Silent Failure Audit · for UK and EU regulated businesses
I find the automated jobs that say they worked, and didn't.
Deleting old customer data, checking customers against sanctions lists, sending e-invoices, moving numbers between systems. If software does it for you, I check it really happened. Five days, a fixed fee, and the answer in writing.
“Most broken systems still report healthy.”
Monitoring catches errors. It can't catch a job that quietly did nothing.
- Takes
- 5 working days
- Costs
- £1,500–£3,000 fixed
- Access
- Read-only
- You get
- A written report
The difference, in one picture
Your dashboard says everything ran. The audit checks whether it did.
Switch between the two views. The jobs are the same, and none of them ever raised an error.
Scheduled jobs, last quarter
Illustrative example
- Sanctions screeningNightlyDashboard saysCompletedActuallyNot checked
- E-invoice batch (KSeF)NightlyDashboard saysCompletedActuallyNot checked
- GDPR erasure jobNightlyDashboard saysCompletedActuallyNot checked
- Payroll exportMonthlyDashboard saysCompletedActuallyNot checked
- Audit log exportHourlyDashboard saysCompletedActuallyNot checked
Why people get in touch
Nobody goes looking for this until something makes them ask.
Something already slipped through.
A penalty letter, a customer complaint, or personal data that should have been deleted. You've fixed that one. The real question is what else has stopped without telling anyone.
Someone wants proof.
An auditor, a regulator, an insurer or a buyer wants you to show that it happens, not tell them it does. A green dashboard isn't evidence.
You've just taken it over.
New in the finance, compliance or tech seat, and answerable for systems somebody else set up. Better to know now what you're signing for.
Your systems just changed.
A new finance system, a move to the cloud, a new e-invoicing rule. Things get switched off and rebuilt during a move, and some of them never come back on.
- Financial services and insurance
- Businesses sending e-invoices (KSeF and EU rules)
- UK companies filing VAT and accounts
- Businesses preparing for a sale
How it works
Three steps. No jargon.
- 1
You tell me what runs on its own.
A 30-minute call about what made you ask, and about the things your systems do with nobody watching: filings, submissions, deleting old data, reports, and the links between systems.
- 2
I check each one actually happened.
Two questions for each one. Did it run when it should have? Did it produce what it should have? I only need permission to look. I don't change anything.
- 3
You get a written list.
What ran, what didn't, what came out empty, and what each gap could cost you. Ranked by risk, so a missed filing sits above a stopped log file.
What I check
If it runs without a person watching, it's in scope.
VAT returns (MTD)
- It says
- Completed
- I check
- Did HMRC send back a receipt for every quarter?
E-invoices (KSeF and EU rules)
- It says
- Batch sent
- I check
- Did every invoice get a number back from the tax portal?
Deleting old personal data (GDPR)
- It says
- Success
- I check
- Was anything actually deleted?
Company filings (CT600, Companies House)
- It says
- Submitted
- I check
- Was it received and accepted?
Reports and data feeds
- It says
- Finished
- I check
- Did the report arrive, with sensible numbers in it?
Links between your systems
- It says
- Sent
- I check
- Did the other system actually get it?
Why nobody notices
Monitoring watches for errors. A job that never starts doesn't make one.
Monitoring tells you when something goes wrong. A job that never starts hasn't gone wrong as far as the software is concerned, so it says nothing. A job that runs and does nothing still counts as a success. So the dashboard stays green, and the problem turns up months later as a penalty, a failed audit or a number nobody can explain.
What is a silent failure?01Green means "no errors"
It doesn't mean the work got done. Those are two different things, and almost nobody checks the second one.
02Nobody checks what came out
A job can finish cleanly and produce an empty file. Unless something looks at the file, that counts as a pass.
03It gets worse quietly
A loud failure is fixed the same day. A silent one carries on for as long as it takes someone to notice.
What you get
One document. Everything that runs on its own. A clear answer for each.
