Read
SAR Narrative Writing: What to Include and What to Avoid
FinCEN & Regulation

SAR Narrative Writing: What to Include and What to Avoid

A SAR narrative has to do three things a reviewer can check: state the facts in order, show why those facts were unusual for that particular customer, and record what the institution did about them. Everything else is style. Most weak narratives fail on the second point rather than the first. They list transactions accurately and never establish the baseline that makes those transactions suspicious, so the reader is asked to take the conclusion on trust. This guide covers what a reviewer needs, a structure that holds up, the details that make a narrative defensible, and the habits that create review problems.

A SAR narrative is not a summary of an alert. It is the written account of what the institution observed, why the observation mattered for this customer, what was checked before deciding, and what the institution did next. A reviewer should be able to read it once, without the file open, and understand the sequence of events and the reasoning that connects them. Most narratives that get criticized in quality control fail on one of two points: they copy the monitoring alert into prose, or they describe activity accurately and never explain what made it unusual. The second failure is the more common one, and the more damaging, because a reader who cannot see the baseline cannot evaluate the conclusion. This guide covers what belongs in the narrative, how to sequence it, the details that make it defensible, and the habits that quietly undermine it.

The Short Version

  • The narrative has to prove three things: what happened, why it was unusual for this customer, and what the institution did about it.
  • Describe facts; conclude only where the record supports it, and label the conclusion as the institution's assessment.
  • Copying the alert is the fastest way to produce a narrative that says nothing a reviewer could not already see.
  • Prior filings, prior alerts, and activity after the review period belong in the story, with dates.
  • Gaps are acceptable when they are named. Guessed facts are not.
  • A second reader who did not work the case is the single most effective quality control step.
  • The narrative supports a filing decision; it does not establish guilt, and it should not read as if it does.

Why the narrative matters more than the alert

The alert was produced by a monitoring system. It is a flag, and it is allowed to be crude, repetitive and generic, because its job is to surface something for a human being to look at. The narrative is the part a person reads, and it is usually read cold: months or years later, by someone who has none of the context the analyst had while working the case. That reader may be an examiner sampling files, an internal quality control lead, a law enforcement analyst working backward from a filing, or a colleague inheriting a portfolio from someone who has left. None of them can ask the analyst what they meant.

So the narrative has to be self-contained. A reader should be able to follow it without the transaction history in front of them, and should be able to point at the sentence where the concern becomes visible. A narrative that records transactions but never explains why they mattered forces the reader to reconstruct the reasoning, and readers do not reconstruct reasoning in the institution's favour. They conclude the file is thin, and they are usually right.

Describing is not the same as concluding

Describing is recording that six deposits of similar size arrived on consecutive business days, from three different branches, into an account that had previously received a monthly payroll credit. Concluding is stating that the customer structured the deposits to avoid a reporting requirement. The first is a fact a reviewer can verify against the statement. The second is an inference, and it belongs in the narrative only when the surrounding record supports it, and only when it is framed as the institution's assessment rather than as established fact.

That distinction is where most of the craft sits. A narrative that jumps to the conclusion reads as advocacy. A narrative that lays out the facts, states the inference briefly, and then says what was checked to test that inference reads as analysis. The second kind is much harder to argue with, and much easier for a second reader to defend two years later.

Why the narrative is the part that gets tested

Reviewers rarely re-perform the transaction analysis. They read the narrative, spot-check the figures against the account records, and look for the reasoning. If the narrative is accurate on the numbers and silent on the reasoning, the review becomes an argument about whether the institution thought about the case at all. If the narrative is complete but the figures do not reconcile with the reported fields, the review becomes an argument about reliability. Both are avoidable with the same discipline: treat the narrative as the primary deliverable rather than the last step before submission.

What a reviewer actually needs from the narrative

Strip away the formatting conventions and a reviewer is looking for five things, in roughly this order.

  • Who and what relationship. The subject, the account or product involved, how long the relationship has existed, and what the institution understands the customer to do. If the subject is an entity, say who the principals are and how that was established.
  • What happened. The activity itself, with dates, amounts, instrument types, counterparties and the direction of funds. In sequence, not in summary.
  • Why it was unusual. The baseline the activity departed from. Prior activity on the account, the stated business purpose, the customer profile, and the specific feature of the current activity that does not fit. This is the section most often missing and the one a reviewer weighs most heavily.
  • What the institution did. The records reviewed, the questions asked, whether the customer was contacted and what they said, whether the explanation was verified, and who reviewed the decision.
  • What the institution concluded and why. A short statement of the concern, expressed as the institution's assessment, and whether a filing decision was made. If the decision was not to file, the narrative should still show the reasoning that led there.

