> For the complete documentation index, see [llms.txt](https://help.smartico.ai/welcome/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.smartico.ai/welcome/products/ai-models/early-vip-detection-model.md).
# Early VIP Detection Model
Early VIP Detection is a feature designed to help operators identify high-value players early — before their value is obvious from deposits alone.
The model flags new depositors whose early behaviour resembles players who later became VIPs. Instead of waiting weeks for a player to "prove" their worth through volume, operators can spot future high-value players within days of their first deposit and act while engagement still feels personal.
This gives operators a critical head start: earlier host contact, smarter budget allocation, and retention saves that happen while a player is still winnable — not after they've gone quiet.
Early VIP Detection is the upstream partner to **Value Score**. Value Score tells you who is valuable *now*; Early VIP Detection predicts who will *become* a Top or High Value player, weeks before Value Score can confirm it.
### Model Activation
Early VIP Detection is activated **on request** — it isn't enabled by default. To switch it on for a label, just request activation for that label.
Once activated, the model calculates scores for **each brand inside the label**. Because scoring is calibrated per brand, this initial calculation is what tailors the model to each brand's own player base.
The feature becomes **usable once this calculation completes**. From that point on, `early_vip_flag` and the tier bands (`D0_WATCH`, `Watch`, `Probable`, `Strong`) are available for that label's brands in segmentation, journeys, and all real-time activities — and all scoring runs, including the same-day updates below, begin automatically. No further setup is needed per brand.
### What the Model Flags
Each eligible player is scored and placed into one of four bands, escalating by confidence:
**D0 Watch:** A same-day signal based on first-deposit size. An early "worth watching" marker before behavioural data exists.
**Watch:** Early behaviour shows some resemblance to future VIPs. Worth nurturing, lower confidence.
**Probable:** Behaviour meaningfully resembles future VIPs. A strong candidate for accelerated treatment.
**Strong:** High-confidence match to the future-VIP profile. Prioritise for immediate host outreach.
The bands are progressive — a player can move up as more of their first two weeks unfolds and the picture sharpens, and this can happen within a single day as they play.
### Where to Find It
Early VIP Detection outputs are delivered directly into the CRM for each eligible player:
* **`early_vip_flag`** — marks a player as an early VIP candidate.
* **Tier band** — `D0_WATCH`, `Watch`, `Probable`, or `Strong`.
* **Probability and confidence** — how strongly the player resembles future VIPs, and how sure the model is.
These are available for segmentation and for all real-time activities — journeys, automation rules, missions, tournaments, and host queues — so early VIP candidates can be actioned the moment they're flagged.
### How the Model Evaluates Players
#### Eligibility
The scored population is **first-time depositors (FTDs)**. Each FTD is evaluated across their first 14 days from the first-deposit date.
#### Scoring
Players are scored at fixed checkpoints across their first two weeks — **D0, D1, D2, D3, D5, D7, and D14** — so a candidate can surface as early as day zero and be re-evaluated as behaviour accumulates.
Two methods run across these checkpoints:
* **D0 (same day):** A fast heuristic based on first-deposit amount, producing an immediate "watch" signal before any behavioural history exists.
* **D1 onward:** A brand-calibrated model that assigns each player a VIP probability with a confidence level, based on their early behaviour.
#### Updated through the day
Scores don't wait for an overnight batch. Every eligible player is **re-scored hourly**, so a player who deposits or plays at 14:00 in a way that resembles a future VIP has their properties updated that same hour — not the following morning. Your team can act while the player is still in session.
Within a day, the flag and tier only move **upward**: once a player is flagged, the flag stays on for the rest of the day, and a band can promote (Watch → Probable → Strong) but won't drop back. This keeps host queues and journeys stable — a player won't appear and disappear from your team's list as the day goes on. A full recalculation runs each night and reconciles every player's standing.
#### Calibrated per brand
Scoring is **calibrated per brand** — the model learns what a future VIP actually looks like for each specific brand, rather than applying one generic profile everywhere.
