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

# Data Overview

> How operational data enters ProDex, gets understood, and becomes durable, auditable knowledge: sources, exploratory analysis, pipelines, reports, and Data Insights.

## Overview

The Data area is where operational data enters ProDex, gets understood, and becomes durable knowledge your team revisits. The Data page has four tabs — **My Data** (the raw inputs), **Reports** (downloadable deliverables), **Insights** (dashboards), and **Pipelines** (the auditable record of how numbers were derived) — and one workflow that ties them together: **Exploratory Analysis**, run in chat with Dexter, which builds trust in a source and populates the dashboards and pipelines.

This page maps those surfaces in the order you touch them, and untangles a few terms that sound like Data concepts but live elsewhere in the product.

## The Surfaces at a Glance

| Surface                                                     | What it is                                                                                                                                                                                                                      | Where it lives                                                  |
| ----------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------- |
| [Data sources](/product/data-sources)                       | The raw inputs: uploaded files and connected integrations. ProDex never modifies them.                                                                                                                                          | The Data page, under the **My Data** tab                        |
| [Exploratory Analysis (EDA)](/product/exploratory-analysis) | When Dexter first encounters a new data source, by default it launches sub-agents, asks you questions about it, and produces an analysis dashboard plus a pipeline documenting its understanding of the source for your review. | Chat, initiated from a source or from the Data page             |
| [Reports](/product/dexter/reports)                          | Downloadable deliverables produced from your data — PDF, Word, PowerPoint, Excel.                                                                                                                                               | The **Reports** tab under Data                                  |
| [Pipelines](/product/templates)                             | A document of a data derivation as a DAG: sources, the transformations that ran (with the code), and the outputs they produced.                                                                                                 | The **Pipelines** tab under Data; opens as a canvas             |
| [Data Insights](/product/insights)                          | Factory-scoped dashboards, the durable home for charts that don't belong to a specific simulation run or experiment.                                                                                                            | The **Insights** tab under Data; charts live inside a dashboard |

## What Each Surface Is

**Data source.** A file you uploaded (CSV, Excel, PDF, CAD, and more) or a connected data integration that pulls data on a recurring basis. Sources are the raw material; ProDex does not modify them. See [Data Sources](/product/data-sources).

**Exploratory Analysis.** A structured workflow, run as a conversation with [Dexter](/product/dexter/chat-and-tasks), that profiles a source, surfaces gaps and ambiguities, resolves them with you, and produces a first dashboard plus the pipeline that documents how it was built. See [Exploratory Analysis](/product/exploratory-analysis).

**Pipeline.** A DAG-shaped record of a derivation: source nodes (bound to an upload, an integration, or any upstream artifact — a model, a BOM, another dashboard, a lookup table), transformation nodes (each carrying the SQL or Python that ran), and output nodes (each bound to the artifact it produced: a dashboard, a simulation model parameter, a BOM quantity, a report, and more). See [Pipelines](/product/templates).

<Note>
  **A pipeline is provenance, not verification.** It shows *how* a value was derived, not whether it matches reality. Trust in the numbers is built during [Exploratory Analysis](/product/exploratory-analysis), where Dexter resolves gaps and ambiguities with you before extracting anything.
</Note>

**Data Insight (dashboard).** A factory-scoped container for related charts, the durable visibility surface for numbers the team revisits between projects. Distinct from run-scoped or experiment-scoped charts, which live with their run or experiment. See [Insights](/product/insights).

**Insight (chart).** A single chart inside a dashboard. The builder covers the usual primitives — bar, line, area, pie, table, scatter, box plot, heatmap, histogram, and gantt — plus higher-level composites like distribution, breakdown, breakdown over time, trend, pareto, and schedule that encode common data-prep patterns. Charts are always children of a dashboard; a chart cannot exist without one.

## Terms That Live Elsewhere

Three terms often assumed to be Data concepts belong to other parts of the product:

| Term          | What it actually means today                                                                                                             | Where it's documented                                       |
| ------------- | ---------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------- |
| **Runs**      | Simulation runs, which belong to a simulation model. Pipelines are not executed by the platform, so there are no "pipeline runs."        | [Runs](/product/runs)                                       |
| **Gates**     | Project sign-off checkpoints: each phase of a Project is closed by a user approval. Gates have no relationship to pipelines or insights. | [Gates](/product/gates)                                     |
| **Templates** | Configuration templates in the BOM configurator. The word is no longer used for pipelines.                                               | [Configuration Templates](/product/configuration-templates) |

## How They Relate

* A **pipeline** reads one or more **data sources** and produces one or more **outputs**. Each output is bound to a specific downstream artifact: a dashboard chart, a simulation model parameter, a BOM value, a report, a constant, or a lookup table.
* A **Data Insight dashboard** contains **insights** (charts) as children. The dashboard exists first; charts are added to it.
* The **EDA workflow** doesn't contain anything. It's the guided process that produces a dashboard and its backing pipeline.
* A pipeline is **keyed by slug and overwritten in place** when the same derivation is re-done against a fresh upload or a corrected interpretation. A genuinely different derivation gets a new slug.

## A Typical End-to-End Journey

1. **Upload.** You drop a file on the Data page, say a quarterly MES cycle-time export. It appears under **My Data**.
2. **Start EDA.** From the source, you ask Dexter to explore it. Dexter profiles the file, asks targeted questions about codes, gaps, and units, and confirms understanding before extracting anything.
3. **First dashboard.** Dexter creates a Data Insight dashboard (e.g. "Cycle times") and populates it with charts: station-level medians, distributions, outliers.
4. **Pipeline lands.** Once the numbers are settled, Dexter documents the cleaning, joins, and aggregations as a pipeline whose outputs bind to the dashboard. From any chart, you can now click into the pipeline and see the file, the filters, and the code behind the number.
5. **Reuse downstream.** The same pipeline can add outputs bound to a [simulation model](/product/simulation-overview) (fitted service-time distributions), a [BOM](/product/bom-overview) (yield rates), or a report. Each output traces back to the same source.
6. **Revisit.** Next quarter, a fresh upload replaces the old one; Dexter re-runs the derivation and saves the pipeline against the same slug. The dashboard updates and the audit trail stays intact.

<Tip>
  **Start with Exploratory Analysis for any source you don't fully understand.** It front-loads the questions that would otherwise surface as wrong numbers later, and every answer gets captured in the pipeline for the next person.
</Tip>

## Where to Go Next

* [Data Sources](/product/data-sources): uploads, integrations, and how sources are documented
* [Exploratory Analysis](/product/exploratory-analysis): the workflow that turns a raw file into trusted numbers
* [Pipelines](/product/templates): reading and maintaining the derivation record
* [Insights](/product/insights): creating and refreshing dashboards
* [Reports](/product/dexter/reports): downloadable deliverables built from your data