A useful test before writing: imagine the reader has the narrative and nothing else, and ask what question they would ask first. If the answer is about the customer's business, put that in the opening paragraph. If it is about why the amounts matter, put the baseline earlier. The narrative should answer the reader's next question before they have to ask it.

A structure that works

There is no single mandated format, but the sequence that holds up under review is consistent, and it is worth using every time so that reviewers learn where to find each element.

  1. Opening identification. One short paragraph: who the subject is, what the relationship is, when it was established, and what the institution does or believes the subject does for a living. No conclusions here.
  2. Account and profile context. The activity the institution expected before the review period. This paragraph is what makes the rest of the narrative legible, because it establishes the baseline.
  3. Chronology of the activity. What happened, in date order, one idea per paragraph or per sentence. Include the instruments, the counterparties, the amounts, and the locations where they matter.
  4. The anomaly, stated plainly. The paragraph that says what does not fit: the pattern that changed, the purpose that does not reconcile, the third party that was never explained, the volume that is inconsistent with the stated business.
  5. What the review found. Documents pulled, public records checked, internal systems queried, the customer's explanation and whether it was corroborated. Include findings that cut the other way, because a narrative that only lists inculpatory detail reads as selective.
  6. What happened next. Escalation, contact with the customer, account restrictions, relationship decisions, and any prior filings connected to the same activity.
  7. The closing assessment. Two to four sentences on why the activity was suspicious and what decision followed. This is the paragraph a reviewer will quote, so it should be precise and free of rhetoric.

Two formatting habits make this easier to read. Use one paragraph per idea rather than one paragraph per document, and keep the chronology strictly chronological inside each section, even when the review itself was not conducted in that order.

What to include

Specifics are what make a narrative defensible, because they convert an opinion into a record. The following belong in the narrative whenever they are available.

The figures, exactly as they appear in the records

Dates, amounts, currency, instrument types, account numbers where appropriate, and the direction of each movement. Round numbers are a warning sign in a narrative: if the analyst writes approximately forty thousand dollars across the period, the reviewer has to do arithmetic the analyst should have done. State the total, state the count, and state the period.

The relationship between the parties

Whether the counterparties are the customer's own accounts, unrelated third parties, entities with common ownership, or parties the customer has declined to explain. This is often the single most useful paragraph in the narrative, because it is where a legitimate explanation and an illegitimate one diverge.

The customer's own account of the activity

What the customer said, when they said it, and whether the institution was able to verify it. If the customer offered an explanation that was checked and did not hold up, say what was checked. If the customer could not be reached, say that instead of writing around it.

Prior filings, prior alerts, and prior reviews

If the subject has been the subject of earlier filings or earlier escalations, the current narrative should say so, with dates, and should describe what the earlier activity was. A reviewer reading the fifth narrative on the same customer should not have to assemble the history themselves.

Activity after the review period

Whether the pattern continued, stopped, changed, or resumed after contact with the customer. Continuation after contact is materially different from continuation before it, and a narrative that ends at the review period leaves the most relevant fact unstated.

Documents and sources reviewed

Statements, onboarding records, beneficial ownership documentation, invoices, public filings, sanctions and adverse media screening results, and any internal notes. Naming the sources makes the review auditable rather than assertive.

Negative findings

What the institution looked for and did not find. If no business purpose could be documented after two requests, say so. If screening produced no match, say so. Negative findings are evidence of a review having happened, and their absence is conspicuous.

What to avoid

Most narrative defects are habits rather than mistakes, which is why they persist across teams.

Copying the alert

This is the most common failure and the easiest to spot. Monitoring alerts are written in system language, organized by rule rather than by event, and often include transaction counts that repeat across alerts. A narrative built from the alert tells the reviewer what the software already said, in the same words the software used, and adds nothing. Start from the records, not from the alert text.

Conclusions about wrongdoing

The narrative explains why activity was suspicious. It does not establish that a crime occurred, that funds were proceeds of a specific offence, or that the customer is guilty of anything. Statements of that kind cannot be supported by the underlying records, and they invite exactly the scrutiny the institution does not want. Suspicion is the standard the narrative is written to, and it should say so in its own language.

Labels without facts

Calling a customer high risk, uncooperative, or evasive without describing the conduct that produced the label is an opinion with no evidence attached. The label should follow the facts, not replace them.

