Leaving this course?

Lesson 1 of 9

What you are actually investigating

An investigation that does not know what it must prove collects the wrong things. So begin with the elements of the offence, because they decide what evidence matters.

English law splits fraud into three offences, and the division is useful well beyond England.

Fraud by false representation. A person is in breach if he dishonestly makes a false representation, intending to make a gain for himself or another, or to cause loss to another or expose another to a risk of loss. Three details repay attention. A representation is false if it is untrue or misleading and the maker knows that it is, or might be, untrue or misleading. A representation may be as to fact or law, including as to anyone's state of mind. And it may be express or implied.

There is a fourth detail that matters more each year: a representation may be regarded as made if it, or anything implying it, is submitted in any form to any system or device designed to receive it. A lie told to a machine is still a lie told, which is why automated applications and self-service channels are squarely within the offence.

Fraud by failing to disclose information. A person is in breach if he dishonestly fails to disclose information which he is under a legal duty to disclose, with the same intention as to gain or loss. The legal duty is the hinge. Silence is not fraud in itself; silence where the law required speech is.

Fraud by abuse of position. A person is in breach if he occupies a position in which he is expected to safeguard, or not to act against, the financial interests of another person, dishonestly abuses that position, and intends to make a gain or cause loss. And then the sentence investigators most often miss: a person may be regarded as having abused his position even though his conduct consisted of an omission rather than an act.

That third offence is the one that catches the attorney, the carer, the trustee, the adviser and the employee. It also explains why those cases feel different to investigate: there is no break-in and no forgery, only a relationship and a decision that favoured the wrong person.

What this means in practice. Every one of the three requires dishonesty and an intention as to gain or loss. Neither is visible in a transaction. They are proved by what the person knew, what they were told, what they said at the time and what they did with the money, which is why the documentary and sequencing work in the following lessons matters more than the transaction itself.

Lesson 2 of 9

The first hour, and the things that spoil a case

Most investigations are damaged before an investigator is appointed, by people acting reasonably in the wrong order.

Decide who owns it, and record the appointment. An investigation run by the manager of the department under suspicion is not an investigation. Someone independent of the activity, with authority to see the records and to require cooperation, should be named in writing, with the scope and the reporting line.

Write the hypothesis down, with its alternatives. A case opened as prove that Ahmed took the money will find evidence for it. A case opened as money is missing from this account, and the possibilities are theft by A, error by B, a system fault, or a transaction nobody has yet explained, will test each. Writing the alternatives down is the cheapest protection against the commonest failure in this work.

Then four decisions, in this order.

Preserve. Logs expire, mailboxes auto-delete, CCTV overwrites, and door records roll. The first action is to stop the clock on anything that might matter, in writing, to the people who run those systems.

Contain. Stop the loss continuing, as narrowly as possible. Removing one authority is better than suspending a person; suspending a person is better than announcing an investigation.

Consider the duty to report. Covered in its own lesson, because in a regulated firm it can arise before the facts are clear and does not wait for the investigation to finish.

Decide who may know. Every additional person is a route to the suspect and a risk to the case.

What spoils cases, in order of frequency. Confronting the suspect early, which converts an evidence-gathering exercise into a contest and invites destruction of records. Letting the suspect keep access to the systems holding the evidence. Asking the suspect's colleagues leading questions, which contaminates what they later recall. Changing the data while examining it. And announcing the investigation internally, after which the suspect's account becomes a prepared one.

One more, specific to fraud. Do not approach the counterparty. The supplier, customer or introducer on the other side of a suspected arrangement may be part of it, and a telephone call to check a detail is the most reliable way to tell the suspect that you are looking.

Lesson 3 of 9

Evidence that survives being examined

A finding that cannot be shown to be reliable is an opinion. NIST's guidance on integrating forensic techniques into incident response describes the discipline as the application of science to the identification, collection, examination and analysis of data, while preserving the integrity of the information and maintaining a strict chain of custody. Each of those four words is a stage, and the two conditions apply throughout.

Chain of custody means a record, from the moment something is collected, of what was taken, from where, by whom, when, who held it since and what was done to it. It is unglamorous and it is the difference between evidence and an assertion. A copy taken from a laptop by a helpful colleague who then used the machine has usually destroyed its own value.

