Decisions, not dashboards.

What Ilara does

Ilara scores every player individually, in their first session, and attaches the next move to that score. The unit is one player rather than a segment, and the timing is the session they are in rather than next week’s report.

The four things it does, in order: Operate, one console across your live titles. Predict, churn, spend, stuck and fraud, scored before the first session ends. Personalise, the move chosen per player rather than authored per segment. Protect, the fraud rules you already own, run on every account.

Who it is for

Studios that are not going to staff a data science team. The alternative on offer is a senior hire you cannot justify, six months to a first model, and a notebook nobody can operate at the end of it.

We are an addition to your stack, not a replacement for it. Your analytics tool tells you what happened. Ilara decides what to do about it, per player, while it still matters.

How we work

Every engagement opens with a first run on your own users, from the event export your analytics tool already writes. No software development kit (SDK), no new tracking, nothing to rebuild. You see the models and what they found on your players before you decide whether to run anything.

Every live play carries a control group. If we cannot show you the difference against a holdout, we do not claim the difference.

How we handle evidence

This is the part worth checking us on.

• We publish what did not move. When a test leaves retention, sessions and revenue per player flat, that is in the write-up, because removing 95% of your offer impressions could have cost you something and the flat lines are the evidence it did not.
• We publish the nulls. On one title, who returns turned out to be predictable and who pays did not. We say so.
• We publish the caveat with the number. Where a figure rests on a proxy, the proxy is named next to it. Where a result is not statistically powered, we leave it out rather than dress it up.
• We do not name clients or quote their economics. Results are anonymised, and figures they have not cleared stay in their data room, not on our website.

You can read the working in the articles and the case studies. Every number in them carries its method.

Press

For press enquiries, briefings or assets, email [email protected] with “Press” in the subject.

Careers

If you work on applied machine learning, live game operations or data engineering, and the evidence standard above is the reason you are writing, send something you have built to [email protected]. We would rather read your work than a CV.

Contact

Everything reaches the same place: [email protected]. If you want a first run, send the export you already have and we will come back with the models on your own players.