Module 03 · Fraud Patterns and System-Enforced Controls

Letting the System Enforce Your Controls — and Verifying a Feature Before You Rely on It

A rule the software refuses to break is worth more than a rule everyone agreed to follow. The catch is knowing which kind you actually have.

6 min readLesson 5 of 5 in this moduleUpdated 2026-09-22

Key takeaways

  • A system-enforced control cannot be skipped on a busy day; a policy control depends on somebody remembering.
  • Move the controls that matter most into the software where you can, and keep a named routine for the ones you cannot.
  • Never assume a feature exists because it sounds like something accounting software should do — check, then test.
  • Test a control the way it would fail: try the thing you expect to be blocked, with a real user account, not the owner's.
  • A warning you can click past is not a block; know which of your controls are warnings.
  • Write down what you tested, what happened, and the date — a feature's behaviour can change with an update.

There are two kinds of control in any business. One asks a person to do something: check the documents match, count the drawer, get a second signature. The other is a rule the software will not let anybody break, whatever kind of day they are having.

Policy controls are flexible and cover judgement — whether a purchase was genuinely urgent, whether a supplier looks real. System controls are rigid and cover repetition. The mistake worth avoiding is believing you have the second kind when you have neither: a control everybody assumes the software is applying, which it is not, and which therefore has no compensating routine either.

What belongs in the software

Move a control into the system when it is high-frequency, easy to skip under pressure, or expensive to reverse:

ControlWhy it belongs in the software
Who can open which part of the systemApplies to every action, every day, with no chance to reconsider
Locking a closed accounting periodProtects figures you have already reported to a bank, an owner or a tax authority
Preventing deletion of referenced recordsA deleted master record can orphan history that cannot be rebuilt
Recording who did what, with individual loginsEvery later review depends on this being reliable
Posting cost of goods sold with the saleKeeps margin comparable month to month without manual work

And what stays with people: judging whether an emergency was real, verifying a new supplier exists, deciding a write-off is justified, counting physical things. No package evaluates those, and buying a more expensive one does not change it.

Do not assume — check

The instinct that causes trouble is reasoning from plausibility: it is accounting software, so obviously it blocks payments over the customer's credit limit or obviously it keeps a full history of every edit. Some packages do. Others warn. Others do neither, and behaviour differs between editions and versions of the same product.

The cost of the assumption is specific: an assumed control gets no compensating routine. A control you know you do not have gets a weekly report or a second signature instead. The gap you do not know about gets nothing.

Three possible answers, three different controls

When you try the action you expect to be prevented, exactly one of three things happens. Refused — a real block. Allowed with a warning — a prompt, which busy people click through. Allowed, recorded afterwards — a detective control, useful only if someone actually reads the record. Know which one each of your controls is.

How to test a feature in ten minutes

  1. Write the claim as a sentence. "A sales invoice cannot be dated into a closed period." Vague claims cannot be tested.
  2. Use an ordinary account. Owner or administrator accounts often bypass restrictions that apply to everyone else, so testing as the owner tells you almost nothing about what staff can do.
  3. Try it in a safe way — on sample data, on a copy of your file, or with a transaction you will immediately reverse.
  4. Record which of the three answers you got, with the date and the version you tested.
  5. Where the answer is not "refused", name the compensating routine and who owns it.
  6. Re-test after significant updates, and as part of your twice-yearly access review.
Two questions worth testing in your own system

Two controls that owners commonly assume exist, and that vary widely between packages: blocking a new order for a customer who is over their credit limit, and keeping a permanent, viewable history of every edit and deletion of a transaction. Do not take either on trust — write each as a sentence, test it, and write down which of the three answers you got. Whatever the result, you will know whether you need a weekly report instead.

What to do with the controls you cannot enforce

Each unenforced cell in your access matrix needs a named routine — not an intention. The three that cover most cases:

  • A report someone reads on a schedule. Cancelled documents this month, refunds by cashier, entries dated into a closed period.
  • A second signature on the small number of actions that are both irreversible and easy to abuse: write-offs, stock adjustments, supplier bank changes.
  • A periodic count or confirmation that compares the record to something physical or external: stock counts, customer balance confirmations, bank reconciliation.

That is the whole design principle of this course in one line: enforce what you can, review what you cannot, and know at all times which is which.

Doing this in Reyuko

Three controls Reyuko enforces in the system rather than asking you to remember:

Closed periods stay closed. Once an accounting period is locked, the posting engine refuses entries dated into it — the refusal comes from the core that writes the journals, not from a dialog in the interface, so it applies regardless of which screen the entry came from. Reopening the period is a deliberate act.

Master data that transactions depend on cannot be hard-deleted. If a product, contact, location, account, payment term or unit is referenced by existing transactions, Reyuko blocks the deletion and offers to mark the record inactive instead — so cleaning up a list cannot silently break the history behind it.

Cost of goods sold posts with the sale. The cost leg is written at the same time as the sales journal, consuming stock by your chosen costing method (average, FIFO, LIFO or first-expiry-first-out), and the running cost is visible on the product record. Margin stays comparable month to month without a period-end exercise.

Access itself is enforced through the seven preset roles and per-user PINs described in the previous lesson. What Reyuko does not do is run an approval workflow with value thresholds — your authorization matrix remains a written policy backed by review. Apply the ten-minute test above to any behaviour you plan to depend on, including the ones described here, on the version you actually run.

Turn this into a one-page register
  1. List the controls you believe your software enforces, each written as a testable sentence.
  2. Test each one with an ordinary user account, and record refused, warned or recorded-afterwards.
  3. For everything not refused, name the compensating routine and its owner.
  4. Note the date and software version you tested against.
  5. Re-test after updates and at your twice-yearly access review.

That completes the course. The controls worth having are the ones you can describe in a sentence, test, and check on a schedule — starting, if you have not already, with the two separations in control foundations.

Frequently asked questions

What is a system-enforced control?

A rule the software itself will not allow to be broken — for example refusing an entry dated into a closed accounting period, or preventing the deletion of a record that transactions depend on. It differs from a policy control, which relies on a person remembering to follow it, and from a warning, which only asks the person to reconsider.

Which controls should be enforced by software rather than policy?

The ones that are high-frequency, easy to skip when busy, or irreversible: access limits, period locking, and anything that would silently corrupt history. Judgement-based controls — whether a purchase was genuinely an emergency, whether a supplier is real — stay with people, because no system can evaluate them.

How do I test whether a feature actually blocks something?

Try it in a safe way with an ordinary user account rather than the owner account, and see what happens. Ask three questions: is the action refused outright, is it allowed with a warning, or is it simply recorded afterwards? Each answer implies a different control, and only the first is a block.

Why shouldn't I rely on a feature that sounds standard?

Because "sounds standard" is not a specification. Packages differ widely in what they enforce, what they merely warn about, and what they leave to the user. Assuming a control exists is worse than knowing it does not — an assumed control gets no compensating routine, so the gap is invisible until it costs something.

How often should software controls be re-tested?

After any significant update, and at least once a year as part of your access review. Software behaviour changes between versions, and a control you verified two years ago may now warn where it used to block, or apply to a different set of users.

Controls the posting engine enforces

Reyuko locks closed periods in the core, blocks deletion of master data that transactions reference, and gives each user their own PIN.

Download Reyuko freeSee what's included