Integrity has to be demonstrable, not assumed. NIST makes the point directly: because electronic logs and other records can be altered or otherwise manipulated, organisations should be prepared, through their policies and procedures, to demonstrate the integrity of such records. Where the whole case rests on a system log, somebody will eventually ask who could have edited it, and the answer should be known before the question.

Volatile first. Some data vanishes when a machine is turned off or a session ends, and some survives on disk. Where both matter, the order of collection is decided by what will be lost first, not by what is easiest to copy.

Say in advance what may and may not be done. NIST asks that policies and procedures explain clearly what forensic actions should and should not be performed in various circumstances. In a bank this is not a technicality: a well-meaning examination of a colleague's mailbox without authority can breach employment and data protection obligations even where the suspicion was correct.

And protect what the tools pick up. The same guidance asks for safeguards for sensitive information that forensic tools might record, such as passwords and personal data. An investigation that leaks the bank details of forty uninvolved customers has created a second incident.

A word on proportion. Most fraud cases are proved with records a bank already holds: statements, authorisation logs, access records, approval trails, emails and the sequence in which they occurred. Full device forensics is occasionally necessary and is expensive, slow and easy to do badly. Establish what the ordinary records show before anyone images a disk.

Lesson 4 of 9

Documents, witnesses and the interview

Evidence in fraud cases is mostly paper and timestamps. People come later, and in a particular order.

Build the sequence before speaking to anyone. What was approved, by whom, against what authority, and what the system recorded at each step. A fraud is almost always visible as a departure from a sequence: an approval granted before the request existed, a record created after the payment, a limit raised the day before it was used. Nobody needs to be interviewed to find that.

Get the third-party records early. Company registries, the counterparty's own filings, land registers, the courts' records of judgments and insolvencies. In arrangements involving a supplier or introducer, the connection between the parties is frequently a matter of public record, and finding it before any interview changes what the interview can establish.

Witnesses before the suspect, always. Interview the people who can describe the normal process first: what should have happened, who usually approves, whether the step taken was unusual. Those interviews establish the baseline against which the suspect's account will be measured, and they are much harder to conduct usefully once the organisation knows an investigation is running.

Then the interview with the person under suspicion, which is the single most misunderstood step in corporate investigation. Three things to be clear about.

It is not a police interview and must not pretend to be. An internal fact-finding interview, conducted by employees, carries none of the protections a criminal interview carries, and conducting it as though it were can taint the evidence and prejudice a later prosecution. Where a criminal referral is likely, take advice on sequencing before interviewing at all.

Its purpose is to obtain an account, not a confession. The valuable output is a detailed, recorded explanation that can be tested against the documents. Many cases are proved not by an admission but by an account that the records contradict.

Put the documents in one at a time. An interview that begins by setting out everything known allows one explanation to cover all of it. An interview that establishes the account first, then introduces each document, tests it.

Record it, and record the conditions. Who was present, what was said about the purpose, whether a companion was offered, when it started and ended. A contemporaneous record written by the interviewer afterwards is worth far less than a recording made with consent, and worth something only if it was made the same day.

Lesson 5 of 9

What you must report while you are still investigating

In a regulated firm the reporting duty does not wait for the investigator. It can arise on the first day, and misunderstanding it is a criminal risk rather than a procedural one.

The duty to disclose. Section 330 of the Proceeds of Crime Act 2002 makes it an offence, in the regulated sector, where its conditions are met. The first condition is that the person knows or suspects, or has reasonable grounds for knowing or suspecting, that another person is engaged in money laundering. The second is that the information came to them in the course of a business in the regulated sector. A further condition concerns being able to identify the person or the whereabouts of the laundered property, or the information being capable of assisting in doing so.

The words that matter are reasonable grounds. The test is not only what the individual actually suspected; it is also what they ought to have suspected on the information in front of them. An investigator who chooses not to think about an obvious inference has not avoided the duty.

Note also what money laundering means here. The proceeds of an internal fraud are criminal property, and dealing with them is money laundering, so a fraud investigation routinely engages this duty rather than exceptionally.

And the duty not to tell. Section 333A makes tipping off an offence in the regulated sector. A person commits it where they disclose that a disclosure has been made under that Part, to a constable, an officer of Revenue and Customs, a nominated officer or an authorised National Crime Agency officer, and the disclosure is likely to prejudice any investigation that might be conducted following it, and the information came to them in the course of a business in the regulated sector.

