Selected work

The numbers, the method, and what we got wrong.

Clients are described by sector, scale and region rather than named — but every figure below is real and traceable to the engagement. Where we publish a forward projection, we label it as one. Where the work hit a ceiling, we say so.

01
Consumer & Beverage · Commercial analytics · 6 weeks · Full write-up
Turning foot traffic into take-home revenue
185 products, 22 months and $2.57M of transactions — reconciled, normalised for footfall, and turned into a tap list that pays for itself.
Take-home sales lift when a product is on tap
2.6×
Take-home sales lift when a product is on tap
Products analysed across 22 months
185
Products analysed across 22 months
Revenue placed under measurement
$2.57M
Revenue placed under measurement
Read the case study
02
Public Sector & Real Estate · Geospatial data science · 14 weeks · 1 min read
638,000 parcels, ranked down to 100 they could actually buy
Every parcel in the state scored on the same 45 measures, so that “which property should we buy next?” finally had an answer that did not depend on who walked in the door.
Parcels scored
638,184
Parcels scored
Ownership records reconciled
249,034
Ownership records reconciled
Measures per parcel
45
Measures per parcel
Read the case study
03
Public Sector · Private AI · 4 weeks · 1 min read
637,000 sensitive records, searchable in plain English — on their own hardware
The obvious tool was off the table, because sending government property and owner records to somebody else’s server was never going to be approved.
Records made searchable
637,358
Records made searchable
Sent to any outside service
Zero bytes
Sent to any outside service
From no system to a working one
4 weeks
From no system to a working one
Read the case study
04
Healthcare · Clinical machine learning · 10 weeks · 1 min read
1.5 million clinical readings, reduced to eight numbers a monitor already has
Sepsis kills faster the longer it goes unnoticed, and the standard bedside scores catch it late. The question was whether the machines already in the room hold enough signal.
Hourly readings processed
1,552,210
Hourly readings processed
Clinical measures assessed
43
Clinical measures assessed
Model types compared
6
Model types compared
Read the case study
05
Healthcare policy · Modelling · 3 weeks · 1 min read
540 model runs for a 0.42% gain. We told them to stop
The job was predicting how long patients stay in hospital. The honest finding was that the data had nothing more to give.
Model versions tested
540
Model versions tested
Improvement over the untuned baseline
+0.42%
Improvement over the untuned baseline
Countries in the panel
32
Countries in the panel
Read the case study

Next step

Find the first workflow worth automating

Bring one process, inbox, document type or bottleneck. We’ll tell you whether it is worth building — and what a first pilot would look like.

What happens next

You pick the workflow. We ask what already exists, who touches it, and where it stalls.
We reply within one business day with a rough shape and a rough timeline — usually smaller than people expect.
If it makes sense, the next step is a working system in two to six weeks, tested by your own staff.
You keep your existing tools. Private deployment is available where data cannot leave.
And if it is not worth building, we say so instead of selling you one.

The numbers, the method, and what we got wrong.

Turning foot traffic into take-home revenue

638,000 parcels, ranked down to 100 they could actually buy

637,000 sensitive records, searchable in plain English — on their own hardware

1.5 million clinical readings, reduced to eight numbers a monitor already has

540 model runs for a 0.42% gain. We told them to stop

Find the first workflow worth automating