# Finding duplicates

A knowledge base built from several questionnaires ends up with the same question
asked more than once, worded differently each time. Duplicates are worth clearing
out: two items answering the same question can give two different answers,
depending on which one gets used.

ResponseHub matches duplicates on meaning rather than wording, so "Do you encrypt
data at rest?" and "Is stored customer data encrypted?" are found as duplicates
even though they share almost no words.

## Run a duplicate check

Duplicates are only found when you ask for them — nothing is scanned in the
background, and the **Potential Duplicates** section on an item stays empty until
a check has been run.

1. Go to **Knowledge Base** and select **Find Duplicates**.

2. Select **Check for Duplicates**.

   The scan covers your whole knowledge base and can take a few minutes on a
   large one. You don't need to stay on the screen.

## Read the results

The results list every item that has potential duplicates, under **Duplicate
Knowledge Base Items**. Select an item to see what it was matched against. If
nothing is found, the screen says "No duplicate items found".

You can also see matches for a single item without running the whole scan again:
open the item and expand **Potential Duplicates**.

**Matches are candidates, not verdicts:** Two items asking a similar question aren't necessarily duplicates — questionnaires
often ask about the same control in ways that genuinely need different answers.
Read both before deciding.

## Resolve a duplicate

Once you have decided which item to keep, delete the other, or edit it so it
answers a distinct question. See
[knowledge base items](/knowledge-base/knowledge-base-items/) for both.

{/* TODO: confirm the intended workflow for resolving a duplicate. There is no merge
    action in the app, so this assumes deleting or editing by hand. */}

## Next steps

- [Knowledge base items](/knowledge-base/knowledge-base-items/)
- [Importing from completed questionnaires](/knowledge-base/importing-from-completed-questionnaires/) — the usual source of duplicates