Findings, ranked by exposure
Illustrative example
- 01HighSanctions screening: no run since 14 JulyCustomers not checked against list changes
- 02HighRetention job: 0 records erased since FebruaryPersonal data held past policy
- 03MediumAudit log export: empty file for 41 daysGap in the audit trail
- 04OKPayroll export: running, output as expectedNo action needed
- Duration
- 5 working days
- Fee
- £1,500 for one area, £3,000 for a full sweep. Fixed.
- Access
- Read-only, removed at the end
- Changes to your systems
- None
- Deliverable
- Written findings and a one-hour walkthrough call
- If nothing is wrong
- The report says so, in writing. Same fee.
Also available
When you need more than an audit.
AI automation with a receipt
Everyone sells AI automation. I build the kind that has to prove it ran. Most automations are judged by whether they threw an error. Mine are judged by whether the work actually happened, so every build ships with its own receipt.
Scoped per build
Fractional CTO and interim technical lead
One or two days a week owning the technical decisions on a .NET and Azure estate: architecture, delivery, cost, hiring, and the judgement calls that need someone who has made them before.
£1,000–£1,200 per day · typically 1–2 days a week
Technical due diligence
Assessment of a .NET and Azure target for investors and acquirers, with particular attention to the automated processes a standard diligence checklist tends to take at face value.
Scoped per deal
Who you'll work with
One person. 25 years of building these systems.
Web Project Studios is me, Arek Adeoye, not an agency. I've spent 25 years building and fixing the software businesses run on, mostly Microsoft .NET and Azure, and around 10 of them leading the teams that build it. Much of that was regulated work, where being able to prove something happened matters as much as it happening. I found this problem in my own systems: a job that had reported success for weeks without running once.
More about meWhere this comes from
doFaktur.pl
E-invoicing under Poland's KSeF mandateAn invoicing product built against a national e-invoicing regime, where a submission that reports success and never lands is a compliance problem, not a bug.
dofaktur.plGovernment filing integrations
HMRC and Companies House APIsIntegration work against Making Tax Digital, corporation tax returns and Companies House. The same failure mode, with a deadline attached to every one.
Automations nobody watches
Multi-step automation on a scheduleAutomations that run on a schedule with nobody watching, built with sign-off steps, testing and monitoring. Running these is how I found the problem this practice is built around.
Common questions
The questions I hear most.
What does the first engagement look like?
The Silent Failure Audit. Five working days, read-only, and a fixed fee: £1,500 for one process area, £3,000 for a full sweep. You get a written findings document listing what should have run, what actually did, what produced output, and the exposure attached to anything that didn't.
I don't think anything is broken. Why would I check?
You probably wouldn't, and that's fine. Most people get in touch after something has made them ask: a penalty, a question from an auditor or a buyer, a new role, or a change of systems. If none of those apply to you, this can wait. If one does, five days gets you a written answer instead of a hope.
We already have monitoring and alerting.
Almost everyone does, and it's usually good at its job. But alerts go off when something goes wrong. A task that never starts hasn't gone wrong, it just hasn't happened, so nothing goes off. That's the gap this looks at. It's how monitoring tools are built, not a sign yours was set up badly.
I'm not technical. Will I understand the report?
Yes. It's written for the person who answers for the filing or the data, in plain English: what should have happened, what did, and what each gap could cost. The walkthrough call is there for anything that isn't clear.
What if you don't find anything?
Then the report says so and you have that in writing, which is worth something on its own if you're heading into an audit or a sale. I'd rather tell you it's fine than manufacture a finding. The fee is the same either way, which is deliberate.
How much access do you need?
Permission to look, not change: the records of what ran and when, any logs, and the places the results should end up. Nothing in your live systems is changed during an audit. If access takes time to arrange, I'll scope around what's available.
Is this only for regulated businesses?
No, but that's where it pays for itself fastest, because the cost of a task that quietly stopped is a penalty or a failed audit rather than an internal annoyance. If nothing you run unattended carries that weight, this is probably not urgent for you and I'll say so.
Start with the audit
How would you know if it stopped?
Pick the process you'd least like to have quietly failed. If you can't say how you'd find out, that's the one to check first.
Five days, fixed fee. You get a written findings document.
Currently working with a small number of clients.