Predictive Segments: Combine Player Models to Decide Who, When and What
One model tells you one thing about a player. Two or three together tell you what to do: who to reach, when to act and what to offer. Smartico's player models give players a value each day, and you combine those values in one segment, after the players you must never contact are taken out.
What is a predictive segment?
Predictive segmentation means grouping players by what Smartico's player models say about them, not only by what they did. Each model adds a value to the player profile once a player has enough data, most of them once a night: a churn risk rank, a value tier, a life-cycle stage, a favorite product. You use that value as a condition in a segment, just like a country or a deposit count.
- Segment on RFM Analysis, Value Score, Churn Prediction, Early VIP Detection, Favorite games and sports, Recommended amounts, Bonus Eligibility and Gamification Penetration.
- An empty value means "not enough data yet". It is never a "no".
- LTV Prediction and Best Time to send are used outside segments: LTV on the player profile and in reports, Best Time in the send settings of a campaign step.
- When a model changes a player's value, Smartico fires an event, so a change such as a rising churn risk can start a journey on its own.
- RFM category
- Loyal Customers
- Value Score (lifetime)
- High
- Churn risk
- Medium
- Favorite product
- Casino Only
- Missions
- Engaged
- Bonus Eligibility
- Unknown (not enough data yet)
Six questions, one segment
Every good campaign answers six questions, in this order. The first one is never a model: it is your own data on who you may contact. Each of the other five has a model that answers it, and the example shows the condition you would add.
May we contact this player?
Self-exclusion, cooling-off, responsible-gambling flags and marketing consent. Build them into one segment and leave it out of every campaign.
Example: not in segment "Do not contact".Why act now?
Risk is rising, a new depositor looks promising, or the player has moved to another life-cycle stage.
Example: churn risk is High or Critical.Is the effort worth it?
How much the player is worth compared with your other players, over the last 30 days or over their lifetime.
Example: lifetime Value Score is Top or High.What will feel relevant?
The product, games and sports the player prefers, and the gamification feature worth showing next.
Example: favorite product is Casino Only, or the recommended feature is Missions.Bonus or not?
Whether a bonus pays off for this player or would simply be spent. Recommended amounts give each player's usual bet and deposit, so a free bet, free spins or a mission goal can match them.
Example: Bonus Eligibility is Eligible.When should the message arrive?
The hour and day each player is most likely to be active. You choose it in the campaign step, not in the segment.
Example: send at the next best hour, any day.
Models with a red dot decide who is in a segment; models with a cyan ring shape the offer for the players inside it.
The models compared
Ten models, each built for one question. Eight of them give a value you can segment on, once a player has enough data. Net deposits, used by several models, means deposits minus withdrawals.
| Model | The question it answers | What it gives each player | Updated | Use it as a segment condition? |
|---|---|---|---|---|
| Churn Prediction | Will this player stop depositing soon? | A risk rank: Low, Medium, High, Critical or Churned, from a daily probability over your churn window of 7, 14 or 30 days. | Every night | Yes |
| LTV Prediction | How much will this player deposit in the coming weeks? | Projected net deposits for the next 15, 30 and 60 days. | Every night | No. It is on the player profile and in reports; use it to plan budgets. |
| Early VIP Detection | Does this new depositor look like your past VIPs? | A band: D0 Watch, Watch, Probable or Strong, for the first 14 days after the first deposit. | Every hour | Yes |
| RFM Analysis | Where is this player in their life cycle? | One of 12 categories, from Champions to Churned by Definition, plus recency, frequency and money scores from 1 to 5. | Every night | Yes |
| Value Score | How much is this player worth compared with the others? | Top (top 5%), High (next 15%), Medium (next 25%) or Low (the rest) by net deposits, for the last 30 days and for the lifetime. | Every night | Yes |
| Favorite games and sports | What does this player like to play? | A favorite product (from Casino Only to Mixed Preferences) and the top 3 games, game types, providers, sports and leagues. | Every night | Yes |
| Recommended amounts | Which bet and deposit amounts fit this player? | One casino bet, sport bet and deposit amount, in the player's currency. | Every night | Yes: to group players by their usual stake, to match free spins or a free bet to it, and to set personal mission goals |
| Bonus Eligibility | Does a bonus pay off for this player? | Eligible, Suppress Bonus or Unknown, plus its four signals, for example whether the player is Self Funded or Bonus Dependent. | Every night | Yes |
| Gamification Penetration | Which gamification feature fits this player next? | A tier for missions, tournaments, jackpots, mini-games and the store (Not started, Lapsed, Starter, Engaged, Power user), and one recommended feature, or Protect. | Every night | Yes |
| Best Time to send | When is this player most likely to be active? | The player's best hours and days. | Every 6 hours for active players, daily for the others | No. You choose it in the send settings of a campaign step. |
Smartico runs the models for your brands; you use their values in segments and campaigns.
Which models each goal uses
Start from the goal, not from the model. Each row is a goal with its own page: the segments to build and how to measure them.
| Goal | Churn | Early VIP | RFM | Value Score | Favorites | Recommended amounts | Bonus Eligibility | Gamification |
|---|---|---|---|---|---|---|---|---|
| Safe targeting | No model. Your own exclusion data: self-exclusion, cooling-off, responsible-gambling flags and consent. Every other row starts here. | |||||||
| New players | ||||||||
| VIP growth | ||||||||
| Reduce churn | ||||||||
| Reactivation | ||||||||
| Bonus efficiency | ||||||||
| Cross-sell | ||||||||
| Gamification engagement | ||||||||
A model can play a different part for a different goal. Bonus Eligibility decides who gets a bonus campaign, but in a churn or VIP campaign it only decides whether the offer includes a bonus. Favorite product decides who is in a cross-sell audience, and elsewhere it picks the games and sports the message is about.
