Regulatory & Conduct Testing
Conduct rules read like principles, not requirements — treat customers fairly, disclose clearly, act in good faith. This lesson teaches you to turn a fair-conduct duty into a concrete, checkable test, and to test that an NZ insurer’s systems treat customers the way the law expects.
1 The Hook
Manuka Life, a fictional NZ insurer, ran a campaign offering a discount to new customers who bought online. The discount was correct, the maths was sound, and the campaign tested clean. It went live.
The problem was who did not get it. The same product, at the same time, was sold to existing customers at the standard price — no discount — while new customers paid less for an identical policy. Worse, the renewal letters for long-standing customers were worded in a way that buried the price increase: the new premium appeared, but the size of the rise and the fact that a cheaper equivalent existed for new buyers were nowhere a customer could easily see. Loyal customers were quietly paying more and being told less.
None of this was a calculation bug. Every premium was arithmetically correct. The failure was a conduct failure: customers in materially the same position were treated differently, and the disclosure was not clear enough for a customer to understand what they were being charged and why. Under the fair-conduct expectations the FMA supervises, that is exactly the kind of outcome an insurer is required to design against — and exactly the kind of thing a tester is now expected to catch.
Here is the lesson hidden in that story. The team tested the numbers and never tested the treatment. They had no test that asked “is a customer in this situation treated fairly, and is the disclosure clear enough to understand?” Conduct testing is testing the outcome and the communication, not just the calculation — and it is now a first-class part of an NZ insurance tester’s job.
2 The Rule
A correct calculation can still be an unfair outcome. Conduct testing checks how the system treats the customer — whether people in materially the same position are treated consistently, and whether disclosure is clear enough for a reasonable customer to understand what they are being charged and why. Test the treatment and the communication, not only the arithmetic.
3 The Analogy
A tradie’s quote with the real cost hidden in the fine print.
Picture a builder who quotes you a headline figure for a deck, but the call-out fee, the disposal cost, and a price rise since last month are tucked into a paragraph at the bottom you are not meant to read. The total might be arithmetically correct — every line adds up — but you walked away not understanding what you agreed to pay. That is not a maths problem. It is a fairness and clarity problem.
Manuka Life’s renewal letter was that quote. The numbers were right; the customer still could not see the size of the rise or that a cheaper equivalent existed. A conduct tester is the person who reads the quote the way the customer would and asks “could a reasonable person understand what they are being charged, and is anyone in the same position being treated worse?” — not just “do the figures add up?”
4 The NZ Conduct Landscape
A tester does not need to be a lawyer, but they do need a working map of who expects what. Keep this conceptual and confirm specifics against current sources — the rules evolve.
The FMA (Financial Markets Authority). The conduct regulator for financial services, including insurers. It supervises how firms treat customers and expects fair conduct to be designed into products and systems, not bolted on. When a system makes or shapes a decision about a customer, the FMA’s interest is in the fairness of the outcome.
CoFI (the Conduct of Financial Institutions regime). A conduct regime that places a duty on relevant financial institutions, including insurers, to treat consumers fairly — supported by a fair-conduct programme that the institution must establish, maintain, and follow. For a tester, CoFI turns “be fair” into something an organisation has to operationalise and evidence, which means its systems must demonstrably support fair treatment.
The Fair Insurance Code. An industry code setting standards for how insurers deal with customers — good faith, clear communication, reasonable timeframes, and a fair claims process. It is the practical articulation of fair dealing that customers and the industry point to.
Prudential supervision (RBNZ under IPSA). Separate from conduct, the Reserve Bank prudentially supervises insurers under the Insurance (Prudential Supervision) Act — concerned with whether the insurer is financially sound enough to pay claims. A tester should know conduct and prudential are different lenses: one asks “is the customer treated fairly?”, the other “can the insurer pay?”
The ACC boundary. NZ’s accident-compensation scheme covers personal injury by accident, which removes much of that risk from private insurers. A tester working on health or life products should understand where the ACC boundary sits, because a product must not imply it covers what the statutory scheme already does, or leave a customer thinking they are covered where they are not.
5 The Fair-Conduct Principle as a Testable Thing
“Treat customers fairly” sounds untestable until you break it into observable behaviours. Fairness, for a tester, becomes a set of properties the system must demonstrate:
- Consistency: customers in materially the same position get materially the same treatment — the Manuka Life failure. Test that two equivalent customers receive equivalent prices, offers, and outcomes, and that any difference traces to a legitimate, disclosed factor.
- No unfair discrimination: differences in treatment must rest on legitimate rating or eligibility factors, not on something the customer would reasonably consider unfair or that the law prohibits.
- Clarity: what the customer is told is clear enough for a reasonable person to understand — the price, the cover, the exclusions, the changes.
