The measurement problem
A studio needs to know which of its games work, and it faces two structural obstacles.
The first is that it does not hold the data. Player behaviour sits with the operator, and what the studio receives depends on what was negotiated, as covered in the previous lesson. Studios distributing through aggregators frequently see less again.
The second is that placement confounds everything. A game in a featured position performs well, and a game buried on page four does not, largely independent of their merits. Comparing raw performance across titles compares the placement they received.
Both are addressable and neither is fully solvable, which means portfolio decisions in this business are made on partial evidence. The discipline is to use the evidence that discriminates rather than the evidence that is available.
The measures worth having
Revenue generated is the headline and is the most placement-confounded figure available.
Unique players indicates reach and is similarly determined by exposure.
Rounds per player per session indicates whether players engage once they arrive, and is much less affected by placement.
Return rate measures the proportion of players who come back to the game after a first session, and is among the strongest available indicators of quality.
Retention curve extends that across weeks and months, showing whether the game holds an audience.
Session length distribution indicates engagement depth, with the caution that longer is not automatically better.
Longevity, meaning how long the title continues generating meaningful revenue after launch, is the single most informative measure a studio can track. It cannot be manufactured by placement, because sustained placement follows sustained performance rather than preceding it.
Placement-adjusted comparison, where performance is assessed relative to the exposure received, is the analytically correct approach and requires data studios frequently lack. Where placement information is unavailable, the proxy is to compare the game's engagement measures rather than its revenue, since those are less exposure-dependent.
The general rule is that revenue tells you about placement and engagement tells you about the game. A studio reviewing its catalogue on revenue is reviewing its commercial function's effectiveness, which is worth knowing and is a different question.
Reading a game's trajectory
Games follow recognisable patterns after release, and identifying which pattern a title is in supports the decisions that follow.
Strong launch, rapid decline. High initial play from featured placement, falling sharply as the placement rotates and players do not return. This indicates a game that attracted attention and did not hold it, and it is the most common pattern.
Modest launch, sustained performance. Lower initial figures with a shallow decline, sometimes growing as word spreads and as operators notice performance and improve placement. This is the profile of the titles that matter, and it is frequently mistaken for failure in the first weeks.
Strong launch, sustained performance. The successful case, rare and worth understanding when it occurs.
Flat and low throughout. The game did not find an audience regardless of exposure.
Delayed rise. Occasionally a title finds an audience later, usually through a specific operator's promotion or a shift in player taste. Studios that retire titles too quickly on early figures sometimes remove games that would have found their moment.
The practical implication is that assessment at four weeks is unreliable, that the shape of the curve is more informative than any point on it, and that the second pattern above is routinely misread as failure by studios that judge on launch figures.
Deciding what to build next
Portfolio decisions follow from the analysis, and several questions structure them.
What worked, and specifically what about it worked? A successful title is evidence about a mechanic, a volatility profile, a theme, an audience or some combination. Establishing which is the difference between a useful sequel and a superficial one.
Where is the catalogue thin? A studio whose portfolio is concentrated in high-volatility feature-heavy titles is not serving the players who prefer steadier play, and operators building balanced lobbies notice.
What are operators asking for? They see performance across many studios and their requests are informative, with the caution that they will ask for more of what is currently working, which is a description of the past.
What can be built efficiently? Games that reuse framework capability, art direction or mechanics cost less and therefore need to earn less to justify themselves.
What would differentiate? The commodity trap described in the first lesson is real, and a catalogue of competent titles resembling everything else has commodity economics.
Will this cannibalise? A new title taking players from the studio's existing games has added less than its figures suggest, and this is worth checking rather than assuming.
Sequels and variants
Given the constraints on modifying live games, sequels are the primary mechanism for acting on what a title revealed.
A sequel takes a successful game's identity and applies what was learned: adjusted volatility, refined features, improved readability, a new theme within the same world. It carries recognition, which helps placement, and it addresses the weaknesses the original exposed.
A variant is a lighter treatment, taking the same mathematics or mechanic into a different theme. This is efficient and risks diluting the original's identity if overused.
A re-release or refresh updates a title's presentation while retaining the model, which requires recertification and is occasionally worthwhile for a game whose mathematics works and whose presentation has dated.
The risk across all three is the franchise treadmill: a studio producing sequels because they are safer than new concepts, gradually losing the capacity to create anything that could become a franchise. Studios recognisable by their sequels frequently have not released anything genuinely new in years, and the catalogue ages with the audience.
Retirement
The decision studios avoid, for the same organisational reasons operators avoid exiting markets.
Titles carrying negligible play still consume something. Certification maintenance where markets require periodic recertification. Technical maintenance as platforms, browsers and devices change. Catalogue attention, since operators have finite lobby space and a studio's weak titles compete with its strong ones for the space allocated to that studio. And support burden when issues arise.
Meanwhile the revenue is near zero, and the game contributes nothing to the studio's identity.
The reasons retirement does not happen are familiar. It is visible and feels like an admission. Someone advocated for the game. The costs are diffuse and the removal is discrete. And nobody's role is to propose it.
The remedy is a scheduled catalogue review with an explicit question applied to every title: if this were not in the catalogue, would we build it today, and what is it costing us to keep. Titles that fail both should be retired rather than left pending further consideration.
Feeding evidence back into design
The final and most valuable use of performance data is improving the studio's judgement.
That requires connecting outcomes to decisions. Which volatility profiles have performed best for this studio's audience. Which mechanics produce retention rather than only initial trial. Which themes have worked and which were fashionable at concept and dated at release. Which features players actually engage with versus which were built because they demonstrated well internally.
