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

# Propensity scores

> Rank the people in a segment by how likely they are to buy, and save the best-scoring people as a new segment

A propensity score helps you prioritize people who are more likely to buy from your business, using patterns in your customer and order data. Scores run from 0 to 100, and a higher score means a person is more likely to buy.

Scores appear on segments and in [Predictive audiences](#predictive-audiences).

## Before you start

* **Scoring has to be turned on for your organization.** It needs your customer and order data to be connected. Once both are connected, your Minerva contact can turn it on. If scoring isn't turned on, a segment's **Score** shows **Unlock**, and Predictive audiences shows **Unlock propensity scores** with a **Request scoring** button.
* **A full score needs the full segment.** A preview can show an estimate earlier, but the full distribution comes after the complete segment is built.

## Score a segment

**1. Open the segment**

Open a segment and find **Score**. It shows where the segment ranks against the US baseline, such as **Top 20%**.

**2. Request the full score**

Select the score. If the segment only has an estimate, the dialog says so. Select **Score segment**. Saving a new segment can also start a full score.

**3. Wait for scoring to finish**

The segment and its scores are built separately. The segment can be ready while scoring is still running.

**4. Read the distribution**

Select the score again to open **Propensity score distribution**. Check the score and how many people were scored, shown as the number scored out of the segment's total.

## How scores are built

Minerva trains a scoring model for your organization on your own customer and order history, combined with Minerva's demographic, household and interest data. The model learns what people who have bought from you have in common and gives each person a score from 0 to 100.

## Read a score

| On screen | What it means |
| - | - |
| Score from 0 to 100 | Higher means more likely to buy. Use scores to rank people and compare segments. |
| Segment score | The average score of the people who were scored. The Segments list shows it as a number. |
| Top or bottom percentage | Where that score ranks against the US baseline, such as **Top 20%**. This is what **Score** shows on the segment itself. |
| Scored count | How many people in the segment received a score. People without a score aren't counted as zero. |
| Estimate | A directional score based on part of the segment, shown before the full score is ready. The Segments list marks it with a **?**. |
| Score distribution | Opens as **Propensity score distribution** and shows the number of scored people in each score range. Changing the bar size (1, 2, 4, 5 or 10 points) changes the chart, not the scores. |

## Save the best-scoring people

**1. Select a score range**

On the distribution, set the **Minimum propensity score** and **Maximum propensity score**. The count beside them shows how many people are selected. **Clear** resets the range.

**2. Create the segment**

Select **Create segment**, enter a **Segment title** and select **Create**.

**3. Activate it**

Open the new segment when it's ready and select **Activate**.

> [!NOTE]
>
> A segment created from a score range is fixed. It keeps the people it had when you created it, even if you later edit or rebuild the segment it came from.

A score range and a top percentage are different selections. The top 20% of scored people isn't the same as everyone scoring 80 or higher.

Selecting more than 1 million people shows a warning that creating the segment may take longer than usual.

## Predictive audiences

**Predictive audiences** shows the propensity score distribution across Minerva's prospects for your organization, under **Third party distribution by propensity score**. Select a range and create a segment in the same way. This distribution covers a different set of people from the one on an individual segment.

Open [**Predictive audiences**](https://app.minerva.io/predictive-audiences), or ask Minerva Agent to open it. It doesn't have its own item in the sidebar.

## How to use a score

* **Prioritize.** Reach the highest-scoring people in a segment first.
* **Select the best prospects.** Choose a score range and create a segment from it.
* **Compare segments.** A segment's score shows where it ranks against the US baseline.

A score ranks people. A score of 80 doesn't mean 80% of those people will buy.

## Limits and billing

Scoring a segment and creating a segment from a score range don't use your record export allowance. Activating that segment follows the same rules as any other.

Scoring ranks the people in a segment. It doesn't add missing contact details or change your limits.

## Troubleshooting

| Message | What to do |
| - | - |
| **Unlock propensity scores** | Scoring isn't turned on yet. Connect your customer and order data, then select **Request scoring** or ask your Minerva contact to turn it on. |
| **No audience members received a score in this run.** | No one in the segment could be scored. Check that the segment contains matched people. If it does and you expected scores, contact Minerva. |
| **Scoring failed. Retry to generate the propensity score distribution.** | Select **Retry scoring**. |

If a segment has a full score but no distribution, select **Rescore** to generate the distribution.


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