The problem, what we did about it, and what changed.

A QR-based asset register that made the annual audit a one-day job

The challenge

Thousands of plant assets were tracked across spreadsheets and paper registers. Audits meant physically walking the site with a printout, depreciation was recalculated by hand each year, and nobody could answer "where is this unit now" without a phone call.

Our approach

We built a QR-tagged asset register with a React Native field app and a Next.js web dashboard on an Express API. Every asset carries a code; scanning it opens its full history. A dedicated audit mode lets a walker confirm presence and condition offline, syncing when signal returns. Depreciation runs on schedule rather than on request, and an AI assistant answers plain-language questions across the register.

The outcome

Audit cycle

Weeks of manual walking reduced to a single guided pass

Asset lookup

Full history available from one scan, on site

Depreciation

Calculated automatically instead of annually by hand

Monthly KPI reporting rebuilt, and 30% of the manual work removed

The challenge

Monthly performance reporting was assembled by hand from several sources. It arrived late, and figures did not always reconcile between teams — which meant meetings spent arguing about the numbers rather than acting on them.

Our approach

We audited every source, agreed one shared definition per metric with the teams that use them, and rebuilt the pipeline so cleaning and validation happen before anything reaches a report. Recurring outputs were automated, with validation checks that flag anomalies rather than quietly publishing them.

The outcome

25%

Improvement in reporting accuracy

30%

Reduction in manual reporting effort

One source

Shared KPI definitions across teams

Inventory that finally matched what was on the shelf

The challenge

Built independently to solve a pattern we kept meeting in distribution businesses: stock maintained in a spreadsheet updated at the end of each day, discrepancies discovered weeks later, and reorder decisions made on numbers nobody fully trusted.

Our approach

A web-based stock management system with role-based access, movement logging on every transaction, and low-stock thresholds per item. Each change is attributable, so discrepancies can be traced to the moment they appeared instead of being written off.

The outcome

Real time

Stock levels accurate at the moment of query

Full trace

Every movement attributable to a user and time

Reorder alerts

Threshold warnings before a line runs out

Finding the churn signal that was already in the data

The challenge

An independent analysis project on subscription retention: customers leaving at a rate a business can feel but not explain, with the data to understand why already present but never assembled into a single view.

Our approach

We consolidated billing, usage and support history into a single customer timeline, then modelled which behaviours preceded cancellation. The output was not a black box — it was a ranked, readable list of warning signs the account team could act on directly.

The outcome

Cohort view

Retention curves by signup month and plan

Early signals

Behavioural warning signs identified pre-churn

Actionable

Ranked intervention list for the account team