# Guild of Guardians Case Study | Data-Driven Game UX url: https://bramha.work/work/guild-of-guardians route: /work/guild-of-guardians format: text/plain; charset=utf-8 llms-txt: https://bramha.work/llms.txt llms-full: https://bramha.work/llms-full.txt description: How data analysis and UX feature design raised D7 retention 25% and D7 LTV 12% on a live mobile RPG — relevant for game product manager and data analyst for games roles. --- Immutable Mobile RPG Data Analysis Feature Design Live Product # Repairing stickiness to improve revenue How data-driven feature design increased D7 retention by 25% and D7 LTV by 12% for a live mobile RPG. Role Feature Owner & Lead UX Designer Timeline 6 months Team 1 UI Artist · 1 UI Engineer · 1 Full Stack Engineer ## Executive Summary +25% Retention (D7) +12% Revenue (D7 LTV) 01. The Problem The game suffered from a lower than standard D7 retention. Users reported that they needed to play 3-4 times a day to maximise their progression when most users only had time to play 1-2 times a day, creating an ill fit for user lifestyle needs. 02. The Data Using SQL (BigQuery), I identified that the most engaged users were pulling the average session/day metric up and that the average experience felt worse to most users compared to what the dashboards indicated. Most users were playing once a day and therefore lagging behind in a progression system that needed them to play 3 times a day for the best results. 03. The Solution - Shifting Player Emotions: Designed an "AFK Reward Feature" that accumulated rewards while the player was away, transforming the mental model from "I'm missing out" when not playing to "I have a gift waiting for me." - Designing for Delight: Changing the game loop to reframe features that interrupted player focus to a loop that creates positive anticipation as a hook for long term retention. - End to end delivery: Designed supporting features — composition recommendation, daily login calendar, and free rewards for achievements to maximise D7 retention. - Bonus Win: Increase D7 LTV by 12% by surgically adding monetisation offers to new features. ## Context Guild of Guardians is a squad RPG game where users create teams of 5 heroes to duke it out against monsters, enemies, and other players. In the first 7 to 10 days, the users are expected to play the Adventure mode where they explore the story, learn about the characters, and fight some easy monsters to learn about the game. We had a problem — our D7 retention, the percent of players still continuing to play by the 7th day since they installed, was low. Which means we lost players who we spend lots of money to acquire, reducing the efficiency of our marketing budget. On top of this, we did not have a dedicated data analyst and user researcher to help us figure out what is happening. The challenges I set out to solve And these outcomes were achieved ## Action ### TLDR Problems analysed through a mixture of quantitative and qualitative data analysis: - Poor structure of sessions → it always ended on a negative note - Lack of recommendations when players get stuck - Lack of visual goal setting - Lack of celebration of milestones How I led the team to fix it: - Make UX and game economy changes to make sure player ends game session on a happy note with anticipation for the next time - Add a strategy recommendation system that nudges and guides while maintaining player autonomy - Create visual UI that creates a path and goal setting - Added moments of celebrations and paired them with in-app purchases to score a double win on retention and monetisation ### Before & After — Quick Overview Old user flow to use adventure energy, get rewards, upgrade, and fight: 7-10 clicks and open app 3 times a day to get rewards that you have already earned New user flow — rewards as soon you open the game: 2 clicks and open app 2 times a day to get rewards that you have already earned Before After Visual goal setting: Players clearly see their progression path. Before After Monetisation Offers: Lore-driven context outperforms generic popups. Before After Recommendations: Showing successful community teams increases win rates. Before After Anticipation: A clear goal and timeline for the next session. ## The Data Being a live product, looking at data is always my first step. D7 retention is an indicator of how much the game has captured the players attention when they are not playing to bring them back and form habits. I was clear from the start — we are optimising for D7 retention and not D1 retention. A good metric to track habit formation is sessions per day in the first seven days. By writing SQL to query our database in BigQuery, I could derive that the average session per day for new players who downloaded the game in the last 90 days was 1.3 . While that sounds great, I trusted my instinct and background in statistics to look into the distribution of this number as averages can often be deceiving. What was surprising was that we had K-shaped behaviour — the players who retained were playing so much that they pulled the average up and practically blindsided the