> For the complete documentation index, see [llms.txt](https://hotro.metric.vn/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://hotro.metric.vn/en/market-analysis/summary-statistics-2.md).

# 1.6. AI Filter Optimization

In e-commerce market research, setting up manual filters typically drains 10 to 15 minutes of an analyst's time, yet still leaves data vulnerable to critical blind spots due to a lack of comprehensive keywords.

More critically, inaccurate filters distort market size and market share reports, triggering severe risks in strategic planning and new product development. To solve this dilemma, Metric releases feature AI Filter Optimization - a solution that delivers tailored filter recommendations in under 60 seconds.

Based on your initial input, Metric’s AI model automatically executes an automated 5-step workflow:\
Step 1: Discovering synonyms, abbreviations, and English equivalents.\
Step 2: Predicting product categories based on the keywords extracted from Step 1.\
Step 3: Analyze actual products to suggest keywords, exclude keywords\
Step 4: Re-evaluating and refining exclude keyword suggestions.\
Step 5: Finalizing the exact category mapping based on keywords and exclude keywords, targeting the highest-revenue segment.

{% hint style="info" %}
Note: This feature yields optimal results when searching for specific product group names rather than generic macro-category terms.
{% endhint %}

The user interface features an intuitive, side-by-side comparison layout designed specifically for advanced data validation.

The left panel displays your Current Filters, while the right panel showcases the AI-Optimized Filters. You can effortlessly evaluate the variance in keyword volume and examine the exclude keyword, add keywords

<figure><img src="/files/xyIIszlzGPR2q0xlEUEM" alt=""><figcaption></figcaption></figure>

To guarantee absolute objectivity and data integrity for your reports, Metric integrates a Human-in-the-loop framework. You retain absolute control to fine-tune and override any AI recommendation

To verify accuracy:

Method 1: In the product preview section, Metric show products for instant cross-referencing. The system leverages a semantic color-coding system for rapid auditing:

* Red (AI Exclude): Misleading products, not belonging to the product group, have been removed from the report by AI.
* Green (AI Added): Targeted products that were missed by manual filtering have now been added to the AI ​​scan by expanding keywords.

<figure><img src="/files/nAUurQ6fYcfb5lkmhyWs" alt=""><figcaption></figcaption></figure>

<figure><img src="/files/UKTPcOHdFmkP4grE3Id6" alt=""><figcaption></figcaption></figure>

Method 2: To check the exclusion keywords in the filter, you can move a keyword from the exclusion keywords box to the keyword box and add a "+" sign to make it a required keyword, then click <mark style="color:$primary;">\[Update Results]</mark>. The system will immediately display a list of products excluded from the AI ​​filter for you to check the data.

<figure><img src="/files/pvAjW9W4RRDOllNxDKtL" alt=""><figcaption></figcaption></figure>

Once the filter structure perfectly aligns with your criteria, click <mark style="color:$primary;">Apply AI filter</mark>. The entire underlying filter architecture will be seamlessly synchronized with the main outer interface.

To optimize your workflow, the system will hold off on running the final analysis immediately. This grants you the freedom to continue fine-tuning other peripheral constraints such as Price Range, Sales Channels, or Date Ranges before executing the Analyze.