Skip links

Transaction Monitoring for AML: Why Static Rules Miss What Perpetual KYC Catches

AML · Transaction Monitoring
Last updated: November 6, 2026
· 11 min read

On October 10, 2024, FinCEN assessed a $1.3 billion penalty against TD Bank — the largest ever levied against a depository institution in U.S. Treasury and FinCEN history. The bank’s transaction-monitoring system had been running, on paper, for years. The catch: the Department of Justice found that between January 1, 2018 and April 12, 2024, the system never evaluated 92% of the bank’s domestic transaction volume — roughly $18.3 trillion moving through accounts it was built to watch, including most check activity and nearly all domestic ACH transfers.

A transaction-monitoring rule can only evaluate the fields it’s configured to read. Zavia resolves the owner behind a counterparty against government registries in 173 countries, so an alert isn’t scored against a blank entity field — it’s scored against who actually controls the account on the other end.

01A rule that works on paper and a rule that works in practice are not the same claim

TD Bank’s system wasn’t broken in the sense of crashing or returning errors. It ran. It generated alerts. What it didn’t do, according to the DOJ’s findings, was look at most of the transactions passing through the bank at all — a configuration gap, not a software failure. That distinction matters for anyone evaluating their own program: a monitoring system can pass every uptime check and still miss nearly everything, because the thing that fails first is almost never the engine. It’s the scope of what the engine was told to watch.

TD Bank: share of domestic transaction volume left unmonitored January 1, 2018 – April 12, 2024

92% unmonitored 8%

Most domestic ACH transfers and the large majority of check activity were never routed through automated alert review, per the Department of Justice — roughly $18.3 trillion in transaction activity over the period.

Source: U.S. Department of Justice and FinCEN, TD Bank enforcement findings, October 10, 2024.

Source: U.S. Department of Justice and FinCEN, TD Bank enforcement findings, October 10, 2024.

Most transaction-monitoring programs still run primarily on rule-based logic: a transaction over $10,000 trips a reporting review, three transfers to one account inside 24 hours trip a velocity rule, a wire to a listed high-risk jurisdiction trips a geography rule. Static rules like these are explainable to an examiner and cheap to run. They are also easy to defeat, because a rule that fires at $10,000 tells the person moving money exactly where to stop — at $9,999.

Behavioral and anomaly-based monitoring works on a different mechanism. Instead of a fixed threshold, it builds a baseline for how a specific customer, a peer group, or a payment corridor normally behaves, then scores deviation from that baseline — including deviation that never crosses any single static limit. FATF’s Recommendation 20 already assumes detection doesn’t stop at a dollar figure: its Interpretive Note extends the reporting duty to attempted transactions, regardless of the amount involved.

02Rule-based vs. behavioral monitoring: two different failure modes

Neither approach replaces the other on its own. Most mature programs run both, with rules handling the clear-cut cases and behavioral scoring handling the ambiguous ones. The two fail in opposite directions, which is the part worth separating out before comparing them in a table.

Detection style vs. data dependency

RULE-BASED Fixed thresholds & velocity caps Static per-customer baseline Gamed by staying just under limits Low dependency on entity/UBO data Tends toward high false-positive volume

BEHAVIORAL / ANOMALY Dynamic peer & self baseline Weighs counterparty risk tier Flags coordinated, below-threshold patterns High dependency on UBO/ownership data Fewer, higher-precision alerts — needs clean data

Characteristics synthesized from FCA thematic-review findings on transaction-monitoring calibration and FATF Recommendation 20 guidance on reportable patterns.

Characteristics drawn from the FCA’s Financial Crime Guide (FCTR 4) and its 2018 thematic review of transaction monitoring in the e-money sector (TR18/3), and from FATF’s Interpretive Note to Recommendation 20 on reportable patterns.
Dimension Rule-based / threshold monitoring Behavioral / anomaly-based monitoring
Detection style Fixed thresholds, velocity caps, static geography and product lists Dynamic baselines per customer, peer group, or network; scores deviation
Typical failure mode Alerts on any transaction crossing the line, context aside — or, misconfigured, doesn’t evaluate the transaction at all Misses the signal entirely if the ownership or risk-tier data feeding it is stale or wrong
Data dependency Low — works on transaction fields alone High — needs current customer, ownership, and counterparty risk data
Vulnerable to Structuring: staying just under the reporting line Data poisoning: a UBO change the model never heard about
Regulatory posture A compliance floor FATF and FinCEN both expect; not, by itself, a ceiling Increasingly expected as a complement, per the FCA’s own calibration findings

03Structuring: the pattern a static rule is built to miss

31 U.S.C. § 5324 makes it a federal offense to structure transactions specifically to evade the $10,000 Currency Transaction Report threshold set at 31 CFR 1010.311 — regardless of whether the underlying funds are legitimate. The statute criminalizes the evasion, not the money.

