WHAT YOU GET
04 deliverables
01A warehouse you can trustModelled to the questions the business actually asks, rather than a copy of every source table with the problems intact.
02Contracts at the edgesSchemas and expectations enforced where data enters, so breakage is caught upstream instead of in a board report.
03Lineage end to endFrom source system to the number on the slide, visible to whoever has to defend it in the meeting.
04Access that passes auditRoles, retention and residency designed with your risk function in week one, not reconstructed for the review.
HOW IT RUNS
One domain at a time, and the old path switched off at the end of each. A migration that leaves two systems running is not finished.
WEEK 01–02Trace the numbers
We follow the figures the business already argues about back to their sources, and find the point where they diverge.
WEEK 03–08Model and contract
Core entities modelled, tests and contracts in CI, and the first domain migrated onto them.
WEEK 09–14Migrate and retire
Consumers moved across, then the legacy path decommissioned with your name on the decision.
WHERE IT LANDS
OUTCOME
One number per question
Ending the meeting where two functions defend different figures for the same week.
OUTCOME
AI becomes possible
Most stalled AI programmes are stalled data programmes. This is the unglamorous prerequisite.
OUTCOME
Audits get shorter
Lineage and access controls designed in, rather than assembled by hand under deadline.
THE SHAPE
DURATIONTypically 10–14 weeks per domain.
TEAMData engineers plus one platform engineer.
DELIVERYRemote, in your warehouse and your cloud. We do not introduce a new vendor to solve this.
PREREQUISITESSource system access and a named owner per domain.
LEADS INTOAI Engineering.
NOT A FIT WHENYou want every table lifted and shifted. That is a migration project, and it will not fix the numbers.
FIX WHAT FEEDS THE MODEL.
Start with the domain the arguments are about. It is usually the one that matters.