Recipes that combine two or three models
Ten audiences that teams build first. A drop in play can also mean a player is cutting down on purpose: your responsible-gambling checks always come before the model.
Valuable players with rising churn risk
Players who matter to you and may stop depositing. Reach them first with something personal about the games they like, add a bonus only when Bonus Eligibility says Eligible, and stop once the risk rank drops.
Reduce churn →Champions whose churn risk is rising
RFM still sees a great past; the churn model sees a change ahead. Check the player's recent play first, then send a light, personal touch, such as a thank-you or news about their favorite games, rather than a bonus.
Reduce churn →New depositors with an early VIP signal
Players in their first 14 days who look like your past VIPs. Give them service priority: a personal welcome, help with verification and deposit limits, and your affordability checks before any VIP treatment. Never a request to deposit more.
VIP growth →Self-funded players at the top
Your most valuable players, who play with their own money rather than bonus money. A cash bonus changes little for them. Recognize them with status, a level or a store item instead.
Bonus efficiency →Offers at the player's own stake
Players for whom a bonus pays off get free spins on a favorite game at the bet they usually place, in place of one value for everyone. Build one branch per stake level. The amount follows the player's habit; it is never a reason to raise it.
Bonus efficiency →Bonus-dependent players with low value
Players who mostly play with bonus money and add little real-money play. Keep them in your regular communication without the bonus step, and never answer bonus dependence with a bigger bonus. Watch bonus cost and real-money play together.
Bonus efficiency →Churned players with top lifetime value
Players who were valuable and have stopped. Favorite product is kept after a player goes quiet, so one win-back message can talk about what they used to enjoy. Make one attempt, then stop, and send it without a bonus where your regulator limits offers to lapsed players.
Reactivation →A casino player at a big sport moment
Players who mostly play casino but already bet now and then on one league. Before a big match in that league, send one message about the match, only to players with sport marketing consent. In Great Britain, promotions that mix casino and sport have not been allowed since 19 January 2026.
Cross-sell →Power users of every feature
Players who are already power users of missions, tournaments, jackpots, mini-games and the store. Take them out of feature promotions. Recognize them, and let your VIP team reach out personally where their value calls for it, with a wellbeing check as part of the contact.
Gamification engagement →Lapsed mission players
Players who used to complete missions and stopped. Offer a mission built on their favorite games, with a goal set from their own usual bet, so it fits how they already play.
Gamification engagement →Measure the combination, not the model
A segment built from three models is only as good as the campaign it feeds. To know whether that campaign worked, compare it with players who did not get it.
- Give the campaign a control group. Smartico holds back a share of the eligible players and measures them against a conversion rule such as "deposit approved". A sticky control group keeps the same players out when they enter again.
- Compare both groups over the same period, on deposits, play and bonus cost.
- Do not judge a campaign by tier movement. Value Score and RFM rank players against each other, so a tier can change when other players change. A falling churn rank does not prove the campaign caused it.
| Judge on | Deposits, active play and bonus cost, campaign group against control group. |
|---|---|
| Not on | Opens, clicks or a model's own tier moving. |
| Keep | One control group for each campaign, held back from the start. |
FAQ
Predictive segmentation is grouping players by what models expect about them, such as their churn risk, value tier or favorite product, rather than only by what they did. In Smartico each model adds a value to the player profile once a player has enough data, and you use those values as conditions in a segment. Combining two or three of them tells you who to reach, when to act and what to offer.
RFM looks back and churn prediction looks ahead. RFM Analysis sorts players into 12 life-cycle categories from how recently and how often they were active and how much value they brought. Churn Prediction estimates how likely a player is to stop depositing in the next 7, 14 or 30 days. When they disagree, let churn decide when to act and Value Score decide how much effort to spend.
Start with RFM Analysis and Value Score: they show where each player is in their life cycle and what they are worth. A common first step for retention is churn risk with the lifetime Value Score, so you reach valuable players whose risk is rising. Then add Favorite product to choose what the message is about, and Bonus Eligibility to decide whether the offer includes a bonus at all.
No. LTV Prediction gives projected net deposits for the next 15, 30 and 60 days, and it is shown on the player profile and in reports, not as a segment condition. Use it to plan budgets for a group. To decide who is worth the effort inside a segment, use Value Score, which ranks players by net deposits over the last 30 days and over their lifetime.
No. RFM Analysis and Value Score already cover life-cycle stage and player value. Add other models when a goal needs them, for example Churn Prediction for retention or Gamification Penetration for missions and tournaments. Start from one goal, use the two or three models it needs, and add more later.
It means the model does not have enough data about that player yet. It is never a negative verdict. A new player has no Value Score until they deposit, and Bonus Eligibility stays Unknown until enough signals are available. Treat these players as a separate group with their own message, not as Low or Suppress Bonus.
Most models update every night: RFM, Value Score, Churn and LTV, Favorite games and sports, Recommended amounts, Bonus Eligibility and Gamification Penetration. Early VIP Detection updates every hour during a new depositor's first 14 days. Best Time updates every 6 hours for active players and daily for the others. Real-time segments use the new values as soon as they arrive.
Yes. A segment built on model values works anywhere Smartico asks who should see, get or join something: automation rules, missions, tournaments, mini-games, the store, banners and bonuses. Each model also fires an event when a player's value changes, so a rising churn risk or a new Early VIP band can start a journey on its own.








