17 September 2026 · 8 min
Data migration support that is not a weekend
A cutover without a rehearsal is a fundraiser for contractors. We map sources to a target grain, reconcile in slices, keep the old system readable, and only switch the report when the hashes match. Public clients need that for audit. Private clients need it so Friday's orders are still there on Monday.
Luke Clark's migration work with us was the unglamorous kind: Oracle, large sets, a catalogue so the new platform did not inherit the old mess. That is still the job.
Gathering business insights without another island
Business analytics fails when process and data disagree. We sit with operations, draw the flow as it is - not as the SOP says - and pin each step to a table. Order to cash, referral to outcome, claim to payment. Then the metric is a step in that flow, not a slide.
A local public body wanted a 'single view of demand'. Three case systems, three definitions of open. We did not start in Power BI. We started with the process, agreed 'open', and only then built the model. Private manufacturers do the same with WIP. The software is interchangeable. The definition is not.
What clients actually feed back
They do not thank us for a colour. They thank us when the number in the room matches the number in the warehouse, when a pipeline failure is a morning ticket not a silent month, and when the bill for Fabric or BigQuery stops being a surprise. Prasad I's note on Python and tough projects, Geoff Jacks on unsupervised complex work - that is the feedback we keep: the work got done, and it did not fall over when we left the call.
Related service: Cloud data