Keep more of the players you’ve already won.
Dividual is a churn diagnostic for live-service games. We find where your players actually drop off, and the one change most likely to hold them.
Methods from the peer-reviewed survival literature. Validated on data where the answer is known.
The work behind it
- Ph.D. — taught at NYU, CUNY, and Virginia Tech
- author of The Loot Loop
- WP-2026-01 — working paper, July 2026
- validation on planted truth
Selected artifacts
- 01 The paper. Matchmaking as performative prediction — the mapping stated, dated, citable.
- 02 The proof. A churn read run blind against a planted truth, and what it accepted, rejected, and refused.
- 03 The demonstrations. The retraining trap, the shutter, the self-exciting loop — formulas in the footer, nothing fitted, nothing hidden.
What you leave with
-
A location.
Where in your game the leaving begins, and the shape it takes across your whole base.
-
A verdict.
Which of your team’s assumed causes survives a test, and which one only looked real.
-
A ceiling.
How early you can see a player going, and where that early read runs out. We tell you both.
How it works
- 01
Share a slice.
A small, anonymized sample of your telemetry — about an hour of your team’s time.
- 02
We diagnose.
A fixed three-to-four-week read. Clear findings, in plain language.
- 03
You act.
The one change to make, and the experiment that confirms it worked.
Proof
We show our work — on data where the answer is known.
Hand a churn model real data and it can score well without being right, because real data has no answer key. So we prove ours on a simulated game where we planted the truth: it finds the real cause, ignores the convincing decoys, and refuses the traps that flatter an ordinary model.
See how it worksThe service
The Diagnostic
A fixed-scope read on a slice of your telemetry, under NDA. You get a short, plain-language report, a working session, and a pre-registered experiment to confirm the cause. Fixed price, fixed timeline, about an hour of your team’s time.
Who
Dividual is built by Adnan Selimović, Ph.D. — a player researcher who has spent a decade on one question from both sides: how live-service systems shape what players do, and how to measure it well enough to act.
The framework beneath this site is a decade of work — the theory of why players stay, written down and made measurable. Dividual is its applied edge, shown in the open. A validation that recovers a planted truth blind. A working paper, dated July 2026, that names the fixed point engagement optimizers converge to. Three demonstrations whose every number reproduces from the formulas printed beneath them.
He taught first — NYU, CUNY, Virginia Tech. Then came humanitarian logistics, where reading a situation correctly is the whole job. After that he consulted for large companies, where a read has to survive production systems and the people accountable for them. The Loot Loop, his field notes on the psychology of engagement, runs alongside.
The name
Why “Dividual”
You arrive here already divided. Before you type a word you are a credit score, a churn-risk flag, a lookalike audience, a feed rank that settles what you’ll see before you’ve decided to look — a spray of numbers that stand in for you and, more and more, act in your place. Deleuze had a word for the person once the machines have finished cutting them up: dividual. An individual is what can’t be divided. A dividual is what already has been, then sold the pieces back as a service. The reader this site assumes is exactly that composite — assembled from data exhaust, scored in the dark, addressed by first name in an ad and never once seen. We named it Dividual because that is who’s reading, most places, most hours of the day. Here we’d like to reach the other one.