Put the two together and the shape of the problem appears. The firm must report early, on a low threshold, and must not let the subject learn that it has. An investigation generates exactly the behaviour that risks telling them: interviews, access revocations, sudden questions about an old payment, a frozen account.

So sequence deliberately. Decide what has been reported and what must stay unsaid before the interviews begin. Keep the report and its existence on a need-to-know basis separate from the investigation file that more people will read. Where an operational step would reveal the report, get advice rather than improvising, because the choice is not between two internal risks but between two offences.

These are British provisions, used here because they state the pattern plainly. Most jurisdictions have both duties in some form, including Kenya, and the sequencing problem they create is the same everywhere.

Lesson 6 of 9

Following the money, and getting some of it back

Recovery and investigation run on different clocks. The investigation can take months; the money moves in hours. Treat them as separate workstreams from the first day, or the second one will be finished before it starts.

Act on the destination before the explanation. The question for recovery is not who did this but where did the money go and can it be stopped. Those are answerable from the payment records alone, usually within the hour, and they do not require a conclusion about anybody.

The routes, roughly in order of speed. A recall or freeze request to the receiving institution, which is fastest where the receiving account is at a bank that answers. A law-enforcement freezing route, which is slower and reaches further, including across borders. A civil application to freeze assets, which is expensive and powerful where the sums justify it. And insurance, which recovers money rather than assets and takes longest of all.

Expect dispersal, and plan for partial success. The companion course on business email compromise sets out the public figures: of the attempted theft put through the FBI's freezing process in 2024, a majority was frozen domestically and much less internationally, and in one case the portion not held had been wired onward immediately on arrival. The pattern is general. Recovery is a function of how fast the first telephone call happens, and partial recovery is the normal good outcome.

Internal frauds have one advantage. Where the perpetrator is an employee, there may be identifiable assets, a salary, a pension and a known address, and the civil routes become realistic in a way they rarely are against an anonymous external fraudster. That advantage is lost by delay and by any settlement signed before the scale is known.

Three traps. Settling early for a repayment, which feels like recovery and can extinguish the evidence and the claim. Accepting an instalment arrangement with no security. And allowing the suspect to resign and leave with a reference, which is dealt with in the next lesson.

Keep the tracing work documented as evidence, not as notes. The chain of payments that shows where the money went is frequently the clearest exhibit in the eventual case, and the version assembled in a hurry on the first day is usually the one that matters. Write it as though a court will read it, because one may.

Lesson 7 of 9

Deciding what happens next

A finished investigation produces a decision, and there are four of them. They are not alternatives to be chosen between so much as questions to be answered separately.

No case. The most under-used outcome. Money missing is not money stolen, and a serious proportion of cases end as error, system fault or a transaction nobody had explained. Saying so clearly, in writing, is a result and protects the person who was suspected.

Internal and disciplinary. The employment standard of proof is lower than the criminal one, so an internal case can succeed where a prosecution would not. It needs a fair process: the allegation put, the evidence disclosed, an opportunity to respond, and a decision by someone who has not already decided.

Civil recovery. A claim for the money, on the balance of probabilities. Independent of whether anybody is prosecuted.

Criminal referral. A matter for the authorities, on their standard and their timetable, and the one over which the organisation has least control.

Two points about running them together. The criminal process may constrain the others: material gathered internally may have to be disclosed, and an internal interview conducted carelessly can damage a prosecution, which is why the sequencing in the interview lesson matters. And an acquittal does not reverse a dismissal, because the questions and standards were different.

Then the ending this course argues against. The quiet resignation. The person leaves, the matter closes, no finding is recorded, and a reference goes out that mentions nothing. It is attractive: it ends the cost and the embarrassment in one afternoon. It also sends a person with a demonstrated willingness to steal into another employer with a clean record, and it tells everyone who knows what happened that the organisation will not act. The catalogue's course on insider fraud makes the same argument from the control side.

So record the finding even where nothing else follows. What was alleged, what was established, what was not, and what the organisation decided and why. That record is what makes a reference honest, a regulatory notification possible, and a repeat pattern visible in three years' time when the same name or the same method appears again.

Lesson 8 of 9

Writing it up

The report is the only part of the investigation most people will ever see, and it is where good work is commonly thrown away.