Passive constructions that hide who did what

Phrases such as the account was reviewed, or the transaction was considered unusual, leave the reviewer unable to tell who reviewed it, when, and on what basis. Name the role, the date, and the record. The narrative is a record of institutional action, and institutional action has actors.

Chained hedging

It may possibly appear that the activity could potentially be inconsistent with the stated purpose. One qualifier where the evidence is genuinely uncertain is honest. A chain of them signals that the analyst has not reached a view, and a reviewer will treat the whole narrative as unformed.

Boilerplate about the institution's program

Descriptions of the institution's commitment to compliance and the strength of its monitoring do not belong in a case narrative. They add length without adding information, and they read as defensive.

Speculation about intent, motive, or personal characteristics

The narrative records conduct and explains why the conduct is not consistent with what the institution knew. Guesses about what the customer was thinking, where they were born, or what group they belong to have no place in it, and they create risk that has nothing to do with the case.

Common mistakes, and what they look like in practice

A few patterns recur often enough that a reviewer will recognize them within the first paragraph.

  • Figures that do not reconcile. The narrative says eleven transactions and the attached schedule lists nine. The narrative totals a different amount than the reported fields. Every number in a narrative should come from one working schedule, and the schedule should be attached or retained.
  • Timeline jumps. Activity described for March, then a paragraph about June, then back to April. Chronology is the cheapest form of clarity available, and abandoning it makes the whole account feel unreliable.
  • No baseline. Six pages of activity with no statement of what the account normally looked like. Without the baseline, nothing in the narrative is unusual, which means the narrative has not explained anything.
  • No statement of what was checked. The narrative asserts that the explanation was implausible without saying what verification was attempted. Reviewers read that as an assertion rather than a finding.
  • Written before the review finished. Narratives drafted at the point of escalation and never updated tend to describe a review that had not yet happened. Revise after the review, not before it.
  • Unexplained internal references. Case numbers, alert identifiers and system codes that mean nothing outside the institution. If a reference is necessary, gloss it in plain language.
  • One enormous paragraph. Length is not the problem; undifferentiated length is. A dense block of text is harder to review and easier to misread.
  • Closing that argues rather than concludes. A final paragraph that editorializes about the seriousness of the conduct weakens an otherwise factual narrative.

Prior filings and follow-up activity: telling the whole story

The activity in the current review period is rarely the whole story, and a narrative that treats it as if it were creates a false impression of an isolated event. If the subject has been reported before, the narrative should say what was reported, when, and what the earlier concern was. More importantly, it should say how the current activity relates to that history: whether it continues the same pattern, whether it changed after the earlier review, whether it paused and resumed.

Follow-up activity deserves the same treatment. If the pattern continued after the institution contacted the customer, that is a different fact from a pattern that stopped on contact, and the narrative should record which one happened, with dates. If the relationship was exited, say when and how. If the account remained open and the activity stopped for reasons the institution cannot explain, that is also a fact, and it belongs in the narrative rather than in a separate note nobody reads.

The practical reason is continuity. A reviewer who reads four narratives on the same customer should be able to see a coherent arc across them without doing any research, because that arc is often the difference between activity that looks explainable in isolation and activity that does not.

When the facts are incomplete

Narratives are frequently written before every fact is available, and the temptation is to close the gaps with plausible inference. That is the one shortcut worth refusing. A named gap is a limitation the reviewer can weigh. A guessed fact is a defect that will be found the moment someone compares the narrative to the file.

Three habits keep an incomplete narrative honest. First, state what is unknown and why: records not obtained, customer not reachable, third party outside the institution's relationship. Second, state what was attempted, so the gap reads as a limitation rather than an omission. Third, state what the unknown fact would change, if anything. Saying that the institution could not verify the source of funds, and that verification would materially affect the assessment, is a stronger position than a confident sentence that later turns out to be unsupported.

It is equally important not to over-hedge. If the available facts support a view, state the view. A narrative that lists uncertainties without ever reaching an assessment leaves the reviewer to make the decision the institution was required to make.

Quality control before the narrative is filed

The most effective control is also the simplest: a second reader who did not work the case, reading the narrative without the transaction history in front of them. Their job is not to rewrite it but to answer four questions. Can I follow the sequence? Can I point to the sentence that explains why this was unusual? Can I see what the institution actually did? And does the closing match what the narrative established?

Four checks should happen before that read, because they catch most mechanical defects.

