Case Studies – Sushrut Studios LLP

[Partner Studio] — Retention Turnaround

Fri, 29 May 2026 18:40:16 +0000

Problem & Solution

Problem: D30 retention dropped 6 points YoY; event engagement was below benchmark.
Solution: New live event framework, churn-prediction model wired to in-game offers, weekly A/B testing cycle.

The Approach

Retention this far into a title’s life rarely fails for one reason. We started with two weeks of pure diagnosis — sessionizing telemetry, mapping the drop-off funnel day by day, and interviewing the live team — before touching a single system. The headline finding: the event calendar had gone stale and the late-game economy starved returning players of meaningful goals.

How We Worked

Diagnose

Rebuilt the retention funnel from raw events and isolated the exact days and cohorts where churn spiked.

Reframe the Live Calendar

Replaced ad-hoc events with a repeatable seasonal framework — themed arcs, predictable cadence, and clear reward ladders.

Wire Prediction to Action

Shipped a churn-prediction model whose scores fed directly into targeted, time-boxed in-game offers.

Test Weekly

Stood up a weekly A/B cadence so every economy and event change was measured, not assumed.

The Results

+20%

D30 Retention

+34%

Event Participation

12 wk

To Stable Gains

Inside the Work

[Partner Studio] — Console Port & Remaster

Sat, 29 Aug 2026 10:00:00 +0000

Problem & Solution

Problem: Tight 5-month window to port, remaster UI, and pass certification on three platforms.
Solution: Dedicated port squad, automated performance regression suite, UI rework for controller-first.

The Approach

A five-month window to port, remaster the UI, and clear certification on three platforms leaves no room for surprises. We front-loaded the risk: a one-week technical spike on the worst-case rendering path and the certification checklists, so the plan was built around the real constraints rather than an optimistic estimate.

How We Worked

De-risk First

Profiled the heaviest scenes on target hardware in week one to find the true performance ceiling before committing the schedule.

Controller-First UI

Reworked a mouse-and-keyboard interface into a console-native layout — focus navigation, safe zones, and platform input conventions.

Automate Regression

Built a performance regression suite that flagged frame-time drops per commit, so optimization never silently regressed.

Certify in Parallel

Ran the three platform certification tracks concurrently against shared checklists rather than one after another.

The Results

60fps

Locked, All Platforms

Platforms Certified

-2 wk

Ahead of Milestone

Inside the Work

[Partner Studio] — AI-Assisted QA Pipeline

Fri, 28 Aug 2026 10:00:00 +0000

Problem & Solution

Problem: Manual QA pass per release took 5 days; flaky builds delayed milestones.
Solution: Automated visual + behavioral regression tests, AI-triaged failure clusters, single-pane release dashboard.

The Approach

QA was the longest pole in the release train, and the slowness compounded: flaky builds meant re-runs, re-runs meant late nights, late nights meant missed milestones. We treated it as a tooling problem, not a headcount problem — automate the repetitive coverage and give humans a single, trustworthy place to triage.

How We Worked

Map the Bottleneck

Instrumented the release pipeline to see exactly where the five-day QA pass spent its time.

Automate Coverage

Added visual and behavioral regression tests for the high-traffic flows that consumed most of the manual pass.

AI-Triage Failures

Clustered failures by likely root cause so engineers saw grouped, deduplicated issues instead of a wall of red.

One Release Pane

Consolidated build health, test status, and deploy readiness into a single dashboard the whole team trusted.

The Results

-65%

QA Cycle Time

5→1.7

Days Per Pass

↑

Release Confidence