BCRAFT CPO Denis Zakalinsky: Speed Is Not a Strategy

Two operators can run the same platform, the same games and the same payment providers, and still end up in completely different places. Denis Zakalinsky, CPO at BCRAFT, thinks that gap has little to do with technology. 

In this conversation he talks about what “fast to market” really costs, why saying no is most of the job, when localisation stops being translation, and how he tells a real AI use case from a demo. 

Clast: At what point does a product being “fast to market” become a liability?

Denis: Speed becomes a liability when we start creating operational debt faster than we are creating market knowledge.

A fast launch exists to validate a market, a proposition and a business model. It is not a way to release an incomplete product and pass the consequences on to players, support teams and future development.

I am willing to trade scope for speed. I am never willing to trade control for speed.

The non-negotiables are regulatory compliance, the integrity of balances and transactions, security, payment and KYC controls, responsible gaming, auditability and measurement. 

If we cannot see what happens after launch, we are not learning, we are simply moving quickly in the dark.

Everything else is negotiable: the number of providers, mechanics, campaigns and interface variants. A two-week launch followed by six months of incidents, manual work and technical debt is not speed. It is a delayed failure.

Clast: How closely do you work with clients nowadays?

Denis: With more than 200 operator projects, it would be neither possible nor healthy for the CPO to take part in every client conversation. But distance from individual tickets must not become distance from the market.

I stay personally involved at high-signal moments: strategic discovery with major operators, entry into a new GEO, the same pain appearing across several clients, significant churn risk, and requests that could change the platform’s architecture or economics.

The team should be able to solve an individual client problem without escalating it to me. My responsibility is to recognise when an individual problem is actually evidence of a systemic platform gap. Client conversations are never the only input, because what clients request, what their players do and what creates economic value are often three different things — and the loudest client should never become the strategy.

Clast: How can one product be noticeably better than another when both brands use comparable technology?

Denis: 

Technology provides capability. Differentiation comes from the quality of the decisions made on top of it.

Two operators can run on the same platform, with the same games and payment providers, and produce completely different results. The stronger one knows exactly whom it serves, how it earns trust, which content is relevant to that audience, how the cashier behaves and what a player sees after a failed deposit. It connects acquisition, product, CRM, payments, support and risk into one coherent system.

The most common mistake is treating the platform as the finished product: rebranding it, loading thousands of games and expecting the technology to create the business.

The best products are not the ones with the most functionality. They remove uncertainty from the moments where trust is won or lost: registration, the first deposit, bonus eligibility, KYC, the first withdrawal and the first contact with support.

Industry benchmarks put the share of casino first-time depositors who return for a second deposit anywhere between roughly 25% and 50%. A spread that wide cannot be explained by technology alone. Most of it is the result of decisions.

Clast: How often does your job involve saying “no”?

Denis: Quite often but a good CPO rarely just says no. The job is to understand the problem behind the requested solution.

Every feature creates a permanent tax. It has to be designed, developed, tested, documented, monitored, supported and maintained, and on a B2B platform that tax is multiplied across hundreds of projects. A platform cannot become a collection of everything every client has ever requested.

I test requests with a few questions:

• Is the problem supported by evidence?
• Does it fit the product strategy? 
• How many clients or markets could benefit?
• Which behaviour and which financial metric should change? 
• Could configuration solve it without touching the core? 
• And what is the total cost of ownership, not just the initial estimate?

The clearest warning sign is a feature requested by one client, for one campaign, with no measurable outcome, that requires a permanent exception in the platform. Often the right answer is a configuration option, an experiment or a smaller version of the idea.

Clast: Is the progression from ready-made technology to custom technology still relevant?

Denis: It does, but I would challenge the assumption that scale means an operator should eventually own everything.

A ready-made platform is usually the rational way to validate a market without spending years rebuilding commodity infrastructure. As the business grows, the operator should take ownership of the layers that create genuine differentiation — not automatically rebuild the wallet, the ledger, the integrations and every back-office component.

An operator has outgrown its platform when the platform consistently limits-proven economics or a strategically important capability: when it blocks meaningful experimentation, restricts access to data, cannot support necessary market mechanics, creates unacceptable integration delays or makes a differentiated experience impossible. “Now we are a large company” is not a sufficient reason to rebuild.

In most cases the right destination is a composable model. Wallet, PAM, payments and game aggregation stay on proven infrastructure. The operator owns the data layer, CRM and bonus logic, personalisation and the front-end experience — the layers where decisions are made and where competitors cannot simply copy the result.

Custom technology should be an investment in a moat. Otherwise, it is often just a very expensive expression of independence.

Clast: At what point does localisation constitute a redesign?

