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

# Preparing Structured Data

> Prepare operational exports so events can be interpreted consistently and data-quality issues are visible early.

## Before export

Define:

* process scope;
* date range;
* case identifier;
* activity/status field;
* timestamp field;
* owner/user field where relevant;
* expected volume.

## Check basic data quality

Look for:

* missing case IDs;
* missing or invalid timestamps;
* duplicate events;
* inconsistent status names;
* timezone ambiguity;
* test records;
* unexplained gaps;
* multiple identifiers for the same case.

## Preserve raw evidence

Keep a copy of the original export. Perform cleaning in a reproducible way and document material transformations.

## Avoid inventing certainty

If timestamps are generated retrospectively or status fields are not used consistently, record that limitation. A process model built from unreliable event data can look precise while being wrong.

## Consultant responsibility

Work with the client data owner to understand field meaning rather than guessing from column names.

<Note>
  Current Iris upload format and validation messages are **\[PRODUCT DETAIL TO CONFIRM]**.
</Note>