changes meaning upstream, the pipeline keeps running, and six months of reporting is quietly wrong before anyone notices.
counter · [ Jason to supply ]Nobody owns data integrity after launch.
The build team leaves. Integrity becomes everyone's job, which means it is nobody's, and the lake degrades one unnoticed change at a time.
counter · [ Jason to supply ]Scoped like an enterprise project the business cannot sustain.
An eighteen-month programme with an enterprise operating model, sold to a company with no data team to run it afterwards. It does not fail at launch; it fails at handover.
counter · [ Jason to supply ]The manual grind persists because the real thing feels risky.
You have heard the stories, so the build stays on the someday list. Meanwhile one person keeps exporting by hand, and that cost never appears on any budget line.
counter · [ Jason to supply ]A data lake does not have to become the story you have heard. The failures have causes, and causes have counters.