Assessing iGaming product managers by the bonus they would stop

Picture two finalists for an iGaming product role. Both can define week-four retention, both know which markets license which verticals, and both talk comfortably about safer gambling. Only one of them would stop a high-margin promotion from reaching self-excluded players. A loop built on conversation alone rarely tells you which one, because it rewards fluency. A bad hire here costs more than a slow quarter: in Great Britain a single mis-targeted bonus campaign can end in a Gambling Commission regulatory settlement.

Test judgment with one decision packet

A good assessment shows how a candidate weighs revenue against player value, risk and market rules that differ from country to country. Generic interviews miss this. Candidates can recite retention formulas, roadmap rituals and regulatory terms without ever showing how any of it changes a decision.

So work from one realistic case. Define a market, a product surface, a commercial goal, a slice of data and the constraints that bind. Ask for a short written answer, then a structured discussion around it. You are probing priorities, rejected options, the evidence they would demand, and the point at which they pull in compliance or safer-gambling specialists.

Keep preparation to 75-90 minutes. Ask someone to design an entire sportsbook or casino app and you reward presentation speed and interview practice; a tight packet tests the work the role involves.

Give them a one-page objective, say, improving first-week return among newly verified casino players without raising bonus cost per retained player. Attach compact data on acquisition source, verification, first deposit, first game session, net gaming revenue (NGR), bonus use and return activity. Spell out the markets and the non-negotiable controls: self-exclusion handling (GAMSTOP in Great Britain, Spelpaus in Sweden), age and identity gates, promotional eligibility, auditability.

Then ask for four things. A problem statement that separates the business symptom from the player problem. Ranked product bets, including one they would deliberately not pursue. A measurement plan with a primary metric, its input metrics and guardrails. A risk log covering compliance, fairness, data quality and operational dependencies.

The assumptions are where judgment hides, so force them into view. A promotion proposal has to name the segment, expected player value, cost exposure, eligibility, measurement window and guardrails. A catalogue change has to say whether it improves discovery, margin, trust or retention, and what it might damage on the way. That produces comparable evidence across candidates without mistaking fluency in one operator's tooling for product ability, which is the trap most panels fall into.

Catalogue choices expose product judgment

A catalogue is a discovery system, and stuffing it with inventory works against that job. Every change lands on commercial performance, player experience, supplier economics, and regulatory exposure at once. Strong candidates start by asking what is available and promotable by market, how provider economics differ, which titles fit player intent, and whether a merchandising move creates repetitive or unsuitable exposure.

Different surfaces do different jobs. Search serves players who already know what they want; curated rails carry discovery; recently played cuts the effort of a returning player; promotional placement can lift conversion while crowding out the titles players trust. Candidates reasoning from the mechanics of an iGaming storefront keep availability, ranking, search relevance and campaign placement as separate levers.

Give them a scenario with real tension: a high-margin provider wants a permanent top rail, while new players abandon the lobby after broad, undifferentiated browsing. A strong answer defines an audience and a hypothesis (a market-specific starter rail for verified new depositors, for instance) without amputating search and category access. It also names a warning metric to watch in case the change does damage: complaint rate, failed searches, players seeing the same game over and over, time from deposit to first play, or the signs of harm that the operator's policy tracks.

Knowing game titles counts for little, since they rotate every quarter. What carries over from one operator to the next is a grasp of catalogue rules, fallback states, market controls and measurement.

Cohort economics tests commercial thinking

A revenue total can hide whether a change built lasting value or just bought a short burst of activity. Ask candidates to work in cohorts: groups of players who all started with the same meaningful event, instead of everyone who happened to join in a calendar month. For casino that event might be verified players who complete a first deposit and an eligible game session. The exact definition depends on product and market, and a candidate who says so is already ahead.

Push for definitions precise enough to compute:

• Activation rate = users reaching first value / eligible new users
• Week-4 retention = cohort users returning in week 4 / users in the starting cohort
• Bonus cost per retained player = bonus cost for the cohort / retained players
• Net contribution per acquired account = (cohort gross gaming revenue minus cohort channel, payment, bonus and variable service costs) / acquired accounts in that cohort, over a stated measurement window

Then hand them a fictional cohort: 1,000 verified arrivals, of whom 420 make a first deposit, 230 complete a meaningful game session and 92 return in week four. Watch what they do with it. The deposit rate is 42%, but only 55% of depositors (230 of 420) reach a meaningful session, so the biggest leak sits after the deposit. Week-four retention is 9.2% of the cohort, yet 40% of players who had a meaningful session came back (92 of 230). That points the work at the first session.

Now add money. Say the cohort received £18,400 in bonuses: that is £200 per retained player. A competing channel brings 1,000 arrivals at higher acquisition cost but only £9,000 of bonus spend and 75 week-four returners, £120 per retained player. The channel that is cheaper to acquire turns out dearer once bonus cost is counted. A candidate who reaches that conclusion unprompted, and then asks whether restricted or self-excluded accounts were removed from both denominators, is doing the job.

Bonus design is no longer only a commercial lever. In Great Britain, the Gambling Commission capped wagering requirements at 10 times the bonus from 19 January 2026 and banned promotions that mix gambling products. In Sweden, a licensee may offer a bonus only at the player's first gambling occasion, a rule in force since re-regulation in 2019, and Spelinspektionen fined operators including Betway for breaching it that year. A candidate who designs a reactivation bonus for a Swedish cohort has failed the case, however good the arithmetic.

