If you read nothing else
A total of 26,747 players responded to a community survey I designed, deployed, and analyzed organically within one week at a studio with no prior research infrastructure. This case study demonstrates the process and outcomes of large-scale mixed-methods research when a single researcher handles instrument development, analytical structuring, and synthesis across four content areas for four internal teams. The research directly informed Vermintide 2’s post-launch content roadmap, and the resulting infrastructure became the template for all subsequent research efforts.
Overview
Fatshark had a vocal, highly engaged player base and no systematic way to hear them.
By 2021, Warhammer: Vermintide 2 had been live for several years and had cultivated a dedicated, highly invested community focused on ongoing content such as new careers, game modes, events, and enemy modifiers. Players regularly expressed their opinions on platforms including Steam, Reddit, and the Fatshark Forums. However, the studio lacked a structured process for translating this feedback into actionable insights for the development team.
I designed and conducted Fatshark's inaugural community retrospective survey, employing a mixed-methods approach to evaluate all 2021 content releases. The study integrated a quantitative Google Forms survey—distributed via the game launcher, Reddit, Steam Forums, and Fatshark Forums—with qualitative analysis of forum and Reddit comments generated in response to the same promotional posts. Within one week, the survey reached 26,747 respondents.
The outcome was a structured, evidence-based account detailing what the Vermintide 2 community valued, their primary frustrations, and their preferences for future content. We delivered it as a comprehensive report with quantitative breakdowns and qualitative themes across four content areas.
The challenge
While player sentiment was abundant across community channels, actionable evidence was notably absent.
The Vermintide 2 community produced a substantial volume of public feedback; however, forum posts and Reddit threads do not constitute systematic research. Such channels tend to overrepresent strong opinions, recent experiences, and vocal minorities. Without a structured research instrument, it was impossible to distinguish majority sentiment from the perspectives of the most vocal participants.
The primary challenge was designing a study that could operate at the scale of a live game community, encompassing tens of thousands of potential respondents, variable engagement levels, and no participant recruitment budget. The study needed to yield findings rigorous enough to be considered evidence rather than anecdote. This required crafting questions that elicited both player preferences and underlying reasoning, and developing a qualitative component to triangulate insights that quantitative data alone could not provide.
Another practical constraint was the study's scope. In 2021, the studio released four major content updates—Chaos Wastes, Grudge Marks, Sister of the Thorn, and Warrior Priest of Sigmar—as well as the Geheimnisnacht event. Evaluating all updates with equal depth would have made the survey too long for most players to complete, while focusing selectively would have deprived some teams of necessary data. Therefore, the research instrument needed to balance precision with brevity to ensure both utility and completion.
The objective extended beyond merely quantifying opinions; it was to establish the studio’s first actionable evidence base to inform content decisions.
Methodology
A mixed-methods approach supported large-scale data collection.
- Total sample: 26,747 respondents; survey open for one week
- Distributed via PC game launcher, Reddit, Steam Forums, and Fatshark Forums — respondents could skip or edit questions
- Primarily PC players (97.8%); console players included as a cross-section; China excluded due to regional platform restrictions
- Core audience: 80.9% with 100+ hours of playtime; 46.2% played at Legend difficulty
- 42.5% of respondents had last played one month ago or more, capturing a meaningful lapsed-player segment
- Comments drawn from Steam Forums, Reddit, and the Fatshark Forums — scoped to responses generated by the survey promotional posts only
- Scoping to promo post comments created partial respondent overlap, allowing qualitative data to triangulate quantitative findings rather than run as a parallel study
- All commenters assumed to be core players with game familiarity; qualitative findings required convergence with quantitative patterns before being treated as conclusions
Process
The process began with instrument design, followed by methodological triangulation.
I wrote the survey instrument from scratch, structuring it in three parts: a player demographics and engagement baseline, a content sentiment overview, and content-specific like/needs-improvement pairs for each of the four major 2021 releases. The like/needs-improvement format gave each content area a parallel structure that made cross-content comparison tractable while keeping individual question sets short.
I excluded China from the survey scope due to region-based platform restrictions, directing that segment to a separate Tencent report. The survey was open for one week — long enough to reach lapsed players, who turned out to be a meaningful segment of the respondent pool.
The qualitative component was intentionally scoped. Rather than analyzing the entire volume of community posts, which would have introduced considerable sampling bias, I restricted qualitative data to comments generated in direct response to the promotional posts announcing the survey. This approach created partial overlap between survey respondents and commenters, allowing forum sentiment to triangulate the quantitative findings rather than serve as an independent data stream.