team. K-shaped behaviour: very few players had the 'average' experience In nerd terms, this is a bimodal distribution and very few players are actually having the "average" experience. Those below the average were definitely not sticking around, which was a clear indicator why having 1.3 sessions per day still resulted in such poor D7 retention. Logistic Regression Analysis I ran a logistic regression test on D7 retention as the dependent binary variable and sessions per day as the independent continuous variable. It explained 24% of the variance in a complex model like retention. Metric Coefficient (β) Odds Ratio T-Statistic R² Intercept (β₀) 0.00334 N/A N/A N/A Sessions/Day (β₁) 0.1501 1.162 98.78 0.2416 Even just 1 additional day was 16% higher likelihood of retaining to 7 days The averages are different and it explains correlation. But how do I know if the problem is with session per day and not with overall ability of the game to delight? If that were true, we would see very low engagement in the first session — but the opposite was true with a healthy average 25 minutes first session length. The game had a healthy D1 retention, just failing to hook players into coming back and forming habits. ## Survey Research To make certain that I was chasing the correct lead, I ran a survey to understand the user stories behind the data. 73% played in-between life activities Unstructured downtime entertainment: - "When picking up my child" - "When I'm on the toilet" - "When commuting" 54% Timers negatively impacted lifestyle Game timers felt punishing and interrupted flow: - "Punishes me for having a job" - "Interrupted just as it got good" - "Use lunch break for boring quests" 21% did not think of game between sessions Driven by loss avoidance: Passive waiting rather than active consideration. 36% passively waited for timer notifications Negative anticipation: - "Forget until notification reminds me" - "Burned a few times trying to time it" - "Anxiety trying to play at work" These were all retained, active players. The experience felt worse to new players. The mismatch in the energy system and user lifestyle was creating negative feelings, and driving anticipation from loss rather than positive anticipation from expectation. This exercise helped me craft a clear persona that represented the data and the feelings — The Casual Connoisseur. They are characterised by someone who spends thousands of hours in a game, but will play casually at the start as a trial before committing. The Casual Connoisseur — our target persona ## AFK Rewards Design In the original flow, users had to come back to take an action and be rewarded. This did not generate any anticipation or excitement at the end of your previous session. Before — an energy system creates a forced action that discourages returning behaviour I redesigned the game loop to start sessions on a happy note and end sessions on moments of anticipation for future rewards. After — a positive loop that encourages habit formation My proposal was to reposition the current energy system into an AFK (away from keyboard) reward system that collects resources while the user is away. When they are ready to leave the game, we seed the anticipation by telling users they will have rewards waiting for them. Wireframing the new screens and the golden path Some hand-drawn throwaway concepts Pushing information heavy concepts Iterating on variants One of the final candidate wireframes A very rough in-engine mockup — originally every interaction was a button While I expected basic usability with this version, my team's artist made a great remark that this screen was missing feelings of delight and joy. After conducting competitive analysis we found that most games in this genre have an incredibly high bar of polish for their rewards screen — the "dopamine" factor. The chest on the screen did not look like a button and did not look clickable — in early internal testing users were confused on how to actually collect rewards. During testing, we found a 100% completion rate for actions related to collecting rewards from any screen in the game. The final mocked up user flow in an interactable Figma prototype The complete AFK Rewards experience ### Post-Launch Survey Question 1 — "What do you think of the new AFK Rewards?" - "I don't have to think about this game during work, which means I can fully focus on it after work" - "The new system is far less annoying, I only have to open once a day" - "It's nice to get all my rewards at the start and then just go into playing" - "Thank god y'all removed the arbitrary timer" 72% Completely agreed that this flow improved the overall game 81% Agreed that the new timers fit their lifestyle better ## Exit Flow & Anticipation To close off the new game loop, I added a quit screen that set goals and anticipation to encourage the user to come back. Before — A simple exit screen that does not create anticipation Using UI data, I found only 12% players closed it using the in-game dialogue . With so few players seeing this screen, I could not justify the ROI of spending lots of effort to develop