CRM for iGaming · powered by ML
See what every player is worth - before the market does.
Sellrise is an ML-based CRM for iGaming. It reads every player and the quality of each traffic source, flags churn risk and hands your team the best next move at every stage. You scale profitable traffic sources and spot VIP players while they're still growing.
- Classical ML - gradient boosting, not LLM guesswork
- Deploys without migration - plugs into your stack and channels
Illustrative. Lead time depends on volume and data quality.
3-arm test on your own players. Illustrative - no published accuracy figure.
The problem
Retention runs on yesterday's averages.
Traffic judged blind
You wait out a 14-day hold before you know if a source paid - so you scale blind or freeze good budget.
Everyone retained the same
Retention sends one offer per cohort - a VIP and a random player get the exact same thing.
Reacting too late
Churn shows up in a report once the player is already gone - too late to win them back.
The product
Players, traffic, retention - the product's three screens.
Predict scores every player, Acquire the traffic quality, Engage runs retention. Each answers its own question and points to the best action.
Predict · Player intelligence
In productionEvery player's value and churn risk
Open a player's card and see predicted value, churn risk, the odds of another deposit, and the best next action.
- Predicted LTV, as a range
- Churn and second-deposit probability
- A recommended action from your own catalogue
The decision it drives
Who to win back now, and with which offer.
Acquire · Traffic intelligence
In pilotSource quality before the hold ends
Affiliate traffic quality is usually judged only after a two-week hold. Sellrise forecasts each source's value by around day 6 - so you scale profitable sources a week earlier and switch off the weak ones in time.
- Scale / Hold / Cut per source
- Fraud and abuse flagged on the funnel
- Forecast calibrated on your data
The decision it drives
Which traffic to scale today, and which to freeze.
| Source | Players | Day-6 forecast | Call |
|---|---|---|---|
| #A7 · aff_kb | 1,204 | $512 PLTV | SCALE |
| #B2 · smrt_x | 860 | $188 PLTV | HOLD |
| #C5 · push_lt | 2,015 | $61 PLTV | CUT |
| #D9 · rev_09 | 540 | abuse pattern | FRAUD |
Engage · Smart retention
In productionThe same chains, only smarter
Build chains with the drag-and-drop your team already knows. Sellrise picks the timing, offer and channel for each player, and a live queue surfaces the offers a manager approves before they send.
- ML timing and dynamic waits
- Recommended, A/B-tested bonuses
- Cheapest-first channel cascade
The decision it drives
Which offer, at which moment, in which channel - per player.
The edge
In pilotA week's head start on every source
While the market waits out the hold, you already see which sources will pay off. You scale the profitable traffic earlier and switch off the weak - without burning budget.
In pilot on your live traffic - a goal, measured on your data, not a guarantee.
Predict and Engage run in production today; Traffic quality is in pilot.
Built for
Markets where Tier-1 CRM is out of reach
Africa, India, LatAm - high volume, thin margins, players worth a few dollars. Tier-1 CRM charges for every player, and at that ARPU it never pays off - so the tooling never arrives. Sellrise stays affordable exactly where per-player pricing stops adding up.
See pricingAffordability is built in - it's about price, not a promised return.
Priced to work where margins are thin.
The economics that break Tier-1 CRM.
Plugs into the CRM you already run - no migration.
Run many brands from one place.
One place
Your whole business, at a glance.
Read value, traffic and retention across brands - with detail down to every player.
Under the hood
From data to decision
We take raw data, not ready-made aggregates - so the model can't peek at the answer and show off an accuracy that falls apart in production.
How it worksFour tables via API: users, transactions, game sessions, game events.
One vector per player, from cohort and activation speed to payments and play.
Separate gradient-boosting models, consolidated into a single decision.
A live queue surfaces an offer in about 6 seconds - a manager approves before it sends.
Manual vs Sellrise
Same team, same process - sharper calls.
Sellrise makes the timing, the offer and the traffic calls smarter - without a migration.
Measured, not claimed
We never publish an accuracy number.
Every deployment runs a three-group test on your own data - so the uplift you see is measured, not a figure we claimed.
No messages - a clean control group.
Your team's current playbook, unchanged.
Sellrise picks the timing, offer and channel.
The number you get is the number measured on your data.
FAQ
Straight answers.
Is this an LLM or generative AI?+
No. Sellrise runs on classical ML - gradient boosting and reinforcement learning. There's no generative model in decisions about money.
Do you hold our players' personal data?+
Personal data never leaves your side - phone and email stay with you. Inside Sellrise a player is just an ID, and data flows via API into ClickHouse on your own infrastructure.
Do we have to replace our CRM?+
No. Sellrise deploys without migration - it plugs into the stack and channels you already run; your team and processes stay.
How fast do we see a result?+
In a pilot, initial training runs in about 30 minutes, and traffic-quality forecasts land around day 6. These are pilot benchmarks, measured on your data - not guarantees.
Is it affordable for Tier-3 ($5-ARPU) markets?+
It's priced to work where per-player Tier-1 pricing breaks - a setup fee plus a subscription that scales with volume, not a percentage of your GGR.
How do you prove it works?+
On your own data: an A/B test with three groups (control, ML and manual) for player value, and the market's real 14-day numbers for traffic. We're honest about what's proven and what isn't yet.
See Sellrise on your own data
In 30 minutes, on your stack and your numbers, we'll show how Sellrise scores player value, forecasts traffic quality and runs retention.