Separate four things, visibly. What is established by evidence, with the evidence identified. What is inferred, with the inference stated as such. What is alleged by someone. And what is unknown, which deserves its own section rather than silence. A report that blends them reads persuasively and cannot be relied on, and the first serious reader will find the seam.

Write the chronology first, and let it do the work. Dates, times, amounts, actors, systems, in order. Most fraud is obvious from a well-built sequence, and the chronology is also what survives if the conclusions are later disputed.

Name the standard you have applied. Whether this report asserts something is established on the balance of probabilities, or merely that there is a case to answer, changes what the reader may do with it. Investigators often leave it to be guessed.

Know who reads it. A board wants the exposure, the control failure and the decision required. A regulator wants the facts, the timeline and what the firm did on learning them. An insurer wants the loss, the proof of it and the policy conditions. A court wants the exhibits and the custody of them. The same findings serve all four; one document rarely does, and a short decision note over a full evidential report usually works better than a compromise between them.

Four failures worth naming. Adjectives doing the work of evidence, where words such as clearly, obviously and blatantly appear exactly where the proof is thin. Conclusions about a person's state of mind stated as fact, when dishonesty is for a tribunal or a court. Omitting the exculpatory material, which destroys the report's credibility the moment it emerges. And no statement of what was not examined, which leaves the reader unable to judge the weight of anything.

Finally, write it so the subject could read it. They may. Employment proceedings, data protection requests and litigation all open the file, and an investigator who would be uncomfortable seeing their own report quoted back to them has usually written something other than an investigation.

Lesson 9 of 9

What the organisation learns

A case that ends with a person and no change to the process has taught the organisation nothing, and the next case is already possible.

Separate the culprit from the cause. Who did it is the investigation's question. What allowed it is the organisation's, and the answers are usually unremarkable: an approval that one person could grant alone, a system right nobody reviewed when someone changed role, a reconciliation performed by the person who created the entries, a limit raised without evidence, a control that existed on paper and had not run for a year.

Ask how it was found, and be honest about the answer. If the fraud was discovered by accident, by a colleague's unease, by a customer complaint or by an external party, then the control framework did not find it, whatever the framework says. Recording the discovery route for every case, and watching the proportion found by design rather than by luck, is the single most informative measure a fraud function can keep about itself.

Measure the function, not the criminals. Useful numbers: time from the first signal to the first containing action, time from discovery to the first recovery call, proportion of cases closed as no case, proportion of recommended control changes actually implemented and still working a year later, and the discovery route above. Counting cases and losses describes the environment; these describe the organisation.

Close the loop in writing. Every finding should produce a named owner, a date and a check that it was done. Control recommendations that nobody owns are the most common output of investigation functions and the least consequential.

And be honest about the limits. Investigation is retrospective by nature. It establishes what happened, recovers some of the money, and tells the organisation what to fix; it does not prevent the case it examines. A firm that invests in investigation while leaving its approval and access controls alone has bought a good account of its own losses.

The three questions worth asking after every case: would the same thing work again tomorrow, would we find it faster, and does the person who could do it now have one fewer way to do it. If the answer to any of them is no, the case is not closed, whatever the file says.

Knowledge check

Thirty questions

Answer all thirty questions, then check your answers. You need 27 out of 30 to pass and receive a certificate. If you score less, you will see which answers were right and wrong, and then go through the course again before you retake the check. Your answers, progress and times are kept only in this browser.

Standards and frameworks covered

The published frameworks this course teaches toward. Saolix is not affiliated with, endorsed by or accredited by any of these bodies; naming a standard is a citation, not a claim about its author.

Sources

The official documents this course relies on. Laws and guidance change, so check the current version.

  1. SP 800-86: Guide to Integrating Forensic Techniques into Incident Response · US National Institute of Standards and Technology
  2. Fraud Act 2006, section 2 (fraud by false representation) · legislation.gov.uk
  3. Fraud Act 2006, section 3 (fraud by failing to disclose information) · legislation.gov.uk
  4. Fraud Act 2006, section 4 (fraud by abuse of position) · legislation.gov.uk
  5. Proceeds of Crime Act 2002, section 330 (failure to disclose: regulated sector) · legislation.gov.uk
  6. Proceeds of Crime Act 2002, section 333A (tipping off: regulated sector) · legislation.gov.uk