> For the complete documentation index, see [llms.txt](https://docs.zingg.ai/zingg/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.zingg.ai/zingg/tuning/improve-accuracy.md).

# Improve Accuracy

Approaches for tuning your Zingg model when match results are not accurate enough.

When your Zingg match results contain false positives (records incorrectly grouped together) or false negatives (matching records not found), there are several ways to improve accuracy\
without starting over.

Start with the approaches in this order; each is less effort than the next.

### Step 1: Add more training data

The most reliable way to improve accuracy is to label more pairs. Return to the `label` phase, run `findTrainingData` again, and focus on labeling the types of pairs that are causing problems.

* If you have too many false positives (records incorrectly merged), add more non-match labels for pairs that look similar but are different entities.
* If you have too many false negatives (matches being missed), add more match labels for pairs that look different but are the same entity.

{% hint style="success" icon="right-long" %}
**Read more**:

* Label training pairs - [Label training pairs](/zingg/running-zingg/create-training-data/label-training-pairs.md)
* Create training data - [Create training data](/zingg/running-zingg/create-training-data.md)
  {% endhint %}

### Step 2: Check your field match types

Review your `FieldDefinition` match types. Common mismatches:

* Using `FUZZY` for a field that should be `EXACT` - for example, SSN or national ID numbers should always use EXACT. Using FUZZY allows near matches that produce false positives.
* Using `EXACT` for a field that has variations - for example, address fields almost always have variations and should use FUZZY.
* Using `DONT_USE` for a field that actually carries strong identity signal - adding it back with the right match type can improve recall

{% hint style="success" icon="right-long" %}
**Read more**: Match types reference - [Match types](/zingg/zingg-concepts/zingg-configuration/field-definition/match-types.md)
{% endhint %}

### Step 3: Remove stopwords

Common words like Mr, St, Street, Pvt add noise to fuzzy matching on address and name fields. Running the stopwords removal step reduces this noise and can improve both precision and recall.

{% hint style="success" icon="right-long" %}
**Read more**: Remove stopwords - [Remove stopwords](/zingg/tuning/improve-accuracy/remove-stopwords-optional.md)
{% endhint %}

### Step 4: Verify blocking

If Zingg is missing matches that you know exist (false negatives), the blocking model may be filtering them out before they even reach the similarity model. Run Verify Blocking to inspect which pairs are being blocked and identify coverage gaps.

{% hint style="success" icon="right-long" %}
**Read more**: Verify blocking - [Verify blocking](/zingg/running-zingg/create-training-data/verify-blocking.md)
{% endhint %}

### Step 5: Custom blocking and similarity

For advanced use cases, you can define your own blocking and similarity functions to replace\
or extend Zingg's built-in approaches. See [Custom Blocking and Similarity](broken://pages/DWdyf7az3MmhJaVca3k7) for advanced tuning approaches.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.zingg.ai/zingg/tuning/improve-accuracy.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.