Squlp

Economic intelligence applied to video games. Scientific method and AI models for economy, balance and monetisation decisions, before they hit production.

Context

Studios iterate their economy on instinct or through slow A/B tests. Every wrong call costs retention and LTV. The production → live → measurement → adjustment loop is far too slow.

Solution

Predictive models that simulate player economic behaviour against design changes. The team sees the expected impact before touching the build — hypotheses validated in hours, not weeks.

The AI inside

What runs on its own.

1. 01

   Predictive player models

   They simulate the economic behaviour of real cohorts against design changes, trained on 18 months of telemetry.

2. 02

   Hypothesis comparator

   Translates the model into expected impact on retention and monetisation before touching the build. Signal in hours, not weeks.

3. 03

   Guardrails in production

   They watch the key metrics live and fire alerts and an automatic rollback if the economy drifts.

Screenshots

• 
• 
• 

Indicators

• AI Predictive player models
• <1d Hypothesis → signal loop
• 3 Dimensions: economy · balance · monetisation

What it solves

• Hours Hypothesis → signal loop, against weeks of A/B testing
• Pre-build The expected impact is visible before touching production
• Rollback Guardrails that revert if the economy drifts

Process

How we solved it.

1. 01

   Data

   Audit of in-game events + cohort mapping

2. 02

   Model

   Training on 18 months of telemetry · validation against real outcomes

3. 03

   Product

   Dashboard that turns the model into live-ops decisions without needing data science

4. 04

   Hardening

   Guardrails, alerts and automatic rollback if the key metric drifts

Arguing about numbers instead of hunches. That is the whole thesis.