We move the systems that cannot stop.
Arcwright modernises platforms for regulated industries — core banking, clinical records, grid settlement, scheme administration. The work is unglamorous by design: we run the new system beside the old one until the numbers agree, and we stay on the pager after the handover.
- Founded
- 2014
- Studios
- Manchester, United Kingdom · Oslo, Norway
- Roster
- 8 listed
- Practice
- Enterprise systems consultancy
A rented room, and one ledger nobody wanted to touch
Arcwright began with a single engagement: replacing a 1987 COBOL savings ledger for a regional building society that had lost its maintenance contract and, with it, the last two people who understood the batch schedule. The replacement ran alongside the original for seven months. Nobody was willing to switch the old one off until the two systems had produced identical statements, member by member, through a full year-end.
That parallel run was not a methodology at the time. It was caution. It became the house method because it kept being the thing that turned a terrifying cutover into a boring one — and because a client who can watch both systems agree for a quarter stops asking whether the migration will work and starts asking what to build next.
Eleven years later the shape of the work has not changed much. We are still brought in when a platform has become load-bearing for a business and impossible to change safely: a mainframe with no vendor left, an integration nobody documented, a batch window that has quietly grown past the start of the trading day. We map it, we prove the replacement against it, and we cut over on a weekend that has already been rehearsed four times.
What has changed is that clients now ask us to stay. Roughly two thirds of our revenue comes from platforms we helped build and now help run, under published service levels and a shared on-call rota. That is a deliberate constraint rather than a revenue line: a consultancy that carries the pager writes different software.
How we decide
Six positions we hold consistently enough that clients can predict us. Each of them costs us something, which is how you can tell they are real.
- VAL-01
Say the number early
The estimate goes in the discovery report, before the contract is signed and while it can still change the client’s mind. We have talked four prospective clients out of programmes in the last two years. Two of them came back with smaller ones.
- VAL-02
Reversible by default
Every change we ship has a way back that has been executed, not merely written down. If the rollback has never been run, there is no rollback — there is a paragraph.
- VAL-03
Write it down for the next person
Runbooks, decision records and control documentation are deliverables with acceptance criteria, not a phase that gets cut when the dates tighten. We test a runbook by handing it to somebody who has never seen the system.
- VAL-04
Prefer the boring option
Proven technology, familiar patterns, and the smallest amount of novelty the problem actually requires. A platform is going to be maintained by people we have never met, for longer than we will be near it.
- VAL-05
Report our own faults first
When we break something or miss something, the client hears it from us before they hear it from an auditor, a customer or a dashboard. We have absorbed the cost of a failed penetration test rather than let it surface on somebody else’s report.
- VAL-06
Hand the pager over
Success is the client’s own team running the platform without us. Every engagement ends with a handover rehearsal, and we measure how many follow-up calls it generates. The target is zero, and we usually get there.
Trajectory
Select a year to read what happened. The rail scrolls; arrow keys move between entries.
2014
One contract, two people
Arcwright starts in a rented room off Deansgate with a single engagement: replacing a 1987 COBOL savings ledger for a regional building society. The replacement ran alongside the original for seven months before anyone was willing to switch the old one off. That parallel run became the house method.
The people you would actually work with
We do not staff engagements from a bench. The partner who writes your discovery report is in the delivery meetings a year later, and everybody on this page bills.
Freya Ashworth
Founder & Managing Partner
Started Arcwright in 2014 with one contract and a rented desk on Deansgate. She still writes the discovery report herself on every engagement over £1m, including the 19-month Nordvik Bank core migration. Eleven years in retail banking change before that, four of them unpicking a failed outsourcing programme — which is why she is blunt about what a rewrite actually costs.
Sindre Halvorsen
Partner & Principal Architect, Oslo
Opened the Oslo studio in 2018 and runs the Nordic practice. He designed the event backbone that carries Kestrel Energy settlement traffic across five grid regions, holding a 06:00 CET clearing deadline every day since 2021. Sindre argues for the boring option roughly nine times in ten, and is usually right.
Priya Raghunathan
Head of Data Engineering
Moved 340 million patient records for Orrery Health across a live estate with zero reconciliation breaks, then turned the dual-run harness she built for it into the tool we now use on every migration. She specialises in the part everyone underestimates: proving the new system agrees with the old one, row by row, before anybody switches anything off.
Callum Reid
Delivery Director
Runs the delivery office — 31 consultants across two studios and the forecast that tells clients the truth about dates. He cut the Vanta Logistics replatform into 14 shippable increments so the business took value from week six rather than waiting two years for a single switch-over. He has never run a big-bang cutover and does not intend to start.
Nadia Whitcombe
Head of Security & Assurance
Takes our work through the audits: DORA, PCI DSS 4.0, NIS2 and the NHS Data Security and Protection Toolkit. She assembled the evidence pack that carried the Cambridge Mutual platform through its PRA review with no findings raised. She spent six years on the other side of that table, which is why our control documentation is written for the person who has to read it.
Tomás Beltrán
Head of Platform & Reliability
Owns the service levels on the 23 production platforms we run or co-run with client teams. His on-call redesign at Heliostat cut paging volume 71% in a quarter, mostly by deleting alerts nobody had acted on in two years. He writes the runbooks that go out with every handover, and tests them by handing his laptop to someone who has never seen the system.
Marte Lindqvist
Principal Consultant, Regulated Change
Translates regulation into backlog items. She mapped the 1,100-page obligations register at Nordvik Bank onto 62 concrete system controls, each one with a named owner and a test — which is how a compliance deadline stops being a quarterly panic. Ten years in Norwegian financial supervision before consulting.
Joseph Adeyemi
Head of Quality Engineering
Builds the test estate that makes a cutover reversible: contract tests, shadow traffic, and a replay of a full trading day before anyone touches production. His replay rig caught the settlement rounding defect at Pillarstone eight weeks before go-live, when fixing it cost three days instead of a regulatory notification.
Inside the studio
Six things that are true about working here, drawn rather than photographed — the same way we document everything else.
FIG-01 Thursday teardown
Ninety minutes a week where one engineer walks the rest of us through a system they did not build. Attendance is optional and nobody skips it.
FIG-02 Two studios, one rota
Manchester and Oslo share a single on-call schedule across a one-hour offset. Nobody carries a pager two weeks running.
FIG-03 Written before spoken
Decisions arrive as a one-page record before the meeting, not as slides during it. Half of them stop needing the meeting.
FIG-04 Four days on site, one on tools
Fridays are for the internal estate: the dual-run harness, the replay rig, the tooling that makes the other four days possible.
FIG-05 Nobody is on the bench
We size the firm to the work we have, not to the work we hope for. It is why we turn engagements down, and why we have never run a layoff.
FIG-06 The post-mortem is published
Every incident on a platform we run gets a blameless write-up, and the client gets the same document we do — including the parts that are about us.
Measured, not asserted
Four figures we publish because clients ask for them in procurement, and because a number you refuse to state is an answer of its own.
- Uptime across the platforms we run
- 99.98%
- Core migrations delivered since 2014
- 41
- Transactions cleared daily on systems we built
- 6.2M
- Countries where we support live systems
- 11
Bring us the platform you are afraid to touch.
Discovery is a fixed-price engagement with a written report at the end of it — including the estimate, and including the parts of your programme we think you should not build.