NEWReplay any past check against today, with the difference shown line by line. See how ›
InackGuard
StatusSign inBook a demoTry free
Platform › Transaction rules
Transaction rules

Test a rule before it raises one alert.

Eight rule types for payment firms and money-service businesses. Backtest each over the past 90 days, see what it would have flagged, then switch it on, with a reason and approver for every threshold change.

Free credits to start. Results are for review by a person, not decisions.
InackGuard · Transaction rules · backtest before go-live
STRUCTURING · 90-DAY BACKTEST
9,000 to 9,999 · 3 in 72 h
12 WOULD FLAG
threshold
JulAugSepOct
Would have flagged12 cases · 0.4% of customers
At 9,700 instead7 cases
Nothing created during the testRecorded in the audit trail
Approve with a reason. “Threshold 9,500, window 72 h, approved by M. Okafor, reason: regulator guidance” is kept with the rule.
VELOCITY · 90-DAY BACKTEST
More than 8 payments in 24 h
31 WOULD FLAG
threshold
JulAugSepOct
Would have flagged31 cases · mostly one merchant
Suggested exclusionKnown payroll account
After exclusion9 cases
Too noisy at this threshold. The backtest shows where the alerts come from before any analyst sees one.
DORMANT ACCOUNT · 90-DAY BACKTEST
No activity 180 days, then a transfer
4 WOULD FLAG
threshold
JulAugSepOct
Would have flagged4 cases
Median time to close in the test5.3 h
Rule statusDraft · not live
Switch on. Every threshold change keeps who, why and when; the next monthly review shows what it raised.
Eight rule types
Reason + approver kept

What is transaction monitoring software?

Transaction monitoring software watches payments for patterns that a compliance policy says to look at: amounts just under a reporting threshold, a quiet account that suddenly moves money, payments to high-risk countries, a first large payment to a new beneficiary. Each pattern is a rule with thresholds. The hard part is setting thresholds that catch risk without burying analysts, which is why a backtest over your own past activity matters more than a long rule library. InackGuard runs the test first, records who approved the threshold, and only then raises alerts.

Guessed thresholds bury analysts or miss risk.

Set a threshold too low and alerts pile up; too high and real activity slips through. A backtest on your own past activity shows the workload and the cases before you commit.

THE AUDIT TRAIL OF ONE RULE
15 Oct 09:12Threshold 9,500 → approvedM. Okafor
15 Oct 09:10Backtest run · 90 days · 12 would flagR. Lindqvist
14 Oct 16:40Draft threshold 9,700R. Lindqvist
14 Oct 16:38Rule created · structuringR. Lindqvist
Every change to a threshold keeps the reason and the approver. The backtest is a record, not a side effect.

How it works

  1. STEP 1
    Upload transactions or connect by API
  2. STEP 2
    Choose rules and draft thresholds
  3. STEP 3
    Backtest over the past 90 days
  4. STEP 4
    Approve, switch on and review monthly

What it does

Six capabilities, each one visible in the report it produces.

Eight rule types

Structuring, velocity, pass-through, dormant activation, high-risk country, new large beneficiary, cash threshold and activity above expectation. Each has a short explanation of what it detects, the parameters it takes and the regulatory reference it serves.

  • Each type documented with an example and a reference
  • Parameters validated before a rule can be saved
  • Start a rule from the typology catalogue with the type preset

RECORDEDThe rule type, version and parameters of every rule, past and present.

StructuringVelocityPass-throughDormant activationHigh-risk countryNew large beneficiaryCash thresholdAbove expectation
Customers flagged12
Cases it would open4
Transactions involved47
Analyst hours, estimated6

90-day backtest

A backtest runs a draft rule over the last ninety days of your own transactions and reports how many customers it would have flagged, how many cases it would have opened and a sample of them. Nothing is created; the result is recorded in the audit trail.

  • Customers flagged, cases it would open, a sample of both
  • Nothing written: no alerts, no cases
  • Every backtest kept in the audit trail with its parameters

RECORDEDThe backtest, its parameters, window and counts.

Governed changes

A threshold change needs a reason and an approver who is not the person proposing it. The old value, the new value, both names and the reason stay on the rule's history, so an examiner sees why a rule was tuned and who agreed.

  • Reason required on every change
  • Approver must differ from the proposer
  • Full history per rule, exportable

RECORDEDEvery version of every rule with reason, proposer and approver.

v1 · threshold 9,000created by A. Rivera
v2 · threshold 9,500reason: 67% false positives
approvedby M. Okafor
v2 activesince 12 Oct
Harbour Point Logisticsexpected 40k/mo · tier 2
Vega Exchangeexpected 900k/mo · tier 3
Alder Supply Coexpected 5k/mo · tier 1

Risk profiles

Each customer carries an expected activity profile: monthly amount, monthly count, a tier. Rules such as activity above expectation read it, so the same payment can be normal for one customer and an alert for another.

  • Expected monthly amount and count per customer
  • Tier feeds rule sensitivity
  • Profiles editable with a reason, like rules

RECORDEDThe profile values used by each alert at the time it was raised.

Recorded tests

Backtests, dry runs and real runs all land in the same audit trail with the same shape: who, when, which rule version, which window, what came out. A tuning decision months ago can be reconstructed from it.

  • Backtests and live runs in one trail
  • Rule version and window on every entry
  • Exportable with the case bundle

RECORDEDEach run, the rule version it used and its counts.

1Backtest · velocity v212 flagged · 02 Oct
2Dry run · structuring v33 flagged · 05 Oct
3Live run · scheduled1 alert · 15 Oct
Scheduled run · 48,100 txdone
Score one transaction2 rules fired · latency reported
Rules evaluated8 active

Fast evaluation

Rules run over the whole book in one scheduled run, and one transaction at a time through the scoring endpoint, so a payment can be checked as it is booked and the answer comes back in the same request, with its latency reported.

  • Scheduled runs over the whole book
  • One-transaction scoring with latency reported
  • No transaction is held by the platform; the answer is for a person

RECORDEDThe run or the scoring request, its latency and what fired.

Specifications

DATA
Transactions with customer, amount, direction, counterparty country and time
RULES
Eight types, configurable thresholds
TESTING
90-day backtest; nothing is created during a test
GOVERNANCE
Reason and approver on every change; review records
OUTPUT
Alerts, cases and an audit trail

Guides that go with it

Written from the product, kept with it.

Reports built on it

Each one shows its price before it runs.

Questions

Do backtests create alerts?+

No. They show what would have been flagged and are recorded in the audit trail.

Can we write custom rules?+

Today you configure the eight rule types; custom rules are on the roadmap.

Who should approve threshold changes?+

Someone other than the person proposing them, as your policy sets.

See transaction rules on your own names.

Free preview, no card.

Create a free accountBook a demo