Denis: Localisation becomes redesign when the same product journey produces different behaviour because the market has a different model of trust.

Language, currency and payment methods are necessary, but they are only the visible layer. Underneath are how players discover brands, what makes an operator feel credible, how comfortable they are with KYC, how they expect deposits and withdrawals to work, which content feels relevant, and whether agents, messengers or offline cash are part of the experience.

At that point we are no longer changing content. We may need to redesign registration, cashier logic, lobby structure, communication, support flows or even the operating model behind the product.

This is why I describe localisation as the economics of player behaviour. A cultural difference matters to a product team when it changes conversion, retention, payment performance, risk or lifetime value.

Clast: Has your view on retention in the regulated German market changed?

Denis: The fundamental view has not changed. Five years of the new regime have made it stronger.

Germany is the clearest European test of what happens when the usual retention levers are removed. GlüStV 2021 introduced a €1 stake limit and five-second spins on slots, a €1,000 cross-operator monthly deposit cap enforced through LUGAS, bans on autoplay and progressive jackpots, and a 5.3% tax on stakes rather than revenue, which pushed RTP down. Aggressive bonuses and CRM pressure could no longer hide friction, weak relevance or a lack of trust.

The tiered stake limit introduced in July 2026 up to €5 for players checked against LUGAS and OASIS — does not change that logic. It raises the ceiling for a verified segment, it does not give operators their shortcuts back. And licensed operators still compete on lower RTP against an offshore market that offers more of everything; industry estimates put channelisation for slots at only 20-40%. A licensed brand cannot win that contest on RTP, so it has to win on safety and predictability.

That means reliable payments, transparent balances and bonus rules,verification that feels proportionate, relevant content within the permitted portfolio and support that resolves problems quickly.

I also measure retention quality, not only return frequency. Are players returning without increasingly expensive incentives? Are second depositsprofitable? Is NGR growing together with retention, or are we postponing churn through bonus cost?

Responsible gaming should not be a compliance layer added after the product has been designed. It should shape segmentation, communication pressureand the mechanics we consider acceptable. Retention does not mean keeping someone gambling at any cost. It means giving healthy customers a credible reason to choose your product again.

Clast: How can you tell the difference between a genuine AI use case and one that has been added just for show?

Denis: I ask four questions:

• Which specific decision or workflow are we improving?
• What proprietary data or context makes AI better than a rule-based system?
• Which measurable baseline should change? 
• And what happens when the model is wrong?

If a team cannot answer those questions, it has an AI presentation, not an AI product. I have rejected use cases on that basis: a generic chatbot without a clear player problem, or an “AI recommendation engine” without enough behavioural data, a feedback loop or a way to measure incremental value. And I would not give a probabilistic model uncontrolled access to high-risk financial actions.

The strongest cases in iGaming pass all four questions: early detection of harmful play, fraud and bonus-abuse detection, and payment anomaly monitoring. They rest on proprietary behavioural data, have a clear baseline and come with an obvious failure policy, a person reviews before anything irreversible happens.

We use AI internally to streamline processes such as market intelligence, discovery, analytics and decision preparation. In several of these workflows, this has produced a measurable improvement of 55-60% in cycle time/analyst hours.

I also build AI products myself — Docaudit for contract review and CV Audit for resume review — so I see the gap between a demo and a product from the inside: evaluation, edge cases, cost per query and what users are actually willing to trust. AI should earn its place in the product with a KPI and a failure policy.

Clast: What would you consider first to determine whether an iGaming product is genuinely beneficial?

Denis: First, I would verify whether the data and metric definitions can be trusted. A dashboard built on inconsistent definitions of FTD, GGR, NGR or attribution is worse than no dashboard, because it creates false confidence.

Then I would go directly to cohort economics. I would segment players by GEO, acquisition source, campaign and registration period, and examine deposit quality, repeat deposits, active days, retention, GGR-to-NGR leakage, bonus cost, fraud, payment acceptance, withdrawal performance, LTV, CAC, contribution margin and payback period.

The key question is whether the next euro of acquisition creates future gross profit or simply buys another unprofitable FTD.

In one business I reviewed, marketing spend more than doubled, from €39,000 to €81,000. FTDs grew by 87%, so the acquisition dashboard looked successful. But GGR grew by only 20%: revenue per new depositor fell by roughly a third, before counting bonus cost, which scales with every new depositor. The operating result became significantly worse. The growth was being bought, not earned.

After the numbers, I would inspect the product and back office: failed deposits,KYC drop-off, withdrawal delays, support contact reasons, manual interventions and bonus abuse. That is usually where the cohort numbers find their explanation.