Xizi (Allen) Huang
Open to New Roles

Data Analysis Portfolio

Xizi (Allen) Huang  ·  Senior Data Analyst

2021–2025 @ Garena Free Fire (~100M DAU)  ·  Specialized in causal inference, A/B testing, and monetization strategy across 15 global markets.

0 Peak DAU
0 Experience
0 Global Scale
01 / Philosophy
01 / Philosophy

How I Think About Analysis

Data analysis is not just about reporting what happened; it's about reducing uncertainty for critical decisions.

01

Bias Over Volume

A small, clean, unbiased sample is infinitely more valuable than a massive, confounded dataset. I prioritize causal clarity over sheer data scale.

02

Decision-Backwards

I start with the decision that needs to be made, then work backwards to the metrics and models required to inform it.

03

Build Reusable

One-off analyses are expensive. I focus on building frameworks and monitoring systems that turn a single insight into a standing capability.

02 / Case Studies
02 / Case Studies

The Pivot Framework

Four deep dives into how analytical pivots moved the needle on product strategy.

Project 01  ·  2023

Combating "The Cheat Market"

From reactive banning to structural deterrence in Trust & Safety

Detection Recall +25%SOP Adopted
Context Reactive banning failing against scale
Pivot Incentive-based detection infrastructure
Impact Structural deterrence and economic friction
Cheater Impact on Network Quality — Indexed Trend
Reactive banning era vs. deterrence-via-friction era (target regions, index = 100 at baseline)
0 50 100 Friction gates live ≈70 (−30%) Reactive banning: impact stuck at baseline Ban waves Structural deterrence
Illustrative trend based on case narrative; endpoint reflects the reported ~30% reduction.
Context

The Trust & Safety team was stuck in an "arms race" — banning cheaters only for them to return with new accounts. The volume was overwhelming, and the reactive approach was failing to protect the player experience. We needed to move from chasing individuals to breaking the cheat economy.

The Pivot

I shifted the focus from banning to deterrence via friction. By analyzing the network patterns of high-frequency cheaters, I identified a specific infrastructure bottleneck they relied on. Instead of just banning, we introduced dynamic verification gates triggered by these network signals, making it economically unviable to run cheat-services at scale.

What Changed

The shift to structural deterrence reduced cheater impact on network quality by ~30% in target regions. More importantly, it produced a scalable SOP for infrastructure-level intervention that the engineering team now deploys automatically when certain network anomalies are detected, rather than waiting for manual analysis.

Cheater Impact −30%
Detection Recall +25%
Scope Global
Project 02  ·  2024

The "Lag" Paradox in LATAM

Why "better" network metrics led to worse player sentiment

Causal InferencePing −15ms
Observation Lower ping metrics but higher churn
Pivot Focus on packet loss jitter variance
Result Stabilized experience alignment
The Lag Paradox — Average Ping vs. 95th-Percentile Jitter
Engineering optimized the wrong metric: ping improved, but jitter — the real driver — was never addressed
What engineering optimized 110ms 95ms Before After Avg Ping −15ms ✓ What players actually felt Jitter before Jitter after Jitter unchanged → churn +2% despite ping gain
Illustrative comparison based on case narrative; ping values reflect the reported ~15ms gain, jitter shapes show variance persisting because it was never the optimization target.
Context

Engineering had optimized for average latency (Ping) in LATAM, successfully reducing it by ~15ms. Paradoxically, player sentiment data showed a spike in "lag" complaints and a 2% increase in churn among competitive players. The "better" the metrics got, the worse the experience felt.

The Pivot

Using causal inference (PSM), I proved that players are far more sensitive to latency variance (jitter) and packet loss than to absolute ping. A stable 100ms connection is better than a 50ms connection that spikes to 200ms every ten seconds. We pivoted the engineering KPI from "Average Ping" to "95th Percentile Jitter," which finally aligned technical metrics with player reality.

What Changed

The KPI pivot led to a routing change that prioritized stability over raw speed. Lag-related churn stabilized, and the "Lag Paradox" was solved. This project established variance-based monitoring as the standard for all 15 global markets, ensuring we never again optimize for a metric that doesn't reflect player experience.