- Good faith in claims and complaints: a claim is handled within reasonable timeframes, declines are explained, and a customer can escalate. Unresolved disputes can reach an external dispute-resolution scheme.
- Vulnerability awareness: the system does not exploit or disadvantage customers who may be in a vulnerable position.
Each of these is something a tester can build a scenario for. The shift is from testing whether a number is right to testing whether an outcome is fair and a message is clear — with a defined, checkable expectation for each.
6 Disclosure and Fair Treatment
Disclosure is where conduct most often becomes a concrete tester’s job, because what the customer is shown is in the software. The testable disclosure questions:
- Is the material information present and prominent? The premium, the excess, key exclusions, and any change to terms must be shown where a customer will see them — not buried, and not in a place a reasonable customer would miss (the Manuka Life renewal letter).
- Is a price change communicated clearly? At renewal, the customer should be able to understand that the price has changed and, where expected, by how much — not just see a new number with no signal.
- Are exclusions and limitations clear before the customer commits? A customer should not discover at claim time that something they assumed was covered never was.
- Is the language plain? Disclosure written so only a specialist can parse it is not clear disclosure, even if it is technically accurate.
- Does the disclosed term match the system? What the customer is told must match what the system will actually do at claim — a disclosure that promises more than the policy delivers is a conduct failure.
7 Turning a Duty into a Test
The skill at the heart of conduct testing is translation: taking a principle and producing a scenario with a checkable expectation. The method:
- State the duty in plain words. “Customers in materially the same position are treated consistently.”
- Pick a concrete pair or scenario. Two customers identical on every legitimate rating factor — one new, one existing — buying the same product at the same time.
- Define the fair outcome. They receive the same price and the same offer, or any difference is explained by a disclosed, legitimate factor.
- Make it checkable. Assert the two prices and offers are equal, or that the difference maps to a documented, lawful reason — not to which channel they came through.
- Capture evidence. The two purchase paths side by side, the prices, the disclosures shown, and the reason for any difference — the kind of record the FMA would expect to see.
This turns “treat customers fairly” from an opinion into a test with a pass and a fail. The same method works for disclosure (“the price rise is shown prominently at renewal”), for claims good faith (“a decline is explained and escalation is offered”), and for the ACC boundary (“the product does not imply cover the statutory scheme already provides”).
8 Building Conduct Test Cases
A strong conduct test case names the duty, defines the fair outcome as a checkable expectation, and captures the evidence a regulator would want — not just “the screen worked.”
Here is a worked test case written to catch the exact Manuka Life failure:
Conduct duty: Consistent treatment of customers in the same position
Risk category: New-vs-existing customer price/disclosure unfairness
Scenario: Two customers identical on every legitimate rating factor;
one is a new buyer, one is an existing customer renewing.
Fair outcome: Same product, same price — OR any difference traces to a
documented, lawful, disclosed factor (not the channel alone).
Disclosure check: The renewal clearly shows the premium and any change to it,
prominently; a reasonable customer can understand what they pay.
Expected result: Prices equal, OR difference mapped to a disclosed legitimate reason;
price change shown prominently and in plain language.
Evidence required: Both purchase paths side by side; the two prices and offers;
the disclosures shown; documented reason for any difference.
Traceability: Risk R-02 (inconsistent treatment / unclear disclosure).
Result: [Pass / Fail] — any unexplained difference flagged.
Notice what makes this catch the Hook failure: it states the duty, defines the fair outcome as a checkable assertion (equal price, or a difference traced to a disclosed lawful factor), and includes a disclosure check on prominence and plain language. The evidence is the side-by-side purchase paths a regulator would ask for. That is the difference between a conduct test and “the renewal screen rendered.”
9 Common Mistakes
🚫 Testing the calculation and assuming a correct number is a fair outcome
Why it happens: Numbers have a clear right answer, so they feel like the whole test.
The fix: The Manuka Life premiums were all arithmetically correct and the outcome was still unfair. Test the treatment and the disclosure as well — consistency between equivalent customers, and whether a reasonable person could understand what they are charged.
🚫 Treating “technically disclosed” as “clearly disclosed”
Why it happens: If the information is somewhere on the page, it feels disclosed.
The fix: Information buried where a reasonable customer would miss it is not clear disclosure. Test for prominence and plain language — read the message as a customer who knows nothing about insurance and ask whether they would actually understand the price, the cover, and any change.
🚫 Confusing conduct obligations with prudential ones
Why it happens: Both are “regulation,” so they blur together.
The fix: Conduct (FMA, CoFI, Fair Insurance Code) asks whether the customer is treated fairly; prudential supervision (RBNZ under IPSA) asks whether the insurer is financially sound. They are supervised by different bodies and fail differently. Most customer-facing tests are conduct — keep the two lenses separate so you test the right thing.
