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

# How Allyy works

> The end-to-end journey from raw supporter data to optimised decisions

Allyy turns your supporter history into decisions. The platform follows the same five stages every time, whatever the business question:

**1. Connect your data**

You connect one or more **data sources** (your CRM, warehouse, files, or a streaming feed) inside a **dataset**, and map them to Allyy's data model. See [Connect a data source](https://docs.allyy.io/processes/data.md).

**2. Build a model**

Allyy trains a **model** on your history to answer a specific question — *will this person donate?*, *will they lapse?*, *how much should we ask for?* See [What Allyy can predict](https://docs.allyy.io/models/what-allyy-can-predict.md).

**3. Generate predictions**

The trained model scores your supporters, producing a **prediction** (a probability or an amount) for each one. See [Generating predictions](https://docs.allyy.io/predictions/generating-predictions.md).

**4. Make a decision**

Allyy combines those scores — for example propensity × expected amount — and **optimises** them into a ranked, budget-aware contact list. See [From predictions to decisions](https://docs.allyy.io/predictions/decisions.md).

**5. Act, then measure**

You **export** the decision to your campaign tools, run the campaign, and use the **dashboards** to compare what actually happened against what the model expected.

## The pieces, and how they fit

Everything in Allyy is one of a small number of objects. If you understand these, you understand the platform:

| Object | What it is |
| - | - |
| **Dataset** | A project-level container holding the data sources for one body of work. |
| **Data source** | A connection to your data (BigQuery, SQL Server, CSV, SFTP, Salesforce, …), batch or streaming. |
| **Model** | A trained predictor for one question (propensity, churn, expected amount, lifetime value, …). |
| **Prediction** | The scores a model produces for a population at a point in time. |
| **Decision** | Scores combined and optimised into a ranked action list for a campaign. |
| **Export** | A decision pushed out to your tools (or pulled via API). |
| **Workflow** | Automation that runs syncs, training, scoring and exports on a schedule. |

> [!NOTE]
>
> The **Monitoring** page tracks all of these in one timeline — Datasets, Models, Predictions, Exports and Workflows — so you can see at a glance what ran, when, and whether it succeeded. See [Monitoring & logs](https://docs.allyy.io/processes/monitoring_dashboard.md).

## Descriptive vs predictive

Allyy gives you two complementary lenses:

> [card] Predictive
>
> Models, predictions and decisions look **forward** — who to contact next and how.

> [card] Descriptive
>
> The [Analytics dashboards](https://docs.allyy.io/analytics/overview.md) look **backward** — what your supporter base looks like today and how it's trending, no modelling required.

The two are complementary: the descriptive analytics show what your file looks like today, and the predictive models act on it. You can use either or both.


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