> ## Documentation Index
> Fetch the complete documentation index at: https://docs.instantcampaign.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Predictive scores

> Every contact gets an engagement score, a churn risk, a purchase propensity, a predicted 12-month value and a tier, recomputed daily from what they did with your email and your store.

**Predictive scores** sort your contacts by how engaged they are, how likely they are to drift away and how likely they are to buy. They are calculated with fixed formulas from data you already have: opens and clicks, bounces and complaints, goals and purchases (including Shopify orders). No AI model is involved, and each score comes with the reasons behind it.

## The scores

| Score | Range | What raises it |
| - | - | - |
| **Engagement** | 0–100 | Opens and clicks in the last 90 days. Recent activity counts more than older activity (an open loses half its weight every 30 days), and a click counts three times an open |
| **Churn risk** | 0–100 | Days without an open or click, engagement falling below the contact's own usual rate, a purchase overdue against how often they usually buy, soft bounces and a spam complaint. Unsubscribed, bounced and complained contacts are 100 |
| **Purchase propensity** | 0–100 | For buyers: how recently, how often and how much they bought, ranked against the other buyers in your workspace. Contacts who have never bought stay at 45 or below, credited for goals they reached and for engagement |
| **Predicted 12-month value** | Your currency | Average order value × the number of orders their buying rhythm suggests for the next 12 months × the chance they stay (100 − churn risk). Shown only for contacts with a purchase |

Opens from security scanners and image prefetching never count, and only marketing email is used: transactional email does not affect anyone's scores.

## Tiers

Every contact is placed in one tier, checked in this order:

| Tier | Who |
| - | - |
| **Dormant** | Unsubscribed, bounced or complained, so you cannot reach them |
| **New** | Added in the last 30 days, no purchase and fewer than 3 opens or clicks |
| **Champion** | Engagement and purchase propensity both 60 or more, churn risk under 40 |
| **At risk** | Churn risk 60 or more, and they used to open, click or buy |
| **Dormant** | Engagement under 15 and no open or click for more than 180 days |
| **Loyal** | Churn risk under 50, and engagement or purchase propensity 50 or more |
| **New** | Never emailed and never bought |
| **Promising** | Everyone else |

## Where you see them

* **Contacts**: the **Predictive tiers** strip above the table shows how your contacts split across tiers. Click a tier to filter the table; the filter is part of the page address, so export and bulk actions use it too. Add the **Tier** column from the column picker.
* **A contact's page**: the **Predictive scores** card shows the tier, the three scores with their reasons, the predicted value and when they were computed.
* **Segments**: the **Predictive score** rule. Pick **Is at least** or **Is less than** with a score and a number (for example *Churn risk is at least 60*), or **Tier is** / **Tier is one of**. A typical win-back segment is *Tier is At risk* AND *Purchase propensity is at least 40*.
* **Spark and MCP**: `get_audience_summary` includes the tier counts, and `create_segment` accepts score conditions. Only counts are shared, never which contacts are in a tier.

## When scores are updated

Scores are refreshed by a scheduled run (daily, once scheduled runs are switched on for your workspace). To rescore now, click **Recompute** on the tier strip (you need permission to edit contacts). It runs in the background; the strip shows progress and the tiers update when it finishes. Contacts added since the last run have no score yet and match no score rule.

<Note>
  In a workspace where fewer than 50 contacts have opened, clicked or reached a goal in the last year, scores are still computed but marked **Low confidence**. There is too little activity for the ranking to mean much.
</Note>

## Troubleshooting

**Every contact is New or Dormant.** Scores need history: send a few campaigns, or record purchases with [goals](/guides/audience-and-campaigns/events-and-goals) or the Shopify integration, then recompute.

**No predicted value.** It needs at least one purchase goal with a value. Purchases in a currency other than the one your workspace uses most are left out of the value.

**The tier strip says scores have not been computed yet.** No run has reached your workspace yet. Click **Recompute**.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.