🚫 Inventing precise regulatory rules instead of staying conceptual where unsure
Why it happens: A specific clause number sounds authoritative in a test case.
The fix: Conduct rules change and the detail matters. Where you are not certain of an exact requirement, frame the test around the principle — consistency, clarity, good faith — and confirm specifics against the current source. A test built on a misremembered rule is worse than one built on the right principle.
10 Now You Try
Three graded exercises across regulatory and conduct testing. Write your answer, run it for AI feedback, then compare to the model answer.
Read the description of a fictional NZ health-insurance renewal and sales flow below. Identify 3 conduct risks — outcomes that are unfair, unclear, or wrongly framed even if the maths is correct — and name the conduct property each touches (consistency, clarity/disclosure, good faith, ACC boundary).
Existing customers renew at a premium that has risen, but the renewal email shows only the new figure with no indication it changed or by how much. The same cover is offered to new customers online at a lower introductory price, identical in every legitimate respect. The product page implies it covers “accidents and injuries” without distinguishing what the statutory accident scheme already covers. When a claim is declined, the customer is sent a one-line rejection with no reason and no mention of how to escalate.
List 3 conduct risks and the property each touches:
Show model answer
There are at least four real risks here; any three well-explained earns full marks. 1. Price rise not disclosed clearly — the renewal email shows only the new figure with no signal it changed or by how much. Conduct property: clarity / disclosure. Why: a reasonable customer cannot understand what they are being charged or that it rose. Fix: show the change prominently and in plain language. 2. Inconsistent treatment of equivalent customers — new buyers get a lower introductory price for cover identical in every legitimate respect to what existing customers renew at. Conduct property: consistency. Why: customers in materially the same position are treated differently with no disclosed legitimate reason. Fix: equalise or trace any difference to a disclosed lawful factor. 3. ACC boundary blurred — the page implies it covers "accidents and injuries" without distinguishing what the statutory accident scheme already covers. Conduct property: ACC boundary / clarity. Why: a customer may think they are buying cover they already have or be misled about scope. Fix: state clearly what is and is not covered relative to the statutory scheme. Bonus: claim decline with no reason and no escalation — a one-line rejection. Conduct property: good faith in claims. Why: the customer cannot understand or challenge the decision. Fix: explain the reason and offer escalation. The trap: every premium and decline may be arithmetically and contractually correct — the failures are in fairness of outcome and clarity of communication, which a numbers-only test never sees.
The conduct “test” below is just a restated principle — it has no checkable expectation. Rewrite it into a real test case, with these fields: Test ID, Conduct duty, Risk category, Scenario, Fair outcome, Disclosure check, Expected result, Evidence required, Traceability. Use a fictional NZ contents-insurance renewal price increase as the context.
“Make sure the renewal is fair and the customer is treated well. Pass if disclosure is good.”
Rewrite as a checkable conduct test case:
Show model answer
Test ID: CON-DISC-013 Conduct duty: Clear disclosure of a renewal price change Risk category: Price increase not communicated clearly (unclear disclosure) Scenario: An existing contents-insurance customer whose renewal premium has risen from $420 to $510 receives their renewal communication. Fair outcome: The customer can readily understand that the premium has changed and by how much before they decide to renew. Disclosure check: The renewal prominently shows the new premium ($510), the previous premium ($420), and the change (+$90), in plain language, where a reasonable customer will see it — not buried in fine print. Expected result: All three figures and the change are present and prominent; the wording is plain; nothing material about the increase is hidden. A reasonable customer could explain what changed and by how much. Evidence required: The rendered renewal communication; a record showing the old and new premiums; a readability/prominence check (placement and language). Traceability: AI/conduct risk register R-02 (renewal price change not disclosed clearly). What makes it a real test: it names the duty, gives a concrete scenario with numbers, defines a checkable fair outcome (the customer can understand the change), and specifies a prominence-and-plain-language disclosure check with evidence. The original was just the principle restated, with no way to pass or fail.
Design a fair-conduct test plan of 5 test cases for a fictional NZ life insurer’s online sales and renewal system. Each case needs at least: an ID, the conduct duty, a checkable acceptance criterion, and the evidence required. Cover consistent treatment of equivalent customers, clear disclosure of a price change, exclusions disclosed before commitment, good faith on a declined claim (reason + escalation), and the ACC boundary not being overstated. Keep regulatory specifics conceptual.