Signals are also **brand-relative**: "high potential" means high against that brand's own players, not against a generic industry benchmark. A player is always judged relative to the right peer group.
#### Precision-first
The model is deliberately **precision-first**: it prefers fewer, higher-confidence flags over noisy volume. The goal is a clean, trustworthy shortlist your team can act on — not a long list of maybes. Operators can confidently invest more per flagged player because the flags they receive are ones the model is genuinely confident about.
### How It Works
As each FTD moves through their first 14 days, they are scored at each checkpoint and assigned the appropriate band. High-confidence candidates (Probable, Strong) surface to the CRM for immediate action; lower-confidence candidates (D0 Watch, Watch) enter lighter nurture tracks.
The flag and band update **as the player's activity accumulates during the day**, as well as at each later checkpoint — so the CRM always reflects the latest read, not last night's picture.
### Marketing Suggestions
| Band | Primary Goal | Best Channel | Suggested Treatment |
| ------------ | --------------------- | --------------------- | ------------------------------------------------------------- |
| **Strong** | Convert to VIP fast | Phone / VIP Host | Same-day host call, premium welcome offer, loyalty fast-track |
| **Probable** | Accelerate onboarding | IM / SMS / Host | Personal host intro, tailored bonus, enhanced journey |
| **Watch** | Nurture upward | Email / Push / In-App | Enhanced nurture track, gamified missions |
| **D0 Watch** | Early soft touch | Push / In-App / IM | Personal thank-you, welcome sequence, monitor |
#### Ways operators use the flag
**Tiered onboarding journeys.** Branch the welcome flow by band — Strong/Probable get accelerated VIP onboarding, Watch gets enhanced nurture, the rest stay standard.
**Early host outreach.** Hand high-potential players to a human host while it still feels personal — a D0 Watch triggers a soft touch, a Strong flag triggers a same-day host call or message.
**Smarter budget allocation.** Spend more per player on fewer, higher-confidence flags, and stop over-bonusing the broad FTD base. Same budget, concentrated where predicted value is highest.
**Early retention saves.** If a flagged player's engagement dips in the first two weeks, trigger a priority win-back. Saving a future VIP at day 5 beats reactivating at day 60.
**Close the acquisition loop.** Tie the VIP flag back to source, campaign, and creative to optimise paid spend on downstream VIP yield — not just cheap registrations.
**Gamification & loyalty fast-track.** Auto-enroll Strong players into higher loyalty tiers or invite-only programs early, with tailored missions and tournaments to build stickiness first.
***Usage in Segmentation:***
***Reporting:***
***User Profile:***
### FAQ
**Q: When does a player first get scored?** On their first deposit (D0), using a fast deposit-amount signal. Behaviour-based scoring begins at D1 and continues at D2, D3, D5, D7, and D14.
**Q: How often are scores updated?** Hourly. Each eligible player is re-scored through the day, so a qualifying deposit or session in the afternoon updates their flag and tier that same hour rather than the next morning. A full recalculation runs each night.
**Q: What makes a player "high potential"?** Resemblance to that brand's own future VIPs. Scoring is calibrated per brand and signals are brand-relative, so a player is measured against their brand's peers — not a generic industry bar.
**Q: Is there one model per brand or one per label?** Calibration is **per brand**. The model adapts to each brand's player base rather than running a separate model per label.
**Q: Why does the model flag relatively few players?** By design. The model is precision-first — it favours fewer, higher-confidence flags over volume, so the flags your team receives are ones worth acting on.
**Q: How is this different from Value Score?** Value Score measures a player's value based on actual recent activity. Early VIP Detection predicts, in a player's first two weeks, who is *likely to become* a Top or High Value player — so you can act before the value shows up in the deposits.
**Q: Can the flag change over time?** Yes. A player is re-scored hourly and at each checkpoint through D14, and can move up in band as their early behaviour develops. Within a single day the flag and band only move upward — a player won't drop off your team's list mid-day. The nightly recalculation reconciles each player's standing.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://help.smartico.ai/welcome/products/ai-models/early-vip-detection-model.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.