Here’s what that looks like inside a system that evaluates one transaction at a time. Six deposits, each between $9,100 and $9,800, land across three days into the same account, originating from accounts that trace back to one controlling party. Individually, every deposit clears a $10,000 rule with room to spare. Only in aggregate — across accounts, across days, tied back to one beneficial owner — does the $56,000 total look like what it is.

Threshold-gaming: deposits structured under the $10,000 CTR line

$10,000 CTR threshold

DAY 1 $9,400

DAY 1 $9,100

DAY 2 $9,700

DAY 2 $8,900

DAY 3 $9,600

DAY 3 $9,300

AGGREGATE 6 deposits $56,000 one counterparty, 3 calendar days

Illustrative pattern. CTR threshold and the structuring prohibition are set out at 31 CFR 1010.311 and 31 U.S.C. § 5324; each deposit alone clears a static rule, only cross-transaction aggregation flags it.

Illustrative pattern, not a real case. The $10,000 CTR threshold and the structuring prohibition are set out at 31 CFR 1010.311 and 31 U.S.C. § 5324.

FinCEN’s own October 2025 SAR FAQs state that a transaction, or series of transactions, at or near the CTR threshold does not by itself require a SAR — a bank needs knowledge, suspicion, or reason to suspect the activity is designed to evade reporting. Put plainly: FinCEN will not let an examiner cite a bank for noticing that $9,999 happens to be a convenient number. A static rule can confirm the deposit was under $10,000. Whether six of them from one controller in three days amounts to structuring is a question about relationships the rule was never given the data to answer.

A counterparty’s ownership doesn’t stay fixed between onboarding and the transaction that triggers a review. Zavia’s company-monitoring API flags ownership and control changes as they’re filed, closing the gap a point-in-time KYB check leaves open — so a six-deposit pattern across “unrelated” accounts doesn’t stay unrelated in the system that’s supposed to catch it.

04Why the transaction only makes sense once you know who’s behind it

A monitoring system that treats each account as its own island will never connect a $9,500 deposit from “Vendor Co.” to the two other accounts the same beneficial owner already controls, because that link doesn’t live in the transaction record. It lives in KYB and UBO data — Know Your Business and Ultimate Beneficial Owner, the entity-side profile that has to be matched against the transaction in real time, not referenced once at onboarding and left alone.

A risk tier set at onboarding and never touched again is a risk tier that’s already wrong by the time a transaction pattern starts to look odd. This is the specific argument for monitoring beneficial ownership on an ongoing basis rather than on a periodic review cycle: the entity a bank onboarded eighteen months ago, under one set of owners, is not reliably the same entity transacting today. Perpetual KYC is the term for closing that gap continuously instead of on a fixed schedule — and it’s the piece a static transaction-monitoring rule structurally cannot supply on its own, because a rule only sees the transaction, never the ownership change behind it.

05From alert to SAR: the filing clock and what triggers it

Once an alert escalates into genuine suspicion, 31 CFR 1020.320 requires a bank to file a Suspicious Activity Report (SAR) no later than 30 calendar days after the date of initial detection of facts that may constitute a basis for filing. If the bank hasn’t identified a suspect by then, it may take an additional 30 days to try — but the total runs no longer than 60 calendar days from initial detection. There is no equivalent grace period for pulling the counterparty’s ownership data; that has to happen inside the same window.

Internationally, FATF Recommendation 20 sets the baseline that national SAR-equivalent regimes implement: it requires financial institutions, by law, to promptly report suspicions to their national Financial Intelligence Unit when they suspect, or have reasonable grounds to suspect, that funds are proceeds of crime or linked to terrorist financing. The UK implements the same underlying obligation through its own suspicious activity reporting regime; the FCA’s own Financial Crime Guide separately found that automated transaction-monitoring effectiveness varies considerably between firms, usually because of calibration rather than a lack of technology.

What the clock actually measures is whether the underlying data can support the decision in time. If an analyst has to manually chase down who controls a flagged counterparty before judging whether a pattern is structuring, thirty days disappears fast — and the TD Bank case is the extreme version of what happens when the data behind the alert was never current to begin with.

06Building a monitoring program against outcomes, not alert volume

Five practices separate programs that keep pace with their own alert volume from the ones that don’t:

  1. Tune against disposition rates, not alert counts. A rule generating thousands of alerts and zero escalations isn’t a safety net — it’s noise wearing a compliance label.
  2. Feed current ownership and risk-tier data into the scoring layer. A transaction from a counterparty re-tiered as high-risk last week should never score the same as an identical transaction from a stable one.
  3. Aggregate across accounts and entities, not within one account at a time. Structuring and layering are, by construction, patterns that span more than one account; a rule scoped to a single account cannot see them.
  4. Treat beneficial-ownership refresh as a monitoring input, not a separate compliance task. When ownership changes, the transaction baseline should change with it, not six months later at the next periodic review.
  5. Document the rationale behind every rule’s threshold. “That’s always been the number” does not survive an exam — TD Bank’s consent order names nearly a decade of underinvestment in the monitoring system itself as a central finding.

