Catalina Pisani — Analytics Portfolio

Selected case studies from SYBO Games (Subway Surfers) · Senior Product Data Analyst

All numbers shown in charts are fictional and used for illustrative purposes only.
Funnel analysis

Game initialization drop-off: diagnosing why users never reached the main menu

We started working on improving the game initialization funnel because there was a drop-off where we were losing users who never even got to load the game.

Subway Surfers main menu
1. Tap app icon
Game loading screen
2. Initialization
Main menu ready
3. Main menu
1. InstrumentProposed tracking each initialization step — success/fail and duration — sampled at 5% of users to control data volume and cost.
2. ModelBuilt an aggregated table in GCP/BigQuery on top of raw events to enable funnel and performance analysis across dimensions.
3. VisualizeLooker dashboard with funnels sliced by game version, platform, device, and geo — plus median duration per step.

Step-by-step funnel

% of sessions completing each step · iOS · v3.61
Median duration (s) Timeout threshold
Asset loading takes longest at 4.2 seconds.
Insights

Timeouts on asset loading: Median of 4.2s on older Android devices, frequently hitting the 5s timeout and forcing a crash.

Config fetch failures: A subset of iOS users was failing the remote config step, causing a silent crash with no retry logic.

Outcome

Engineering fixed the timeout logic and added retry handling. Drop-off improved measurably and the dashboard became an ongoing monitoring tool for new releases.

All numbers shown in charts are fictional and used for illustrative purposes only.
Engagement analysis

Mission funnel drop-off: finding the friction point blocking level progression

The mission system drives engagement by guiding players through game features. To advance levels, users must complete sets of 3 missions across 35 total levels. The PM asked for a progression analysis to understand where players were getting stuck.

Mission Set 1
Each level has 3 missions to complete.
Examples: play N runs, collect N coins, open a mystery box.

Completing all 35 levels unlocks special events and game modes.

% of users completing each mission level

Sharp drop-off between levels 5 and 6 flagged for investigation
Completion drops sharply between level 5 and level 6.
Insights

Sharp drop-off between levels 5 and 6: From ~71% to ~34% — much steeper than any other transition. This flagged a specific mission at level 6 as the likely cause.

Completion rate per mission within level 6

Mission 3 stands out as the blocker
Mission 3 has only 41% completion.
Insights

Mission 3: "Buy a mystery box from the shop." After simulating the action myself, I found the mystery box was positioned very low in the shop scroll — not visible on first open, with no visual cue pointing to it.

Outcome

When a user had this mission active and opened the shop, the game automatically scrolled to the mystery box and displayed a badge highlighting it.

  • Mission 3 completion at level 6 improved from 41% to 74%.
  • Overall drop-off between levels 5 and 6 decreased significantly.
All numbers shown in charts are fictional and used for illustrative purposes only.
Monetization · A/B test

Welcome Pack: investigating low conversion through exposure analysis

The Welcome Pack is a $4.99 limited-time offer available to new users during their first 3 days. It is surfaced in the shop and also shown as a pop-up after the user's first 3 games. We observed conversion was lower than expected, so we analyzed whether the percentage of users exposed to the pop-up was aligned with expectations — which should equal the % of users who play at least 3 games on day 1.

Welcome Pack
Pack price
$4.99
Offer window
3 days
Conversion
1.8%
Users exposed
31%

Welcome Pack was shown late in the pop-up queue

How I analyzed the exposure gap

Dashboard (standard monitoring) A/B test dashboard built with a Python script that calculates statistical significance for core metrics (retention, engagement, monetization) and compares them across variants.
Ad hoc query (custom analysis) Custom SQL query joining activity events with experiment assignment data to compute the % of new users shown the pop-up per variant, compared against the expected baseline.

Hypothesis: "If we show the Welcome Pack pop-up earlier in the session, we'll increase exposure and that will drive higher conversion and ARPDAU."

Control
Current order — Welcome Pack shown after 3 games
Variant A
2nd pop-up — after tutorial completion
Variant B
2nd pop-up — after 3 games played
MetricControlVariant AVariant B
Pop-up exposure31% 78% +151% · significant 71% +129% · significant
ARPDAU (IAP)$0.031 $0.032 · not significant $0.031 · not significant
D1 Retention38.2% 37.9% · no impact 38.1% · no impact
Outcome

Exposure increased significantly for both variants, but ARPDAU and retention didn't move. The team decided not to ship either variant and redirected focus toward the offer design itself.