Studios that record their expectations at concept, and compare them to outcomes afterwards, build a genuine institutional understanding of what works for them. Studios that do not rely on the recollection of senior people, which reliably reconstructs the past to match the present and attributes successes to intention.
This is the decision-record discipline from the Operations Strategy course applied to creative work, and it is unusual in studios precisely because creative judgement is regarded as the kind of thing that cannot be examined. It can, and the studios that examine it improve faster than those that do not.
A worked portfolio review
To make the analysis concrete, consider a studio reviewing five titles released over the preceding eighteen months. The figures are illustrative.
Title A launched into featured placement at a major operator, produced strong first-month revenue, and by month four was generating a small fraction of that. Return rate after first session was low. The engagement measures indicate a game that attracted attention through exposure and did not hold it.
Title B launched with modest placement and modest revenue. Its monthly figures have risen slowly for a year. Return rate is high, retention at ninety days is well above the studio's average, and two operators have improved its placement without being asked. This is the profile that matters, and at month four it looked like the studio's weakest release.
Title C produced middling figures throughout with flat engagement measures. It found no audience regardless of exposure.
Title D performed poorly for eight months and then rose sharply when one operator ran a tournament around it, sustaining at the higher level afterwards. It had an audience that had not found it.
Title E launched strongly and has held roughly half its initial level for a year, with good retention. A solid performer.
The decisions that follow differ from those a revenue ranking would produce. Title B is the studio's most important release and warrants a sequel, whatever its cumulative revenue says. Title A's mechanic should be examined for why it failed to retain, and repeating it would be a mistake despite its launch figures. Title C is a retirement candidate. Title D indicates the studio's placement work is underperforming its content, which is a commercial finding rather than a creative one. Title E is the model to understand and repeat.
The general lesson is that the ranking by revenue and the ranking by what the studio should learn are substantially different, and only the second improves the next release.
Building the data capability
A practical note, since much of this lesson assumes information many studios do not receive.
Negotiate for it. As the distribution lesson established, data rights are a contractual term and studios that do not ask do not receive. The argument to operators is straightforward: a studio that can see how its content performs builds better content for that operator.
Normalise across sources. Data arriving from several operators and aggregators in different formats, with different definitions and different periods, requires consolidation before it means anything. This is unglamorous and is the prerequisite for everything else.
Establish definitions. Return rate, retention window, active player and session all need consistent meanings, and the sources will not agree.
Record expectations at concept. What the studio expected this title to do, for whom, and why. Without this, outcomes cannot be compared to intentions and the studio learns nothing systematic.
Review on a schedule. A catalogue review that happens when someone raises a concern is a review that happens too late.
Share findings internally. Performance data that reaches only the commercial function does not improve the games. Designers who see how their previous titles performed, in terms of retention rather than revenue, calibrate differently.
Studios doing this have a genuine advantage, and it is available to any studio willing to negotiate for the data and do the unglamorous work of making it usable.
Release strategy
The final portfolio decision is cadence and timing, which interacts with everything above.
Cadence must be sufficient for portfolio effects, as the first lesson established, and sustainable given the studio's capacity and framework quality. A studio releasing faster than its pipeline supports produces rushed titles that fail at the vertical slice stage nobody held.
Timing against the calendar matters commercially. Operators plan promotional periods around major sporting events, seasonal peaks and their own campaign schedules. A release arriving in a congested window competes with everything else; one arriving in a quiet period may receive placement it would not otherwise get.
Spacing within the catalogue avoids cannibalisation, since two similar titles released close together compete for the same players and the same placement slots.
Sequencing by market follows the certification logic from the production lesson, with markets added as a title demonstrates it justifies the cost.
Balancing the portfolio across volatility profiles, mechanics and audiences, so that the catalogue serves the range of players operators need rather than concentrating in whatever the studio most enjoys building.
The recurring tension is between cadence and quality. Portfolio effects require volume; volume under fixed capacity means less time per title. The resolution is not to choose one but to improve the framework and tooling so that the same capacity produces more without reducing what goes into each game, which is the argument for framework investment made in the production lesson.
What this course has argued
Drawing the studio material together before the final lesson.
A studio is paid from revenue share on games that are played, which makes placement decisive and commercial work as consequential as development. Game performance is severely skewed, which makes portfolio construction unavoidable and hit-picking unreliable. The mathematics is the product and the presentation is the invitation, and both carry design responsibilities. Certification makes post-release iteration expensive, which pushes discipline earlier in the pipeline. Distribution costs margin and reach must be earned. And performance data, where a studio can obtain it, is the mechanism by which judgement improves rather than merely accumulating.
The final lesson addresses the obligations that come with all of this, which fall on studios more directly than is generally acknowledged.
Common analytical errors
A short catalogue of the mistakes studios make when interpreting their own performance data.
Ranking titles by revenue. Measures placement, as established throughout this lesson.
Judging at four weeks. Too early to distinguish a title that will sustain from one that will not.
Attributing success to intention. A hit is evidence that something worked and rarely evidence that the reason it worked was the reason the studio intended.
Ignoring the operator dimension. A title performing well on one operator and poorly on another is telling you about placement and audience fit, not only about the game.
Comparing across markets without adjustment. Player preferences differ substantially by market, and a title underperforming in one may be succeeding in another.
Treating aggregate player numbers as engagement. Reach and depth are different, and a game played once by many people is not the same as one played repeatedly by fewer.
Sampling from success. Analysing the characteristics of the studio's hits without comparing them to its failures produces conclusions that do not discriminate, since the failures frequently share those characteristics.
Confusing the sequel effect. A sequel performing well may be doing so because of the original's recognition rather than because of anything the sequel did.
Each of these leads to a wrong conclusion about what to build next, which is the decision the entire analysis exists to inform.