None of this argues for abandoning rule-based monitoring. Static rules stay fast, explainable, and cheap to run, which is why they still catch the obvious cases. The argument is against mistaking the rules a bank can explain for the risks it has actually covered.

07What to check this quarter

Four checks are worth running against an existing setup before an exam cycle runs them instead:

  1. Aggregation scope. Confirm whether the system evaluates transactions across related accounts and entities, or only within one account at a time.
  2. Ownership visibility at the point of alert. Confirm an analyst can see a counterparty’s current beneficial owner and risk tier from inside the alert screen, not in a separate system they have to go find.
  3. Threshold re-justification. Pull the date each rule’s threshold was last reviewed, not just carried forward from last year’s configuration.
  4. Ownership-change latency. Measure how long a counterparty’s ownership change takes to reach the monitoring layer — hours, or the next scheduled KYB refresh.

For most programs, the honest answer to that last question is “whenever the next refresh happens to fall,” which can run months behind the actual ownership change. That gap is what continuous ownership monitoring is built to close, and it’s worth reviewing alongside rule-tuning work rather than as a separate project.

Closing that gap is a data problem before it’s a policy problem. See what Zavia’s registry-sourced plans cover before the next exam cycle makes the question urgent rather than routine.

Frequently asked questions

What is transaction monitoring in AML?

It is the ongoing review of customer transactions — deposits, withdrawals, wires, card activity — to detect patterns indicating money laundering, terrorist financing, or other financial crime, so the institution can investigate and, where warranted, file a Suspicious Activity Report (SAR) or Suspicious Transaction Report (STR).

Why do rule-based transaction-monitoring systems generate so many false positives?

A static rule alerts on any transaction crossing a fixed threshold or velocity limit, regardless of context. It cannot distinguish a customer’s routine payroll run from a one-off anomaly that happens to cross the same line, so it alerts on both equally.

Is structuring illegal even if the money is legitimate?

Yes. 31 U.S.C. § 5324 makes it a federal offense to structure transactions to evade the $10,000 Currency Transaction Report threshold, regardless of whether the underlying funds are legal. The statute targets the deliberate evasion of the reporting requirement, not the source of the money.

How long does a bank have to file a SAR after detecting suspicious activity?

31 CFR 1020.320 requires a bank to file a SAR no later than 30 calendar days after the date of initial detection of facts that may constitute a basis for filing. If the bank hasn’t identified a suspect, it may take an additional 30 days to do so, for a total cap of 60 calendar days from initial detection.

Does being close to a reporting threshold automatically require a SAR?

No. FinCEN’s October 2025 SAR FAQs state that a transaction, or series of transactions, at or near the CTR threshold does not by itself require a SAR. The bank needs knowledge, suspicion, or reason to suspect the activity is structured to evade reporting — proximity to the number alone is not evidence of intent.

How does beneficial ownership data improve transaction monitoring?

It supplies context a transaction record cannot provide on its own: whether a “new” counterparty is actually linked, through common ownership, to accounts already under review, and whether a counterparty’s ownership changed shortly before the activity that triggered the alert.

What’s the difference between a SAR and an STR?

They are the same obligation under different national labels. The US Bank Secrecy Act calls it a Suspicious Activity Report (SAR). FATF Recommendation 20 sets the international standard, which most other jurisdictions implement as a Suspicious Transaction Report (STR) filed with their national Financial Intelligence Unit.

Can behavioral monitoring fully replace rule-based monitoring?

Not in most programs. Rule-based monitoring stays useful for clear-cut, easily audited cases. Behavioral and anomaly-based monitoring adds coverage for coordinated or below-threshold activity a static rule structurally cannot see, but it depends on current customer and ownership data to work, which is why most institutions run both.

Primary sources cited in this article:
FinCEN, “FinCEN Assesses Record $1.3 Billion Penalty against TD Bank,” press release, October 10, 2024, fincen.gov · U.S. Department of Justice, TD Bank N.A. plea agreement filings, October 2024 · FinCEN, SAR FAQs, October 2025, fincen.gov · 31 CFR 1020.320, SAR filing deadlines, eCFR · 31 CFR 1010.311, Currency Transaction Report threshold, eCFR · 31 U.S.C. § 5324, structuring prohibition · FATF, Recommendation 20 and its Interpretive Note, Reporting of Suspicious Transactions · FCA, Financial Crime Guide (FCTR 4) and Thematic Review TR18/3, transaction monitoring in the e-money sector, 2018.

This website uses cookies to improve your web experience.