A company that couldn’t run its own software — until it could.
A live regulated system, four missed launch dates, one supplier holding every key. Here’s the value we helped them take back: full control, a priced way out, and no interruption to service.
Four questions. No evidenced answers.
Every answer available came from the vendor — so the only people who could evaluate the work were the people being evaluated. Not a trust problem. A structural one.
If the vendor stopped work tomorrow, could we keep operating?
Are we getting what we are paying for?
Is the next date real, or is it the fifth one that slips?
If we wanted to leave, what would it actually cost us?
A live system with a legal deadline attached.
4/4
Launch dates set and missed. The fifth shipped only because a regulatory deadline forced it — not because anything was ready.
Live
Already in production carrying regulated filings when the review began. Any misstep reached customers with a legal deadline.
1
One person understood the system end to end. No documentation existed that would survive their departure.
0
Automated checks standing between a bad change and production. No rollback anyone in-house could execute.
One number ended the debate.
Components only the supplier could reach.
Without that access the platform could not be built. Not slowed down — unbuildable. Three years of funding had produced a product the business could not compile.
Test files protecting 380,000 lines of production code
Checks in the pipeline that could stop a bad release
Recovery docs, monitoring, or infrastructure definitions
Contracted deliverables actually live — all 31 invoiced
We built the ability to leave first, and left last.
The review gave them the picture. They then asked us to act on it — a separate engagement they were free to give to anyone. Firing the vendor overnight would have meant losing the only people who could run a live regulated system. That is how a difficult situation becomes an outage.
Secure the assets
Source estate, private components, cloud accounts, and build images moved under the client’s own control. Every exposed credential rotated.
Build the quality gate
Four environments stood up on their own accounts, with automated tests over the critical paths blocking any release that failed.
Stabilise
Releases made repeatable and reversible. Silently failing data pipelines fixed. Monitoring added where there was none.
Transition
Delivery moved to a team they control, with the vendor stepping back in planned stages rather than all at once.
What it was worth to the business.
Renewal moved from a conversation they could not win to one where leaving was a priced option on the table.
An outage stopped depending on one supplier’s availability, on a system with a legal deadline attached.
Stale and wrong data stopped reaching customers unnoticed — the failure that costs standing rather than money.
The vendor relationship ended on their timetable, with no interruption to the service.
Client name withheld under NDA. Every figure is as measured. The review findings came from analysis of the code and delivery record — no runtime testing or interviews. The transition that followed was a separate, optional engagement; its phases overlapped rather than running cleanly in sequence.
How many parts of your system can only one party reach?
If that takes more than ten seconds to answer, that is the finding. Tell us the decision you’re weighing and we’ll tell you what we’d want checked first.
Independent · Fixed fee, not contingent on findings · We never sell you the fix