Skip to content
Case studies

Work, measured in outcomes

Representative engagements, anonymised for confidentiality — client names and references are available on request, on a call. Every study names the metric the system was accountable to.

Custom softwareDedicated teamLogistics

A freight operator replaced twelve spreadsheets and three legacy tools with one quoting and dispatch platform

Mid-size freight company, ~200 employees. Quoting took hours and lived in inboxes. The platform unified quoting, dispatch and invoicing into one source of truth, in production from week six of a two-year dedicated-team engagement.

Read the full study →
4h → 6min
median quote time
7-figure
annual operating savings
AI solutionsPer projectLegal services

A law firm first-passes every commercial contract through a review model tuned on its own playbook

A 40-lawyer practice. Lawyers now review flagged clauses instead of reading page one to signature, with confidence-based routing to human counsel. Delivered in four milestones over five months, on the signed estimate.

Read the full study →
−82%
first-pass review time
< 1 year
payback on project cost
AutomationPer projectRetail · multi-store

A retail group automated order-to-cash across four systems and eleven stores

Fulfilment ran on exports, email and heroic effort. Storefront, warehouse, accounting and logistics now share one automated flow; humans handle exceptions only — a small fraction of order volume.

Read the full study →
months
payback, not years
3.5 FTE
capacity redeployed
Partnership · equityHealth tech

A clinical founder with distribution and no product — built under a hybrid equity deal

Reduced price plus equity. We built the patient-intake platform; she filled it with clinics. The company has since raised institutional funding — with SigmaJunction still on the cap table.

Read the full study →
100+
clinics on the platform
funded
institutional round closed
AI solutionsDedicated teamEnergy · trading

An energy trading desk embedded demand forecasts directly into its scheduling tools

The previous forecasting dashboard was ignored for a year. Rebuilt inside the scheduling workflow — with error ranges shown honestly — the forecast now informs every dispatch decision, and the override rate falls month over month.

Read the full study →
−4pt
forecast error vs. prior method
daily
use by the trading desk
AutomationPer projectFinancial services

An investment firm automated client onboarding and compliance document flows

KYC document collection, verification routing and ledger setup ran across email and four systems. Now one flow with human checkpoints on flagged cases — onboarding measured in days instead of weeks, with a full audit trail.

Read the full study →
weeks → days
client onboarding time
100%
audit-trail coverage
Custom softwareDedicated teamiGaming & betting

A casino & sportsbook operator ran its in-play book on a trading-desk engine we built

Live odds management built like a trading floor: real-time exposure dashboards per market, automated risk limits with trader overrides, and sub-second bet settlement under peak-event load — serving hundreds of thousands of end users. Responsible-gaming controls and full audit trails designed in from milestone one.

Read the full study →
<300ms
bet-to-settlement latency
100k+
end users served
AutomationCustom softwareOil & gas

An oil & gas operator unified terminal scheduling, trade capture and field reporting

Cargo schedules lived in email threads; trade positions in spreadsheets; field data arrived days late. One operations platform now carries scheduling, trade capture and daily field reporting — with position and exposure visible the moment a deal or delivery changes.

Read the full study →
days → hours
field-to-report cycle
live
position & exposure view
Custom softwarePer projectConsumer · publishing

A publisher of personalised books turned manual design into a compose-to-doorstep platform

Every order used to buy hours of a designer's time. Customers now compose a personalised book — from children's stories to serious titles — with a live preview; the platform renders a print-grade file and routes it through integrated print-shop APIs, so a bound copy arrives at the customer's door with no manual step in between.

Read the full study →
minutes
from order to print-ready file
to the door
printed & shipped via print-shop APIs
AI solutionsDedicated teamIn progress

A company is moving its complete AI workloads onto custom on-prem hardware

Metered cloud APIs had become a structural dependency: linear costs, external data boundary, someone else's capacity decisions. We specified custom inference hardware and are migrating workload by workload — each benchmarked against its cloud baseline before cutover.

Read the full study →
ongoing
migration, in production
in-house
where the data now stays

How every study gets its numbers

The metric is agreed in the estimate phase, measured before the build as a baseline, and measured again in production. No before-number, no claim.

Baseline first

We measure the process as-is — hours, error rates, cycle times — before a line of code is written.

Measured in production

The after-numbers come from the running system's own telemetry, not from a survey or an estimate.

Verifiable on request

Client references will confirm the numbers on a call — that's part of every engagement agreement.

Want the unredacted versions?

Client names, architectures and real numbers — shared on a call, with references.