animations. After — An exit screen that creates a timeline and clear goal for return time ## Recommendation System I found through analysing further data that users would get stuck at specific dungeons and quit the game after trying a few more times . This was a highly correlated factor to low D7 retention. Average attempts per dungeon vs players who quit after failing — many dungeons where users were hitting their head against a wall 46% of churned users had attempted and lost at the same dungeon 2+ times in their first week 22% of churned users had attempted and lost 3+ times After analysing what heroes they were using, it turned out they were using weaker strategic teams despite being at an adequate power level. Clear difference in heroes used by successful vs unsuccessful players about 50% users who quit Made no changes to their comp between attempt #1 and #2 We arrived at adding a recommendation system using server-wide hero usage data. Initial wireframe for recommended heroes Before After Recommendations guiding players using community data without forcing teams. ~70% users Changed their comp after loss #1 (up from 50%) And 63% of those users Completed the dungeon on Attempt #2 ## Goals & Progression Players quit after losing because whenever they lost in a dungeon, they had to begin from the start. I collaborated directly with the game designer to add a new checkpoint dungeon progression system . Initial wireframe of visual goal setting Before After A clear visual path with checkpoints replaced the old system where players start from the beginning. ### 7-Day Reward Calendar I created a 7 day reward letting players pick 1 legendary guardian , giving players choice and autonomy at the end of a long commitment. The UI was specifically designed as a countdown calendar. Daily instant gratification rewards + a big delayed gratification reward at the end of the week +3% increase in day 6 to day 7 retention ## Delight & Monetisation I added an explicit scripted moment to get the first legendary hero after defeating the first boss — a moment of delight to feel celebratory and achieved . This decision was not informed by pre-prepared data, so we A/B tested this implementation. I created this design in Figma with existing design system components +38% higher likelihood A user who received the hero would retain compared to one who did not After the success, I added an introductory offer paired with the delight moment. It performed 162% better against the old offer and increased odds of a purchasing player retaining to D7 by 65%. Before — The old offer had no context. 99% of players closed the popup in under 1 second 99.9% users Closed this popup instantly After — The new offer with context and clear value proposition More than double purchases compared to the old offer ### Iteration Process Stage 1 — Low fidelity wireframes focusing on user flow and value proposition Stage 2 — Polished by UI artist. We did not release this iteration Stage 3 — 61% better than old offer, but I wanted more lore and context The final version outperformed Stage 3 by 4% on claiming the free hero and by over 50% on purchasing the offer! ## Results +25% D7 Retention Uplift +12% D7 LTV (Revenue) The bimodal distribution eventually regressed to a normal distribution, and fewer overall players churned after solving user pain points. Impact on D7 retention followed an 80/20 principle — major impact came from the earliest, biggest changes. At that point, I proposed to dissolve the "D7 Retention" strike team and reallocate people to other features. The 'average' is lower but the outcome is more consistent for more players Mocked visualisation of approximated retention data (real data is proprietary) ## Learnings This was a major block of work that took multiple months and was shipped over multiple releases. The discovery phase started by understanding what users are lacking is beyond the screen — it's a mismatch of the game's user experience and their habits and lifestyle. With hypothesis and insights backed by data about player behaviour, we could confidently commit to this long endeavour to deliver value with constant iteration. Confidence in vision is key to keep the team motivated when the development period is long and uncertain. On a small, under-resourced team I had to basically also be the lead data analyst on this feature. If you want to see my full in-depth analysis, you can read the full report here (~15 min read). ## Testimonials Josiah Wallace Senior Game Design Manager > Data driven design is now a keystone in modern game development and there are few others like Bramha who combine the UX and Game Design know-how with the Data Analysis process & procedure as well as he does. The decisions, features, and projects he stands behind always produce measurable results, drive revenue growth, and increase player retention. Daniel Paez VP of Revenue, Immutable > Bramha was always able to breakdown each design decision into its core target audiences and their motivations, helping cut through the franticness and rash decision-making and driving towards elegant, effective solutions.