Behind the Curtain: The Anatomy of a Brown Envelope

I can tell you what they cannot know: that letter is not the beginning of anything. It is closer to the end of a story that started more than a year ago - a story they were never part of, even though it was entirely about them.

I spent seven years at HMRC, the last stretch of it as an operational lead in the Fraud Investigation Service. The decision that letter represents - the decision to investigate - was my job. Since leaving, I have spent a few years on the other side of the table, sitting next to people in the days after the envelope arrived. Having seen it from both seats, I think taxpayers have a right to understand how their government approaches this, to the extent I am able to share it. Some things stay behind the curtain, and rightly so. But the shape of the machine, how it thinks, what it is doing long before anyone knows it is there - that much, people are entitled to know.

So let me take that letter and run the film backwards.

Six weeks before the envelope: the sign-off

The last meaningful thing to happen before that letter was printed was a decision - and for cases in my team, the decision was mine.

Signing off an investigation is a strange weight to carry. You know, in that moment, that you are about to put a person through one of the most stressful experiences of their life - years of it, potentially. And you sit with that alongside the other truth: that when we were right, and we often were, we were identifying someone genuinely defrauding the country while everyone around them paid their share. It is a peculiar moral position - being the bad guy while being, underneath it, the good guy. People hate the taxman. People also hate watching someone glide past in a Lamborghini having paid no tax at all. We were the answer to the second feeling and the target of the first, and it was a fairly thankless place to stand.

The job also does something to you over time. The moral payout is real, but so is a slow decay in your sensitivity to what an enquiry actually feels like from the other side. It became a file, an outcome, a statistic. It was only when I moved into the private sector - into the seat beside the person holding the envelope - that the sharp edges got rounded off. In that seat you are a tax adviser second. First, you are a consoler, and on the hard days something closer to a therapist. Nothing in my HMRC years prepared me for how heavily an enquiry sits on a person's shoulders, because from the inside, I never had to watch.

Three months before: the case that nearly wasn't

Rewind further, and the file that became that letter is sitting in a stack with others - most of which will never become anything at all.

Most files died. A case could fall for any combination of reasons: the tax at stake was too small, the evidence pointed to non-compliance but not fraud, the wealth had plausible sources outside the UK, or there was clear family wealth that explained the picture. Politically sensitive cases were not off the table, but they carried a weight the others did not - with those, you simply could not afford to be wrong, and you felt that every time one crossed the desk.

What kept a file alive ran the other way. Previous non-compliance was a strong indicator. The number of tax years mattered enormously - our powers allowed us to go back up to twenty years in the right circumstances, and we wanted the hardest-hitting cases, not the ones that scratched two or three years and stopped.

HMRC knows the tax gap is vast and its investigators are finite. The whole exercise existed to find the top percentile - high value, high profile, or serious enough that pursuing it was plainly in the public interest. So not every case with fraud in it gets worked as a fraud case. The machine is selective by design. Each investigator, worked properly, returns far more to the Exchequer than they cost - which is exactly why investigator numbers are growing now.

And here is something the person holding the envelope never sees. Sometimes a file that looked strong but lacked evidence of fraud did not die at all - it was referred to a lower compliance team to open the enquiry and gather information, while FIS monitored progress in the background, waiting to see whether the evidence it expected would surface. People took real comfort from which team a letter came from; a general compliance letterhead felt survivable in a way FIS did not. The letterhead told you who was writing. It did not always tell you who was watching.

Five months before: the presentation

Further back still, and the file is being presented to me by one of my team.

What they were working from was someone's entire financial footprint - and it contained no opinion. Just data. Whether that data amounted to fraud was a judgement, and the judgement belonged to us. It was almost never one thing. It was the pattern - what the bank records said next to what the lifestyle suggested next to what the returns declared. The skill was in reading the whole picture, and that skill was human, not algorithmic.

Sometimes it came down to the right case finding the right investigator - a file that matched someone's particular skillset, or one they had genuine passion for. An investigator who cares about a case works it better; that is as true in HMRC as anywhere else. But passion cuts both ways, and part of my job was cutting through it - an enthusiastic presentation of a file is not the same as a strong one, and someone had to hold the balance. My team brought me their conviction. I owed them, and the taxpayer, a colder read.

Book a demo

A year before: the SIP

Keep rewinding, and the file itself comes apart into its raw material.

It arrived as what we called a SIP - a Standard Intelligence Package, though the name may well have changed since my time. Financial records, banking information, probate, credit history, flight data, open source material, social media. A person's whole footprint, lawfully gathered, compiled into a single electronic profile - built over anywhere from one to six months depending on complexity, during which the subject had no idea any of it was happening.