The sharpest candidates also separate early signals from final results. The time from verification to the first meaningful session moves within days, so it can steer weekly onboarding work. Cohort retention and contribution move slowly and tell you whether the business actually got healthier. Each still needs eligibility rules, a time window and an owner.

Personalisation needs explicit governance

Personalisation in iGaming shapes player behaviour, so it cannot be treated as one more conversion layer. Test whether a candidate uses context to remove friction without applying pressure around sensitive behaviour. Ask them to pick a surface (catalogue ordering, education, offer visibility, lifecycle messaging or account prompts) and lay out the data it needs, the exclusions it must respect and the review process around it.

Bring in AI personalisation in iGaming as behavioural governance, because model quality never settles the product decision on its own. A candidate has to speak to consent and notice, who is eligible and who must never see an offer, human review, audit trails, models that get worse over time, and how a harmful rule gets reversed.

A useful prompt: a model predicts a group will accept a bonus. What has to be true before that bonus appears? A strong answer opens with eligibility, with whether policy and market rules permit using this behavioural data, with lifecycle suitability and with protection for players showing markers of harm, long before it reaches expected conversion. In Great Britain that last point has a named instrument: the customer interaction requirements of LCCP social responsibility code 3.4.3 oblige operators to identify and act on such markers. The answer defines a holdout group and reads the guardrails next to the conversion number. A weak answer treats one-to-one targeting as self-evidently good; a strong one knows relevance has boundaries, designed with legal, compliance, data and safer-gambling partners.

Compliance changes the product shape

Regulatory constraints belong in discovery, requirements, design, release planning and measurement. Stapled on at the end as a final approval, they breed rework and unsafe flows. Watch whether a candidate folds constraints into the very first problem statement.

Ask them to map a feature across three layers. The player flow: which screens, messages, deposit and withdrawal steps or game-access points change. The control logic: which checks, exclusions, limits, records or approvals apply in each target market. The operational response: who receives the alerts, who can pause the feature and who documents the decision.

Product managers are not there to write legal opinions. Their job is to spot and document uncertainty, then get a decision from the accountable legal or compliance owner. Score them on asking for market-by-market rules, on keeping global intent separate from local configuration, and on choosing safe defaults when a requirement is unclear. The bonus rules above are the test case: one global "welcome journey" spec, configured per market, beats three hand-built flows that drift apart. The revealing mistake is treating every control as friction to be filed off; good judgment tells avoidable effort apart from necessary protection and can explain the difference to a commercial stakeholder.

One rubric, evidence first

Run one rubric, and make interviewers record evidence before anyone discusses the hire. That discipline blunts the pull of charisma, shared career history and familiar vocabulary.

A five-point scale only works if each score describes what the candidate actually did. A one lists features with no evidence behind them. A three links each action to evidence and names its trade-offs. A five also sees the knock-on effects, spots the missing data and lays out a practical way to test the idea or drop it safely.

The weights also give you a decision rule. A candidate who scores 4 on cohort economics and 2 on compliance lands within a tenth of a point of one with the reverse profile, but they are not the same hire. Set a floor: anyone below 3 in either compliance or personalisation governance does not proceed, whatever the weighted total.

Calibrate before the interviews start. Have two assessors score a sample response independently, compare the evidence each cited and tighten any rubric language that stayed vague. If they land more than one point apart on the same area, the rubric is measuring taste.

False positives, and the case against the case

Vocabulary can counterfeit depth. There is the metric recital that never reaches a decision: retention, lifetime value and NGR all named, with no cohort, denominator or action attached. There is revenue-first merchandising that pushes promoted content while ignoring player intent, market availability and player-protection effects. There is personalisation mistaken for targeting volume, where more offers count as success with no suppression or control group. There is the compliance handoff ("legal will handle it") with no risk register, release gate or accountable owner. And there is the experiment for show: an A/B test that never says who is in it, how long it runs, what the main result is or when to stop.

Guard against the opposite error too. Do not reject a candidate for lacking local jargon or for arriving from an adjacent regulated product such as payments or insurance. Disciplined questions, instinctive user protection and structured handling of uncertainty often predict a faster ramp than fluent language wrapped around weak judgment.

The design has a fair weakness: a 90-minute take-home favours candidates with free evenings, and a senior PM with eight years at regulated operators has a track record that says more than any fictional cohort. For senior hires with verifiable public work, replace the take-home with a live 45-minute walk-through of a real decision they made, probed with the same rubric. For everyone else, keep the packet, cap it at 90 minutes and say so in the invitation.

Turn assessment evidence into a hire

Spend the live interview probing the written choices. Ask what evidence would change the top priority, what they would ship behind a feature flag, which stakeholder would push back hardest and which control they would refuse to weaken. Answers to those reveal whether the reasoning is genuine or memorised.

Close with a decision record: category scores, the evidence cited for each, the concern still unresolved, the role-specific risk and the conditions under which this hire succeeds. For senior roles, add a review by a compliance or safer-gambling lead, whose job in the loop is to test whether the candidate treats player protection as a design responsibility or as someone else's checkpoint. If that reviewer and the hiring manager disagree, the compliance floor decides.