- Established baseline demographics: playtime, platform, preferred difficulty, genre overlap
- Measured content preference at scale — most and least favorite releases, specific feature ratings
- Surfaced distribution of opinion, not just direction — revealing internal splits within content areas
- 26,747 respondents; questions skippable to reduce abandonment
- Explained the why behind quantitative patterns — what players meant when they rated something poorly
- Surfaced nuance the survey format couldn't capture (e.g. distinguishing Grudge Marks concept approval from specific mark complaints)
- Scoped to promo post comments only — same distribution event as the survey, partial respondent overlap
- Treated as triangulation, not a parallel study — conclusions required convergence across both methods
Key findings
Four content areas. Four distinct player relationships with each.
The study engaged a highly invested sample: 80.9% of respondents reported over 100 hours of playtime, and 62% had played within the past week. This was not a casual audience. Their feedback showed deep familiarity with the game's systems, making the nuanced findings more meaningful than they would have been with a broader, less engaged sample.
A total of 64.1% of respondents identified Chaos Wastes as the best release of 2021, representing the largest share among all content updates. Players appreciated the roguelike run structure and build customization, viewing these as significant innovations compared to the traditional adventure map loop. The primary criticism concerned map variety: nearly one-third of respondents said the maps needed improvement, citing novelty fatigue as a potential long-term risk.
Quantitatively, players valued the challenge and variety Grudge Marks offered—yet Grudge Marks was also the most-cited least-favorite release. Qualitative data resolved the contradiction: players liked the concept but found specific marks (particularly Illusionist, cited by 36% as their least favorite) to feel punishing rather than skillful. Illusionist and Rampart were almost always mentioned together in forum comments, suggesting a combinatorial problem as much as an individual one.
Players welcomed the new career and appreciated the new weapons. However, 23.3% felt that Sister of the Thorn did not fulfill her described class role—a sentiment strongly reflected in qualitative data, where players consistently characterized her as excessively powerful for a support class. The primary complaint centered on the discrepancy between her marketed identity and in-game behavior, indicating a communication and balance issue rather than an isolated design flaw.
36.1% of players said nothing needed improvement — the highest "nothing" response of any content area. Players felt he fulfilled his support role as described, in explicit contrast to Sister of the Thorn. The main complaints centered on weapon similarity and a finicky passive ability. These were specific, addressable criticisms rather than systemic concerns.
Players loved the event's atmosphere and the optional difficulty modifier — 39.1% cited look and feel as their favorite aspect, 24.1% appreciated the skull difficulty modifier. But 37.1% identified the limited-time structure as the main thing needing improvement. Forum comments surfaced FOMO around rewards and desire for the difficulty modifier to become a permanent fixture. The event itself was well-received; the delivery format created friction.
Outcomes
The dataset comprised 26,747 responses, providing a coherent account of what the live game community values.
The retrospective gave Fatshark something it hadn't had: a structured, traceable account of player sentiment that was too large to dismiss as anecdote and rigorous enough to inform planning. Before this study, development decisions were informed by community telemetry, internal playtesting, and the loudest voices on community platforms — all of which skew in predictable ways. A survey at this scale, with this level of demographic grounding, offered a different kind of signal.
The findings were organized and delivered as a formal report — the same document that became the artifact submitted to my mentor and now serves as a reference for this case study. It established a repeatable format for community research at Fatshark: a structured instrument, a consistent question architecture across content areas, and an integrated qualitative layer.
- 26,747 survey responses collected over one week across game launcher, Reddit, Steam Forums, and Fatshark Forums
- Demographics baseline established for Vermintide 2's core audience: 80.9% with 100+ hours, 97.8% PC, skewing toward Legend and Champion difficulty
- Four content areas analyzed with parallel like/needs-improvement structure, enabling direct cross-content comparison
- Qualitative triangulation from forum comments scoped to promotional post responses, connected to the same distribution event as the survey
- First community survey instrument designed and executed at Fatshark, producing a replicable format for future retrospectives
Reflections
The advantages and limitations of large-scale research.
A 26,000-respondent sample is unusually large for game UX research and offers significant advantages: the distribution of opinion becomes visible in ways that smaller studies cannot reveal, and findings are less likely to be dismissed as unrepresentative. However, scale also has limitations. This audience self-selected through community channels, resulting in overrepresentation of players already engaged with studio communications. The lapsed player perspective, with 42.5% of respondents having last played a month ago or more, is present in the data, but the most disengaged players are structurally unlikely to have seen the survey.
Limiting forum comments to responses on promotional posts was appropriate for validity, but it also constrained the richness of the qualitative layer. A study incorporating dedicated open-text fields or follow-up interviews could have explored the underlying reasons behind some of the more surprising quantitative splits. This represents the natural next step that this instrument was designed to support.
The broader lesson from this project concerns the unique value of community research compared to telemetry. While player behavior data reveals what occurred, survey data indicates whether players noticed, cared, and whether their experiences aligned with their expectations. For a live game with a highly engaged player base, this reasoning layer often contains the most actionable information.