How do you raise a defect's perceived impact honestly instead of inflating it?
answer
- Evidence raises impact, adjectives do not
- Swap the contrived path for a realistic one
- Count who is actually exposed
- Keep asking: and then what?
- State the limitation that weakens your case
basics
~20 sRaise impact with evidence, never adjectives: replace a contrived path with the most realistic one that triggers the same defect, count who is exposed, follow the consequence to money, data or safety, and state the conditions that limit it.
solid answer
~50 sThe honest levers are all factual. First, generalise the path: if I found it through an odd sequence but it also fires on a route real users take daily, report the realistic one. Second, quantify exposure - how many accounts, records or trips actually meet the conditions. Third, follow the consequence to something the reader already cares about: lost money, corrupted data, a blocked support queue, a regulatory obligation. Fourth, say what the system would notice on its own, since an undetectable failure costs more. Inflation is the opposite move: claiming 'all users' without a count, hiding the unusual precondition, loading the wording, or nudging a rating to force attention. It works once. After that every report from you is discounted, a far bigger loss than one unfixed defect; and when a defect really is minor, saying so is what makes the rest believable.
go deeper
Recall that impact is argued with facts: who is affected, how often, and what happens as a result. Avoid words like catastrophic or unusable unless you can show the effect they describe, and always include the condition the defect needs.
Explain the honest levers - generalising a contrived path to a realistic one, counting the exposed population, following the consequence to money or data, stating detectability - and be able to contrast each with the inflation move it is often confused with.
Show that you treat your own credibility as the asset being spent, including reporting negatives and limitations voluntarily. Interviewers want to hear you separate the evidence you gathered from the reading you put on it, so the reader can disagree with the second without doubting the first.
Own the incentives. If your reporters have to inflate to be heard, the problem is the intake, not the reporters - so decide what evidence you require, how impact claims are checked, and how you keep an accurate reporter from losing to a louder one.
### Advocacy is an evidence problem, not a persuasion trick A defect gets fixed when a busy reader with a full queue concludes that fixing it is worth more than the alternatives in front of them. Nothing about that is dishonest; a report that buries a serious problem in flat, contrived prose fails the product just as much as one that exaggerates. The skill is making a true thing legible - and the reason it is a skill at all is that the tempting shortcuts all work in the short run and cost you your credibility in the long one. ### The four honest levers **1. Generalise the path.** Testers find defects by unusual routes because unusual routes are what they are paid to walk. If the same defect also fires on an ordinary one, report the ordinary one and mention the contrived route as an aside. The strongest version of this move is a realistic scenario the reader recognises: not 'if a supervisor edits a trip that was already settled', but 'when a rider's promotion is applied after the trip starts, which is the normal flow for the reactivation campaign'. Nothing is exaggerated - you replaced an unrepresentative example with a representative one. **2. Quantify exposure.** Replace 'many users' with a number you actually obtained: how many records match the triggering conditions, how many accounts hold that role, how often the flow ran last week. If you cannot get a count, say so and give the bound you can defend - 'at least the three regions where the campaign is live'. A defensible small number beats an indefensible large one, because the first survives being checked. **3. Follow the consequence outward.** Whatever the defect does technically, keep asking 'and then what?' until you land on something the reader already values: money moved wrongly, data that cannot be reconstructed, a support queue that fills, an obligation the organisation has to a regulator or a partner, a user who cannot complete the one task the product exists for. Stop at the first genuine one; chaining speculative consequences is inflation wearing a suit. **4. State detectability and reversibility.** Whether the system itself would catch this, and whether the damage can be repaired later, are checkable facts and often the two that matter most. A wrong value that a nightly comparison catches is a different defect from the same wrong value that nothing ever compares. ### The four inflation moves Knowing the honest levers is half of it; recognising the dishonest ones in your own drafts is the other half: - **Fabricated frequency.** 'Happens constantly' after four observations, or 'all users' from one account. - **The hidden precondition.** Omitting that it needs a role three people hold, or a configuration nobody runs. The precondition belongs in the report even when it weakens your case. - **Loaded wording.** 'Catastrophic', 'unusable', 'data loss' where the data is recoverable. Emotional vocabulary is the reader's cue that the facts were not enough. - **Rating games.** Nudging a field upward to buy attention. It is the most common one, and the most corrosive, because it degrades a signal the whole team relies on. ### Why credibility is the real asset A reporter's history travels with every report they file. Once a reader has been burned by one inflated claim, they start their next read from your reports by discounting everything, and the cost lands on the serious defect you file six weeks later. The compounding runs the other way too: a tester known for exact counts, stated preconditions and voluntarily reported limitations gets read carefully and gets defects fixed with far less argument than a louder colleague. This is also why reporting negatives matters. 'I could not reproduce it with an ordinary rider account' and 'this only affects the display, the stored value is correct' cost you nothing on that defect and buy you the benefit of the doubt on the next one. ### A worked example A dispatcher defect is first written up as 'fare shown incorrectly on the trip card'. The honest raise, with no new adjectives: the value is stored, not just shown; it flows into the weekly settlement total; nothing in the product compares it against what the rider was charged, so the difference is silent; 1,873 trips in the last full week met the triggering condition, averaging 0.07 each; and the trigger is the promotion flow the current campaign relies on, not an exotic sequence. Every clause is checkable and the report ends with the honest limitation - it only occurs when the promotion is applied after the trip starts, and riders are billed correctly. That last sentence is what makes the rest believable.
- What do you do when your own probing shows the defect really is minor?Say so explicitly in the report - the observed effect, the narrow conditions, and the fact that I found nothing worse behind it. Then state what would change my mind: which downstream consumer I could not check, or which configuration would make it worse. Volunteering that a defect is small is the cheapest credibility you will ever buy, and it is what makes people take the next report seriously without argument.
- How do you report exposure when you genuinely cannot get a count?Give the tightest bound I can defend and label it as a bound: 'at least the three regions where the campaign is running', or 'every account with the supervisor role, which I could not enumerate'. I also name what would produce the real number and who could run it. A hedged, checkable statement invites someone to sharpen it; an invented number invites someone to disprove it and discount the rest.
- Why is nudging a rating upward more damaging than overstating impact in the prose?Because the rating is a shared signal that other people's planning depends on, and it is consumed without the prose attached. Inflated wording is at least visible to anyone who reads the report; an inflated rating quietly distorts ordering for everyone and cannot be checked without re-doing the investigation. It also tends to trigger a defensive correction, after which your accurate ratings are the ones nobody trusts.
It is the difference between a doctor saying 'this looks bad' and one saying 'the reading is 1.8 above the safe range, in an organ we cannot repair afterwards' - the second is not louder, it is checkable.
saying these in an interview costs you the question
- Escalates with adjectives instead of evidence
- Claims 'all users' from a single account
- Omits the precondition that narrows the defect
- Nudges a rating upward to force attention
- Chains speculative consequences to sound serious
- Never reports that a defect turned out minor