Why Game Studios Miss Their Biggest Problems | Jakob Berglund

Summary

Jakob Berglund argues that studios rarely lack data or possible projects. The harder problem is turning information into comparable, actionable choices. Teams often prioritize visible feature opportunities over less visible costs—such as technical debt—because they do not consistently estimate the value of both.

Decision-Making Framework

Berglund describes four elements for making better decisions:

• Visibility: Map the drivers of change to the outcomes they affect. A dashboard showing that revenue fell is not enough; teams need to know which underlying factors might explain the change.
• Context: Record what the business has learned about those relationships, so the knowledge is not held only in individuals’ heads.
• Comparability: Estimate different kinds of work using consistent definitions, formulas, and time windows. Roughly comparable estimates are more useful than precise-looking numbers produced in different ways.
• Actionability: Connect estimates to decisions, assign owners, and check afterward how accurate the estimates were.

He also groups business activity into three pillars:

• Customer opportunity: Customers or players who could generate value.
• Business capability: The product, processes, and systems that convert that opportunity into value.
• Business capacity: The people and resources available to perform or accelerate that work.

For prioritization, consider the current constraint while also preparing for the next one. For example, increasing customer acquisition can create a later delivery or service-capacity bottleneck.

Practical Playbook

• Map cause-and-effect links. When a metric changes, investigate its possible drivers and record validated relationships. Build a “tree” of factors that can be checked when similar changes occur.
• Test hypotheses. Correlation alone does not establish causation. Change a suspected driver and see whether the intended outcome follows, using an A/B test where practical.
• Use consistent estimates. Define KPIs clearly and make sure teams use the same formulas. Otherwise, estimates from different projects cannot be fairly compared.
• Estimate size, not false precision. Use a reasonable range or rough scale—such as approximate money, time, or percentage impact—rather than presenting an uncertain forecast as an exact figure.
• Include costs and risk reduction. Compare revenue-generating work with avoided costs, reduced losses, time savings, and improvements to capability or capacity.
• Review decisions after the fact. Track how estimates compare with outcomes. Over time, this helps identify reliable sources and improve confidence in future forecasts.
• Make retrospectives more useful. For recurring issues, identify the circumstances that caused them, estimate their impact, and assign actions to address those circumstances. Start with a simple, consistent method rather than trying to precisely apportion every impact.
• Connect roles to value chains. Show how individual work contributes to a larger outcome—for example, ad creative quality → click-through rate → return on ad spend → revenue. Leadership should also recognize work that prevents losses or improves speed, not just work that directly generates revenue.

Berglund’s leadership lens is to reduce negative outcomes, slow negative trends, add positive outcomes, and accelerate positive trends. This helps prevent teams from focusing only on growth while overlooking costs, reliability, or customer retention.

Examples and Metrics

• At King, Berglund said players were leaving because some devices could no longer support the game. The impact was spread across engagement, installs, spending, conversion, and player cohorts, so standard KPIs showed only a gradual decline. Once the underlying issue was identified, the company redirected its priorities for three months to address it.
• The host’s meeting-booking test compared a more casual outreach message with the previous approach: the reply rate rose from 40% to 65%. This was presented as an example of testing a hypothesis rather than relying on an assumption.
• Berglund illustrated the risk of overlooking technical debt with a hypothetical comparison: a recurring issue could cost $1 million over six months, while a game feature might be expected to generate $1 million over six months. The point is to make recurring costs visible alongside potential revenue.
• In another example, an incident costing $300,000 involved three contributing circumstances. He suggested initially allocating $100,000 to each, rather than claiming unsupported precision about which mattered most. Repeated incidents can then reveal which circumstances deserve priority.
• A hypothetical 0.1% retention uplift was used to illustrate how retention might be translated into lifetime value and estimated revenue; it was not presented as a measured result.
• Berglund said long-term investments such as articles or brand-building may need to be evaluated over a longer window—potentially a year—rather than judged only by immediate leads.
• The host mentioned that King’s Candy Crush Saga was played by roughly 300 million people monthly at the time Berglund joined. This was background context, not a decision-making KPI discussed as a target.

Product and AI

Berglund is building Atlas One, a decision engine intended to help people evaluate products based on their circumstances and needs, rather than simply surfacing the most popular or highest-rated choices.

His broader warning is that AI can make information and analysis cheaper, but more dashboards or recommendations will not improve decisions without business context and clear links between value drivers and outcomes.

Presenters and Sources

• Jakob Berglund — Founder of Atlas One; formerly worked at King and on the MONOPOLY GO! team at Scopely.
• Harry Phokou — Host/interviewer.
• Examples and frameworks are drawn from the interview and the host’s business experiences.