It is worth knowing how some of that banking information arrives. Since 2021, HMRC has been able to issue what is called a Financial Institution Notice - a formal demand to a bank for a customer's statements and account information - without tribunal approval, and without the customer ever being told. I issued them; they were among the most quietly powerful tools we had. It meant that by the time HMRC asked a taxpayer a question, it very often already held the answer - or at least a significant part of the story - and the question existed to see what came back. If any part of you assumes your bank would be on your side in that exchange, or would at least mention it, that assumption is wrong. They must comply, and they must do it silently.

The SIPs came out of HMRC's Risk and Intelligence Service, who ran distinct ongoing projects into live issues - a data leak HMRC had gained access to, say - with the Fraud Investigation Service getting first pick of the resulting packages before anything passed down the line. I helped develop that pipeline into my own team, so I say this with some affection: RIS were good at their jobs, but they were intelligence collectors, not enquiry-trained investigators - and they were fragmented almost by design. RIS, RIS Offshore, this project, that project. Even from the inside it was hard to grasp the full breadth of the operation. I sat on calls with thirty or forty people, each pitching their own project as the one to pick cases from - a strange experience, being sold to from inside your own organisation. The mandate, at least, was never unclear: deliver cases, and make sure they are high value. Everything was tracked as a statistic, and projects were started, stopped and revisited on the strength of those numbers. Admirably flexible - and genuinely hard to keep consistent, when an enquiry takes two or three years to resolve and truly show what it was worth.

That was one route of three, in my time. Cases also came from our own team's research - identifying prospective taxpayers, including people with no tax footprint at all, and deep diving using every source lawfully available. The third route was through professional contacts on the other side of the fence - advisers looking to place disclosure cases with the right team. That road still exists, but it has narrowed considerably - HMRC moved over time to get a tighter handle on what each team was working and to stop teams effectively collecting the cases they wanted through relationships, and what was once common is now much harder and much rarer. I understood the thinking. I still think it has a place, with the right guardrails - some of the best-worked cases I saw came through advisers who knew exactly what a good case looked like.

Before any of it: a name in a dataset

And now we are at the true beginning, which is the quietest moment in the whole story.

Before the SIP, before the projects, before anyone with judgement was involved at all - there was simply a name surfacing in data. A mismatch between declared income and visible life. A connection to a structure that fitted a known pattern. A record in a leak. Nothing dramatic happened. Nobody decided anything. A name just started to sit differently in a dataset than the names around it.

This is where I should say something about the machine, because when I moved into the private sector I kept hearing HMRC's central intelligence system described as an all-seeing oracle. The reality is more modest and, I think, more interesting. It is genuinely powerful - but it only shows data. It has real limitations - fuzzy name matches, outdated records that need verifying before anything can be relied on. Whether any of it amounts to fraud is entirely down to the skill and experience of the people reading it. The strength of the system was never really the technology. It was pattern recognition - dedicated teams building deep expertise in specific fraud types, the structures, the indicators, the commonalities. In recent years that expertise has extended to labour supply chains and mini umbrella company arrangements. When a sector develops a known fraud pattern, HMRC does not investigate it randomly. It studies the pattern until its people can recognise it at a glance - and then the data does the finding.

Every industry has its patterns - that is simply what happens when enough cases pass through enough experienced hands. Construction happens to be one of the sectors HMRC knows better than most.

Back to the doormat

So that is the film, run backwards. A name in a dataset. A package built in silence. A presentation, a judgement, a sign-off. And then, more than a year later, a brown envelope - landing on the doormat of someone who believes, entirely reasonably, that this is day one.

From the seat beside them, I heard the same explanations again and again, and I understood every one. Someone has tipped them off - tip-offs were far rarer than people imagined, and where they existed they were usually one line in a SIP to be corroborated rather than the spark. They have an agenda against my company - in my experience files were assessed on risk and value, and a case built on feeling rather than evidence tended to fall apart; if anything, the absence of feeling in the process was the most unsettling thing about it. They clearly haven't done their homework - by the time contact was made the homework was usually long finished, often including bank information the taxpayer never knew had been requested.

I will not pretend the machine was infallible - it was not, and anyone who has dealt with HMRC will have their own story of something misjudged or mishandled, and plenty of those stories are fair. But the mistakes I saw were almost never the ones people assumed. They were rarely mistakes of malice or laziness. When HMRC got it wrong, it was usually a judgement call that did not land - not a vendetta, and not a guess.

The truth is less personal than any of the explanations people reach for, and more thorough. And the most useful thing I can say runs the other way, before any envelope exists. The whole process I have described was, at its heart, a search for people whose affairs could not answer questions. The files that never became cases were so often the ones where the picture held together - where the records existed, the explanations were evidenced, and pursuing it further was plainly not worth the resource. You cannot control whether your name ever surfaces in a dataset. You can control what the picture looks like when it does.

I sat in that chair. I never once opened a case because someone's paperwork was too good.