> ## Documentation Index
> Fetch the complete documentation index at: https://docs.insight.nobly.dk/llms.txt
> Use this file to discover all available pages before exploring further.
# Cascading datasets
> Chain keyword values across levels so choosing a value narrows the options on the next keyword — and manage the dataset's levels, values, and document types.
## What a cascading dataset does
A cascading dataset chains two or more keywords together so that choosing a value on one narrows the options on the next — pick a *Customer* and the *Contract* dropdown only offers that customer's contracts; pick a contract and *Policy* narrows again. Each keyword in the chain is a **level**, and every value on a level is linked to the values it allows on the level below.
Cascading datasets guide data entry and keep combinations valid. Like [autofill keyword sets](/configuration/autofill-keyword-sets), they don't change which keywords a document type has — they constrain the values offered for keywords that are already assigned.
Manage them under **Admin settings → Keywords → Cascading datasets** (open **Admin settings** from your profile menu at the bottom-left of the sidebar). The list shows each dataset's chain of keyword types and the document types it applies to.
## How a dataset is built
| Part | What it is |
| ------------------ | ------------------------------------------------------------------------------------------------------------------- |
| **Levels** | The ordered chain of keyword types, from the first (top) level down. Levels are fixed once the dataset is created. |
| **Values** | The allowed values on each level. |
| **Links** | Which values on the next level each value leads to — a *Customer* value links to that customer's *Contract* values. |
| **Document types** | The document types the dataset applies to. On every other document type the same keywords behave normally. |
## Cascading datasets and document types
A cascading dataset only takes effect on the document types it's assigned to:
* **On an assigned document type**, the chained keywords offer only the dataset's values, narrowed by what's chosen on the level above.
* **On any other document type**, the same keyword types are unaffected — users enter values as usual.
* A keyword type on a given document type can be governed by **at most one** cascading dataset. Nobly Insight blocks assignments that would make two datasets overlap on the same document type and keyword.
You can add and remove document types at any time from the dataset's detail screen. A dataset that isn't assigned to any document type is valid but dormant — it doesn't apply anywhere until you assign one.
## Creating a dataset
1. Open **Admin settings → Keywords → Cascading datasets** and choose **New cascading dataset**.
2. Name the dataset and pick the **keyword types** to chain, in order from the first (top) level to the last. A dataset needs at least two levels.
3. Choose the **document types** it applies to.
4. Open the dataset and add **values** to each level, then **link** each value to the values it should allow on the level below.
The levels — which keyword types are chained, and in what order — can't be changed after creation. To restructure a cascade, create a new dataset and delete the old one.
## Managing values
From the dataset's detail screen you can:
* **Add values** to any level, and **link** a value to the values it allows on the next level. A value on a lower level that no parent links to is flagged — it can never be offered during entry until you link it.
* **Rename a value.** If documents already hold the old value, Nobly Insight offers to update those documents in the background as part of the rename; you can follow the job on the **Mass value editing** screen.
* **Deactivate a value** to stop offering it during entry without deleting it — useful for values that are obsolete but still present on existing documents.
* **Delete a value**, which also removes its links. Documents that already hold the value keep it.
## Deleting a dataset
Deleting a cascading dataset removes the dataset itself — its levels, values, links, and document-type assignments; it requires the **Keyword configuration administration** permission, and you're asked to confirm first.
It does **not** delete the keyword types in the chain, the document types it was assigned to, any documents, or the keyword values already stored on documents. The only effect users see is that the chained keywords on the assigned document types stop being narrowed and behave like ordinary keywords again.
## Where to read next
Make a chained keyword required, read-only, hidden, or unique on the document types the dataset covers.
Define the keyword types a cascading dataset chains together.
Make sure the chained keywords are assigned to the document types the dataset covers.
Fill related keywords automatically from a single look-up value — a complement to cascading narrowing.