  • Arithmetic and dates. Totals, counts and periods reconcile with a single working schedule, and the chronology has no unexplained gaps.
  • Names and identifiers. The subject is named consistently, including entity variations, trade names and any earlier names found in the records.
  • Consistency with the reported fields. Dates, amounts and activity types in the narrative match what was entered in the filing itself. A mismatch between narrative and fields is one of the fastest ways to damage confidence in a file.
  • A read-aloud pass. Reading the narrative aloud surfaces run-on sentences, missing antecedents and the places where the reasoning actually skips.

It helps to write the review step into the program rather than leaving it to individual diligence. An examiner who asks how narratives are quality controlled is asking whether the institution has a repeatable process, and the answer is much stronger when the process is described in the procedures and can be evidenced with review notes. Our analysis of AML program documentation and what examiners want to see covers how those procedures and evidence trails fit together.

Where this fits in a program

Narrative quality is downstream of everything else. If the risk assessment is stale, the baseline paragraph is guesswork. If customer due diligence records are thin, the reviewer cannot tell whether the activity was consistent with the customer. If escalation is undocumented, the narrative cannot describe the review. Improving the narrative in isolation therefore has limits, which is why it is worth checking the surrounding controls at the same time.

The 24-item AML compliance checklist is a reasonable starting point for that check, because it covers the program elements that feed into every filing decision. For a structured view of where the gaps sit, the free BSA/AML Audit Simulator asks questions across the five BSA pillars and produces a scored report, which turns a general concern about documentation into something specific you can act on. Where the gap is broader than narratives, or where a banking partner has asked for documentation that does not exist yet, the compliance services page sets out what Soflo Consulting does across training, policy review, risk assessment support and independent review guidance.

This article is general educational information about writing SAR narratives. It is not legal advice, and it does not determine whether a particular transaction is reportable. Applicable thresholds, timing, form requirements and confidentiality obligations depend on current FinCEN guidance and on the institution's own program, and the decision to file or not to file should be made with a qualified compliance professional or counsel. Filing is not a safe harbor, and neither is declining to file.

Frequently Asked Questions

There is no target length, and any number offered as a rule is a convention rather than a requirement. Length should follow the number of facts a reader needs in order to follow the sequence and reach the same view. A simple case with a short review can be handled in a few paragraphs. A case involving multiple entities, several counterparties and a long review period will need more. What matters is that every paragraph carries information a reviewer would otherwise have to look up. Padding a narrative to reach a word count makes it harder to review, not safer.

Tags

SAR narrative writingsuspicious activity report narrativeBSA AML documentationSAR filing quality controlFinCEN SAR requirementsSAR escalationtransaction monitoring alertsAML narrative structureBSA compliance programSAR review process
SC
Soflo Consulting

AML/BSA Compliance Platform and Professional Services · Soflo Consulting

LinkedIn
Soflo Consulting Team
Soflo Consulting

Soflo Consulting provides an AML/BSA compliance platform and professional services for regulated businesses across the United States, including online training with certificate tracking, policy review, risk assessment support and annual compliance support. Articles published under this byline are general educational information and are not legal advice.

BSA/AML ProgramsCompliance TrainingRisk AssessmentPolicy Review
In This Article

8 sections

Key Takeaways

  • 1A SAR narrative has to establish what happened, why it was unusual for that customer, and what the institution did about it.
  • 2The alert is a flag, not a draft. Narratives built from alert text repeat what the software already said and add nothing.
  • 3Describing activity and concluding from it are different acts. Conclusions belong in the narrative only where the record supports them and they are framed as the institution's assessment.
  • 4The baseline is the section reviewers weigh most heavily, and it is the section most often missing.
  • 5Prior filings, prior escalations and activity after the review period belong in the story, with dates, so the history stays coherent.
  • 6Name the gaps rather than filling them. A stated limitation can be weighed; a guessed fact will be found.
  • 7Mechanical defects such as totals that do not reconcile with the reported fields damage confidence faster than imperfect prose.
  • 8A second reader who did not work the case is the most effective quality control step available.

Need Expert Guidance?

Put these insights into action. Talk to a compliance professional at Soflo Consulting compliance specialist.

Stay Ahead of Compliance

Get FinCEN updates, BSA/AML guidance, and federal compliance news delivered to your inbox - no fluff.

We'll send a 6-digit code to confirm your address.

No spam. Unsubscribe any time.

Category

FinCEN & Regulation
Continue Reading

You Might Also Like

Handpicked articles to deepen your compliance knowledge

Browse all insights

Explore the full Insights library

50+ articles on BSA/AML compliance, FinCEN requirements, and industry-specific guidance

View all articles