Designing for
Human Minds

The Problem

Deadlock is Valve's competitive MOBA currently in closed beta. After each match, the game offers thirteen different statistical breakdowns: lane stats, damage dealt, gold economy, kill timelines, and more. On paper, that sounds like a rich feedback system.

In practice, the players we researched used an average of two to four of those thirteen tabs. Most of the analytics were simply ignored, not because they were broken, but because they weren't compelling enough to explore.

The problem isn't missing data. It's that the data doesn't answer the questions players actually care about in a way they can act on.

This project explores how Deadlock's analytics experience could be redesigned to support player learning, performance reflection, and strategic decision-making, without adding more complexity to an already information-dense interface.

User Research

Methods

Our team used two research methods to understand how players currently engage with Deadlock's analytics. First, we conducted semi-structured interviews with ranked players, specifically those at Ritualist rank or above, who have enough game time to have formed real opinions about the post-match system. Second, we conducted field observation through live gameplay streams on Twitch, watching how players interact with in-game data in real time.

What I found most valuable about the interview method was that it forced us to listen for behavior, not just preference. We didn't ask "do you like the analytics screen?" We asked players to walk us through exactly what they do after a match ends, step by step, and let the patterns reveal themselves.

Participant 1: Kenny, 23, Phantom Rank

Kenny has over 300 hours in Deadlock and plays at Phantom, the third highest rank in the game. I expected a player at his level to explore more of the post-match data than a newer player. What I found was the opposite.

Kenny told us he consistently returns to exactly two post-match tabs: Lane Stats and Damage Dealt by Type. Lane Stats tells him how the first ten minutes of the match unfolded. Damage by Type tells him how effective his build actually was. Everything else sits unused, not because it's bad, but because it doesn't directly answer "why did this match go the way it did?"

Participant 2: Jeremy, 21, Archon Rank

Jeremy plays at Archon rank with 250+ hours. He used four of the thirteen post-match tabs in our session, slightly more than Kenny, but his attitude was telling. He said outright that he's "not trying to check all that" and would rather move on after a match. He did express genuine interest in analytics that don't currently exist in the game, like tracking teamfight participation or resource denial.

What struck me about Jeremy's session was the combination of disengagement with the current system and interest in something better. He wasn't opposed to analytics; he was under-served by the ones available.

Key Insight

The clearest finding from both interviews: expertise doesn't lead to broader analytics engagement; it leads to more selective engagement. An experienced player doesn't explore more tabs over time. They learn which two or three are worth their attention and ignore the rest. A better design doesn't add more; it curates better.

High-Fidelity Prototypes

After our initial research and low-fi sketches, we built a high-fidelity prototype that shifted the core design direction from a post-game analytics dashboard to a live in-game tactical assistant. The key insight driving that shift: if players only engage with analytics for a second in the middle of a match, the interface needs to meet them there, not ask them to navigate somewhere else.

The prototype spans five screens. Two are shown here. The system is framed as a "Combat Assistant," always on, always contextual, rather than a static report you review after the fact.

User Testing

Methodology

After building our high-fidelity Figma prototypes, we returned to the same two participants, Kenny and Jeremy, and ran informal remote user testing sessions over Zoom screen share. We showed each participant two distinct prototype flows: a post-game analytics flow focused on AI-driven recommendations and training drills, and an in-game analytics flow designed as a real-time tactical assistant.

We started each session by asking for first impressions without any prompting. Then we walked participants through specific tasks: finding a key weakness from the match, generating AI recommendations, understanding why a recommendation was made, and locating a training drill from that recommendation. Finally, we asked them to compare both flows directly.

Because Kenny and Jeremy were also our original interview participants, they gave us the unusual advantage of being able to compare what they said in research to how they actually behaved with a prototype in front of them.

What Kenny Responded To

Kenny understood the post-game prototype immediately and responded well to the AI Insights screen, specifically because it surfaced the Damage by Type chart, which he already trusts, as supporting evidence for the AI recommendations. That design decision paid off. Making the analytics he already valued feel like proof rather than extra noise was the right call.

He responded more strongly to the in-game prototype overall. The Tactical Brief screen, which described the match situation with language like "You dominated early engagements, but tactical errors cost mid-game control," felt more useful to him than a stat breakdown. His main critique was visual: the UI felt too much like a separate dashboard rather than something embedded inside Deadlock's world. He suggested placing the interface over a blurred in-game background to make it feel like an overlay rather than a standalone app.

What Jeremy Responded To

Jeremy's most positive reaction was to the Recommended Training screen in the post-game flow. In our earlier interview, he had specifically asked for analytics that connected to things he could actually do differently. Seeing a drill card with a severity tag, time estimate, and "Start Training" button built from the exact weaknesses found in his match, which addressed that directly.

Jeremy raised the most useful structural critique of the session. On the in-game screens, he kept asking: "When would this actually show up?" He said the full-screen tactical interface would feel overwhelming during active combat but would be genuinely useful during a death screen, at the shop, or in the downtime between objectives. That timing question reshaped how we thought about the in-game flow entirely.

Cross-Participant Findings

Personal Reflection

The finding that surprised me most was how strongly timing shaped the perceived value of the in-game interface. The same information felt useful or intrusive depending entirely on when in the match it appeared. That's a cognitive load problem, not a content problem, and it pushed us toward designing around game states rather than just designing screens.

I also came away from these sessions more confident in the hybrid direction: light tactical signals during the match, followed by deeper AI analysis and personalized training after. That structure mirrors how real athletes consume feedback: brief real-time cues during performance, fuller debriefs afterward. Building that rhythm into Deadlock's analytics felt like the most honest design response to what our participants told us.

Before & After Stories

Based on user testing feedback, our team made two significant redesigns. Each change was driven directly by what participants told us and what we observed in their behavior.