Show model answer
CON-01 | Conduct duty: consistent treatment of equivalent customers | Acceptance criteria: two customers identical on every legitimate factor receive the same price and offer, OR any difference traces to a documented, disclosed, lawful factor (not the channel alone) | Evidence required: both purchase paths side by side; the two prices/offers; reason for any difference CON-02 | Conduct duty: clear disclosure of a price change | Acceptance criteria: the renewal shows the new premium, the prior premium, and the change prominently and in plain language | Evidence required: rendered renewal; old/new premium record; prominence and readability check CON-03 | Conduct duty: exclusions disclosed before commitment | Acceptance criteria: key exclusions and limitations are shown clearly before the customer confirms purchase, not only in post-sale documents | Evidence required: pre-commitment screen showing the exclusions; placement evidence CON-04 | Conduct duty: good faith on a declined claim | Acceptance criteria: a decline states a clear reason and offers a route to escalate / external dispute resolution; handled within a reasonable timeframe | Evidence required: decline communication with reason and escalation; timestamp vs the service standard CON-05 | Conduct duty: ACC boundary not overstated | Acceptance criteria: product wording does not imply cover the statutory accident scheme already provides; scope relative to the scheme is clear | Evidence required: product wording; a clarity check against the stated scheme boundary Strong plans: each case names a duty, has a checkable criterion, names concrete evidence, and together they cover consistency (CON-01), price-change disclosure (CON-02), pre-commitment exclusions (CON-03), good faith in claims (CON-04), and the ACC boundary (CON-05). Weak plans say "make sure it is fair" five times with no checkable outcome — that is the difference being marked. Keep regulatory detail conceptual where unsure.
11 Self-Check
Click each question to reveal the answer.
Q1: Why can a correct calculation still be a conduct failure?
Because conduct is about the treatment and the communication, not the arithmetic. Every Manuka Life premium was correct, yet equivalent customers were treated differently and the price rise was not disclosed clearly. A right number can sit inside an unfair outcome or an unclear message, and conduct testing checks those, not just whether the figure adds up.
Q2: How do you turn “treat customers fairly” into a test?
State the duty in plain words, pick a concrete scenario (two customers identical on every legitimate factor), define the fair outcome (same price, or a difference traced to a disclosed lawful factor), make it a checkable assertion, and capture the side-by-side evidence a regulator would want. That turns a principle into a test with a pass and a fail.
Q3: What is the difference between conduct and prudential supervision?
Conduct — supervised by the FMA, with CoFI and the Fair Insurance Code — asks whether the customer is treated fairly. Prudential supervision — the RBNZ under IPSA — asks whether the insurer is financially sound enough to pay claims. Different bodies, different lenses. Most customer-facing tests a tester writes are conduct, so keep the two separate.
Q4: When is information “disclosed” but not “clearly disclosed”?
When it is technically present but buried, or written so only a specialist could parse it. Clear disclosure means the material information — price, change, cover, exclusions — is prominent and in plain language, where a reasonable customer who knows nothing about insurance would see and understand it. Test for prominence and plain language, not just presence.
Q5: Why stay conceptual when you are unsure of an exact regulatory rule?
Because conduct rules change and the precise detail matters; a test built on a misremembered clause is worse than one built on the right principle. Where you are not certain, frame the test around the principle — consistency, clarity, good faith — and confirm the specifics against the current source before stating them as fact.
12 Interview Prep
Real questions asked in NZ QA interviews for insurance roles. Read the model answers, then practise your own version.
“How do you test something as vague as ‘treat customers fairly’?”
I translate the principle into observable behaviours and build a scenario with a checkable expectation. For consistency, I take two customers identical on every legitimate rating factor and assert they get the same price and offer, or that any difference traces to a disclosed lawful factor — not just the channel they came through. For disclosure, I read the renewal as a customer who knows nothing about insurance and check the price change is prominent and in plain language. For claims good faith, I check a decline gives a reason and an escalation route. Each becomes a test with a pass, a fail, and the side-by-side evidence a regulator would expect — that is how fairness stops being an opinion.
“A customer complains that new buyers pay less than they do for the same cover. Is that a defect?”
It can be a conduct defect even if every premium is calculated correctly. My first step is to compare the two customers on every legitimate rating factor — if they are materially the same and the only difference is new-versus-existing, that is inconsistent treatment unless it traces to a documented, disclosed, lawful reason. I’d also check the disclosure: was the existing customer’s price rise shown clearly, and could they see an equivalent cheaper option existed? The arithmetic being right does not make the outcome fair, and fairness of outcome is exactly what the FMA and CoFI expect us to design and test for.
“How do you keep conduct and prudential concerns straight, and where does ACC come in?”
I hold two questions apart: conduct asks ‘is the customer treated fairly?’ and is supervised by the FMA with CoFI and the Fair Insurance Code; prudential asks ‘is the insurer financially sound to pay claims?’ and sits with the RBNZ under IPSA. Most of what I turn into customer-facing tests is conduct. ACC matters at the product boundary — the statutory accident scheme covers personal injury by accident, so for health or life products I test that the wording does not imply cover the scheme already provides or leave a customer thinking they are covered where they are not. Where I am unsure of an exact rule, I test the principle and confirm the specifics against the current source rather than inventing detail.