What is Minimum Match Score?

Minimum Match Score is a confidence threshold for your AI chatbot. It defines how closely a knowledge base entry must match a customer’s question before being used in a response.

When should I raise or lower the Minimum Match Score?

Raise the score if your chatbot gives irrelevant or incorrect answers. Lower the score if it often says 'I don’t know' to questions it should answer. Adjust by small increments and test for best results.

Which tone should I choose for my chatbot?

Choose a tone that matches your brand and makes your customers feel comfortable. Options include Friendly Helper, Expert Advisor, Luxury Concierge, or No-Nonsense Support.

What's a good starting range for Minimum Match Score?

For most WooCommerce stores, a starting range of 70–80% offers a balance between accuracy and responsiveness. Adjust as needed based on your chatbot’s performance.

Set Helpmate Chatbot Tone And Minimum Match Score

Your AI chatbot is live, and your knowledge base is trained. Now it’s time for the crucial final touches that transform a good bot into a great one: setting its personality and tuning its accuracy. Just like a human sales associate, how your chatbot communicates is as important as what it says.

This guide will show you how to choose the right tone for your brand and master the Minimum Match Score. These two settings work together to ensure your chatbot provides accurate, on-brand responses that build customer trust and drive conversions.

1. Choose a Tone That Matches Your Brand #

Your chatbot’s tone is its personality. It should feel like a natural extension of your brand, creating a consistent experience for shoppers. A mismatched tone can feel jarring and unprofessional. Choose a voice that resonates with your target audience.

Here are four common tones to consider:

The Friendly Helper: Casual, warm, and approachable. Perfect for everyday retail and building a community feel.
- Sample Reply: “Hey there! I can totally help you with that. What are you looking for today?”
The Professional Advisor: Confident, knowledgeable, and precise. Ideal for stores selling technical or specialized products.
- Sample Reply: “Certainly. The specifications for that model include a 2.4 GHz processor and 16GB of RAM.”
The Casual Concierge: Formal, elegant, and attentive. Best for high-end brands selling premium or bespoke goods.
- Sample Reply: “Welcome. It would be my pleasure to assist you in finding the perfect item.”
Technical Support: Direct, efficient, and to the point. Great for B2B stores or sites focused on quick, factual resolutions.
- Sample Reply: “Your order #12345 has shipped. The tracking number is ABC-123.”

Select the tone that best reflects your brand’s values and customer expectations.

2. What Is the Minimum Match Score? #

The Minimum Match Score is a confidence threshold for your chatbot. When a customer asks a question, the AI searches its knowledge base for the most relevant answer. The score you set tells the bot how closely a piece of information must match the query before it can use it in a response.

A high score (e.g., 90%) means the bot will only answer if it’s very confident it has the right information. This reduces the risk of wrong answers but may lead to more “I don’t know” responses.
A low score (e.g., 65%) allows the bot to answer even if the match isn’t perfect. This increases the number of answered questions but raises the risk of providing irrelevant information.

Think of it as adjusting your bot’s level of strictness. Your goal is to find the sweet spot that delivers maximum accuracy with minimum frustration.

3. Recommended Starting Ranges by Store Type #

The ideal Minimum Match Score depends on your inventory and the type of questions you expect. Here are some recommended starting points:

General Retail & Fashion Stores (70–80%): This range provides a great balance for broad queries about products, styles, and store policies. It’s flexible enough to understand variations in how customers ask questions.
Technical or Regulated Goods (80–90%): For stores selling electronics, industrial parts, or items with precise specifications, a higher score is critical. It prevents the bot from making inaccurate claims or providing dangerously wrong information.
Content-Heavy Blogs or Service Sites (65–75%): If your site has a lot of overlapping content, a slightly lower score can help the chatbot connect concepts and find answers in broader articles.

Start with the recommended range for your store type and fine-tune based on performance.

4. When to Raise or Lower Your Score #

After your chatbot has been active for a few days, review its chat history. The data will tell you whether you need to adjust its confidence threshold.

Raise the score if you see:

The bot is providing answers that are completely irrelevant to the question.
It combines information from two different topics into one confusing response.
It confidently gives a wrong answer to a factual question.

Lower the score if you see:

The bot frequently says “I don’t know” to questions you are certain are in the knowledge base.
It can only answer questions that are phrased exactly as they appear in your Q&A data.
It fails to answer questions that are slightly vague or use synonyms.

5. Your 10-Minute Testing Script #

Run this simple test weekly to check your bot’s performance. Keep a simple log of the question, the bot’s answer, and whether it was successful.

The Hyper-Specific Question: Ask for something you know is in your Q&A data (e.g., “What is your return policy for international orders?”).
- Expected Result: A perfect, accurate answer.
The Broad Keyword Question: Use a general term that appears in multiple products or pages (e.g., “Do you have anything in cotton?”).
- Expected Result: A helpful list of suggestions or a clarifying question.
The “Weird Phrasing” Question: Ask a common question using unusual language (e.g., “How long till my stuff gets here?” instead of “What is the shipping time?”).
- Expected Result: The bot should understand the intent and provide the correct shipping information.
The Boundary-Pushing Question: Ask something you know isn’t in the knowledge base (e.g., “What’s the weather like in Paris?”).
- Expected Result: A polite “I don’t know” or an offer to connect with a human agent.

6. Quick Troubleshooting Tips #

False Positives: If the bot answers incorrectly, the score is likely too low. Raise it by 5%.
“I Don’t Know” Loops: If the bot can’t answer simple questions, the score might be too high. Lower it by 5%. Check if the information exists in the knowledge base.
Conflicting Sources: If the bot gives an answer from the wrong document (e.g., quoting a blog post instead of your policy page), it means your knowledge sources are overlapping. Refine your training data to be more distinct.

Take Action Now #

Your chatbot’s tone and accuracy are not “set it and forget it” features. Go into your settings now and adjust the tone and Minimum Match Score. Use the testing script to check performance and commit to a 10-minute weekly review. This small investment will pay massive dividends in customer satisfaction and sales.