Avg Ping −15ms
KPI Pivot Jitter
Churn Impact Stabilized
Project 03  ·  2024

Combat Balance Guardrails

AUG controversy & Wukong Skill Skin launch strategy

KD Gap 28%29% New Buyers
Conflict Aggregate stats vs. community perception
Pivot Same-player controlled comparison + fairness guardrails
Impact Evidence-backed fix, pay-to-win crisis avoided
Context

The design team believed AUG was balanced based on aggregate KD stats. Community pressure said otherwise. These two positions were both right in their own frame — aggregate metrics are confounded by player-skill selection bias, so they can't settle a balance dispute.

For the Wukong Skill Skin launch, the risk was different: monetization and fairness were in direct tension. The question was how to set pre-launch guardrails that let us capture revenue without crossing into pay-to-win territory.

The Pivot

For AUG: I proposed a same-player controlled comparison — comparing each individual's KD with AUG versus a neutral AR baseline — to remove the skill confound entirely. AUG outperformed the baseline by ~28%, moving the design team from "community perception" to an evidence-backed fix.

For Wukong: I set fairness guardrails before launch — a hard ceiling on per-player KD uplift. Monetization success that violates that ceiling is not a win. This prevented a "pay-to-win" crisis while maximizing revenue.

Same-Player Controlled Comparison — AUG vs. Neutral AR Baseline
Each player's KD with AUG measured against their own baseline, removing skill-selection confound
Neutral AR baseline Same player w/ AUG +28% mean KD gap KD (indexed)
Illustrative paired design based on case narrative; the +28% mean gap is the reported result. Aggregate stats looked balanced — the controlled comparison did not.
What Changed

These two cases produced a reusable pre-launch evaluation template: quantify gameplay impact via controlled comparison, define acceptable outcome ranges before release, and treat fairness guardrails as a launch gate rather than a post-hoc audit.

AUG KD Gap 28%
Wukong Pick Rate +4.8%
New Buyers 29%
Project 04  ·  2025

Revenue Decline Diagnosis

Vietnam market: revenue falling despite stable activity

Content SaturationNew Content →50%
Anomaly Revenue falling while DAU stays stable
Pivot Item ownership-rate → content supply gap
Impact Pacing policy; new-content revenue ≈50%
Context

Vietnam showed a multi-month revenue decline while DAU and session time stayed stable. The revenue team's initial read was weak player sentiment or market softness. I pushed back: a DAU–revenue divergence of this magnitude rules out an engagement problem. The cause had to be structural, inside the monetization system itself.

The Pivot

I introduced an item ownership-rate metric — measuring what share of featured items were already owned by active high-value payers. The data showed that 80%+ of re-featured content was already in their inventories. The problem was a content supply gap, not a demand problem. We pivoted from "how to re-engage payers" to "how to ensure high-value payers always have something new to buy."

Vietnam Revenue Mix — New Content Share Before & After Pacing Policy
DAU stable throughout; decline driven by ownership saturation (>80% of featured items already owned)
0% 50% 100% Pacing policy live Re-featured inventory swells New content share: dips → recovers to ≈50% Saturation phase Recovery phase
Illustrative mix shift based on case narrative; endpoints reflect the reported recovery of new-content revenue to ~50% of total.
What Changed

The content team adopted a revised pacing policy: enforced cooldown periods on re-featured items and a guaranteed monthly new SKU for the high-value segment. After the change, new-content revenue recovered to ~50% of total Vietnam revenue. The ownership-rate metric is now a standing leading indicator in market monitoring.

Activity Trend Stable
Ownership Rate 80%+
New Content Rev ~50%
03 / Capabilities
03 / Capabilities

Technical Toolkit

Methods and tools used to drive decision-making across global markets.

Analysis Methods

Causal Inference A/B Testing PSM Cohort Analysis Funnel Decomposition

Data Stack

SQL Python Hive/Spark Tableau Excel

Product Domains

Trust & Safety Network Experience Combat Balance Monetization

Cross-functional

Product Strategy Engineering Partnership 15 Global Markets Monitoring & SOPs