Case studies

Problems we were asked to solve.

Each of these is a real engagement, described with the client's permission. Names are used where they agreed to it.

ManufacturingData analytics

Cutting a weekly reporting cycle from 30 hours to 4

Halgrove Industrial produced a weekly production report by hand from six systems. Three analysts spent most of Monday and Tuesday on it, and the numbers still disagreed with finance.

The problem

No agreed definition of a “completed unit”, so each system counted differently.

What we did

Ran a two-week discovery, defined 14 metrics with named owners, then built a scheduled pipeline with reconciliation checks against the finance ledger.

Outcome

Report generated overnight. Analysts moved to exception handling. Discrepancies now surface as alerts instead of arguments.

LogisticsReliability

Six-hour outages down to twenty minutes

Ostend Freight had a dispatch system that failed roughly monthly. Each failure took most of a working day to resolve because nobody was sure which component had actually broken.

The problem

Monitoring measured hosts, not user-visible behaviour. Alerts fired constantly and were ignored.

What we did

Defined four service-level objectives on the dispatch path, rebuilt alerting around them, and introduced a blameless review after each incident.

Outcome

Median time to restore fell to 20 minutes over the following quarter. Alert volume dropped by around 90%.

Public sectorAdvisory

Stopping a rebuild that was not needed

Meridian Health Trust had approved budget for a full replacement of its patient-record integration layer, on the basis that it could not scale.

The problem

Load testing showed the existing system handled four times the current peak.

What we did

Ten-day review of architecture, load profile and vendor proposal. Found the real constraint was a single unindexed query path.

Outcome

The rebuild was cancelled. A three-week remediation delivered the required headroom, releasing most of the budget.

Financial servicesPlatform

Nine teams, one deployment path

Calder & Voss had grown from two teams to nine in eighteen months. Each team had invented its own release process.

The problem

Release lead time varied from two days to six weeks depending on the team.

What we did

Built a service template and a single pipeline, then migrated teams one at a time with pairing sessions.

Outcome

All nine teams on the same path within a quarter. Lead time settled at under a day.

Next step

Have a similar problem?

If you recognise one of these situations, tell us about it. Most engagements start with a short conversation, not a proposal.

Get in touch