I had just started working on an established freemium web game targeting children in the 6 - 10 year old range when, senior leadership became interested in getting more users to check their notifications page. This page was both a vector for marketing our paid subscription offers and where subscriber benefits were delivered to the players. I was personally interested in seeing if there was something we could do to meaningfully affect player engagement patterns, since many of our new features launched to relatively lacklustre engagement. While our most active cohorts always checked out the new updates, we always seemed to have significant issues with feature discover in our core users.
Article Summary
Domain: Mobile/Web Gaming Analytics & UX Optimization
Core Problem: Low feature discovery among core child cohorts (ages 6–10) and poor traffic to the notification hub (the primary delivery vector for subscriber perks and conversion marketing).
Methodology: Single-variate A/B test (Red-dot Badge MVP), full-funnel telemetry tracking (Exposure -> Click-Through Rate -> Scroll Depth -> Downstream Conversions).
Key Findings:
Micro-Behavioral Lift: The notification badge drove a statistically significant surge in Notification Center visits and increased mean notifications opened (shifting distribution mode from $0 \rightarrow 1$).
Macro Disconnect: Zero movement in macro metrics (CvR, D7/D30 Retention), proving visual cues affect feature discovery and navigation mechanics, not intrinsic downstream conversion intent.
Outcome:
Proved behavioral modification capabilities in young player cohorts.
Established the "Attention Manager" Framework: Results served as the foundational research for a full UI overhaul, deploying system-wide visual highlighting to guide player focus toward high-value feature loops.
Read more: Directing Focus: How a Basic UI Test Shaped an Attention Framework
I was working on a freemium web and mobile video game targeting an audience of children and teens. Up to this point, the monetization model had been purely subscription-based, but the publisher wanted to boost revenue by introducing a premium currency and an in-game micro-transaction shop.
This involved massive structural changes to the game. Many features and items that were previously subscriber-only were unlocked to non-subscribers in exchange for premium currency. On top of that, basic progression loops were integrated directly into the premium currency spend cycle.
Article Summary
Domain: Mobile & Web Gaming / Virtual Economy Analytics
Core Problem: A 6-month post-launch collapse in quarterly overall revenue following the introduction of a freemium premium currency and shop.
Methodology: Cross-platform cohort segmentation, platform penetration audits, time-series cannibalization analysis, cross-feature currency sink/faucet balancing, and iterative A/B testing.
Key Findings:
- Uncoordinated Faucets: Multiple development teams built independent reward loops (Daily Logins, Goals, Live-Ops), creating an economy where 2 hours of monthly play yielded a 50% surplus of free currency over progression needs.
- Subscription Cannibalization: Power users leveraged free currency surplus to buy subscriber-only perks without paying, tanking core subscription ARR.
- Cognitive / UX Mismatch: Younger cohorts failed to understand mobile currency tropes, resulting in lower progression and unspent balances, while marketing space was stripped from the proven subscription model.
Outcome:
- Rebounded core subscription sales by re-establishing primary marketing triggers.
- +14% increase in premium currency participation via onboarding explainers and video tutorials.
- +11% boost in overall play intensity after re-tuning Live-Ops currency depth and capping free faucets.
I was working on a large educational game targeted at young children. At the time, the developer had discovered issues with with tutorial or FTUE (First Time User Experience) completion rates and I asked if I could look more deeply into the issue. From everyone's experience at the time, this issue had been around for some time, but they didn't have good tracking of the onboarding funnel, so they couldn't see what the underlying issues were.
Article Summary
Domain: EdTech / Gaming Analytics
Core Problem: Low FTUE completion rates and high early-session churn in a gamified math platform.
Methodology: Funnel drop-off analysis, cohort binning (accuracy vs. retention), cross-tabulating accuracy thresholds against 7-day / 14-day / 28-day retention metrics.
Key Finding: Unshielded initial placement tests caused an immediate accuracy drop (<25%) in struggling students, driving high FTUE bounce rates. Ideal retention peaked at a 70% accuracy threshold.
Outcome: Rescheduled the placement engine to fire asynchronously post-FTUE, preserving onboarding conversion while maintaining pedagogical profiling.
The way the game worked, you could explore everything quasi-freely without engaging in any educational content, but if you wanted to progress you had to engage in combat encounters, and to do anything in an encounter you needed to answer educational questions.
Read more: Educational Gaming and the Impact of Existing Knowledge
This was something I put together for the interest of colleagues at work, and I got some good feedback from it so I thought I'd share it publicly. I want to preface this with the fact that I'm neither a physician, an epidemiologist, a virologist, nor any other kind of medical expert (I don't even play one on TV!), and I don't work even tangentially in healthcare. I'm just a numbers monkey who reads the news.
I've seen a lot of people on every possible social network posting graphs of the local COVID-19 infections. Sometimes they're line graphs, sometimes they're bar charts, sometimes they're even line graphs and bar charts, but I haven't seen much in the way of model fitting. So, I thought I'd throw a couple of basic models at the data and see if they met expectations.
Article Summary
Domain: Public Health Analytics & Epidemiological Time-Series Modeling
Core Problem: Aggregate provincial COVID-19 daily case totals in Nova Scotia exhibited anomalous tail behavior and peak deviations, failing to fit standard single-distribution epidemic models.
Methodology: Non-linear parametric curve fitting, time-series disaggregation into sub-populations, cross-referencing public transit/retail exposure advisories, and composite distribution summation.
Key Findings:
Sub-Population Splitting: The outbreak was driven by two distinct, asynchronous transmission vectors: broad community spread versus insulated institutional spread (Long-Term Care Homes / Northwood).
Distribution Divergence: Community daily cases followed a standard Gaussian/Normal distribution (reflecting early social distancing efficacy), whereas LTCH cases fit a Log-Normal distribution (rapid acceleration and delayed decay characteristic of high-density institutional environments).
Outcome:
Composite Model Accuracy: Summing the Gaussian (Community) and Log-Normal (LTCH) curves resolved the aggregate model residual errors, accurately tracking both the mid-April peak and the tail decay into mid-May.
Transmission Lag Mapping: Mapped the temporal offset between initial community exposure advisories (Halifax Transit, local retail) and the subsequent onset of institutional outbreaks.
Read more: Modelling the NS COVID-19 Outbreak - 12 May, 2020