Analytics out of the box
Sessions, retention, cohort LTV and ROAS dashboards live in days — not a quarter of data-engineering. dbt models and Superset boards ship ready, running on your own warehouse.
Analytics · ML · LiveOps — for game studios
Player analytics and prediction, running in your own cloud.
Keel gives game studios one stack — event analytics, predictive ML and lifecycle messaging — deployed into your own Google Cloud. Works standalone; plugs into your MMP when you run paid UA. Your data stays in your account.
Deployed to your Google Cloud
BigQuery
PostgreSQL
dbt
Apache Superset
Unity SDK
The stack
One layer for measurement, prediction and reach.
Attribution when you need it
No MMP required. Running paid UA on AppsFlyer, Adjust or Tenjin? Keel plugs in — a sticky snapshot for clean experiments, a live layer for UA reporting, and ad-revenue / predicted-LTV postbacks back to your MMP for Value Optimization.
Predictive ML, built in
Churn, predicted LTV and send-time models — with champion-gate promotion, calibration and drift monitoring. Off by default until your data matures, then switched on model by model.
Your cloud, your data
Deployed into your own GCP project. Postgres, BigQuery and backups stay in your account. Source-available, with a continuity guarantee — if we ever step away, your install keeps running.
How it works
Assisted plug-and-play. We do the wiring.
STEP 01
Provision
We stand the stack up inside your Google Cloud project — managed by us, owned by you. No data leaves your account.
STEP 02
Connect
Drop the Unity SDK into your build. Events start flowing; optionally wire your MMP for paid-UA attribution and postbacks.
STEP 03
Read & act
Universal dashboards go live in days. Predictions and lifecycle messaging switch on as your data matures.
A realistic timeline
Keel is not a magic switch. Dashboards are live within days. Attribution settles over the first weeks as postbacks finalize. Predictive models learn from your players and turn on once the data matures — typically a few weeks of installs. Setup, tuning and calibration are a hands-on collaboration: we ramp it up with you, not over the wall.
How the prediction works
In plain terms.
Keel's models learn from your own players — not a generic benchmark.
Every session, purchase and return is a signal. Once enough history builds up, the platform learns what separates a player who is about to leave from one who is about to pay — on your game.
From then on, every player gets a live score: how likely they are to churn, what they are likely to be worth, and the best moment to reach them. Those scores drive decisions — who gets a push, what predicted value we report to your ad network.
Models are calibrated, continuously checked for accuracy and drift, and stay off until the data says they are ready. No black box.
01
Collect
Sessions, purchases, returns — captured as first-party signals in your warehouse.
02
Learn
Once history matures, models learn the patterns behind churn and spend on your game.
03
Score
Every player gets a live, calibrated score — leave risk, predicted value, best time to reach.
04
Act
Scores drive pushes and the predicted value reported back to your MMP.
Get in touch
See Keel on your game.
Tell us about your game and we will walk you through the platform on your own data. Early studios get design-partner terms.
◆ Deployed to your Google Cloud
◆ No MMP required to start
◆ Design-partner terms for early studios
◆ No MMP required to start
◆ Design-partner terms for early studios