Too Many False Positives

How To

Your dashboard is flooded with irrelevant events. Your keywords technically match, but the matches aren’t actually about what you care about. This is the most common problem with social monitoring, and there are three main tools for fixing it: compound keywords, Lenses, and ignored users.

Prerequisites

You need at least one active keyword with existing events to review.

Diagnosing the problem

First, figure out where the noise is coming from:

1. Go to Keywords, open the noisy keyword, and scan through recent matches on its Events page
2. Look for patterns in the irrelevant events:
   • Are they from the same user? → Use ignored users
   • Do they contain a specific word that’s causing the match? → Use compound keywords with NOT
   • Are they about a different topic that shares your keyword? → Use Lenses
3. Filter events by platform to see if one platform is noisier than others

Solution 1: Compound keywords

Best for: Keywords that match in multiple unrelated contexts.

Using NOT to exclude terms

If your keyword matches posts about an unrelated topic, add a NOT condition:

• Problem: “acme” matches both your company and Acme cartoons
• Fix: acme NOT roadrunner NOT cartoon

Using AND to require context

If your keyword is too broad alone, require additional context:

• Problem: “apple” matches the fruit, the company, recipes, and more
• Fix: apple AND iphone or apple AND macbook

How to apply

1. Go to Keywords and click the keyword to edit
2. Modify the keyword text to include AND/NOT logic
3. Save
4. Watch for a few hours to see if false positives decrease

Make one change at a time so you know what helped. Don’t add too many NOT conditions: each one risks filtering out relevant posts. And check that your compound keyword still matches the posts you DO want.

Solution 2: Lenses

Best for: False positives that can’t be eliminated with keyword logic alone, or that require understanding the post’s meaning.

Lenses use AI to classify each event based on a prompt you write. They can understand context that keyword matching can’t.

Setting up a Lens for noise reduction

1. Go to Keywords and open the noisy keyword
2. On the keyword’s Events page, use the Lenses card to add a Lens
3. Write a prompt that describes what makes a mention relevant:

Example prompt:

“This post mentions our brand ‘Acme’. Is the author discussing Acme as a software tool or SaaS product? Keep posts that explicitly reference Acme as a product, service, or company. Exclude posts where ‘acme’ is used as a common word, such as ‘the acme of perfection’.”

4. Save and run the Lens
5. Review Lens results after a few hours and adjust the prompt as needed

When Lenses are better than compound keywords

• The false positive pattern is semantic (about meaning, not just word presence)
• Different false positives have different causes that can’t be covered by one NOT condition
• You want to keep some matches that would be filtered by a NOT condition
• The noise pattern changes over time and keyword logic can’t adapt

Solution 3: Ignored users

Best for: Specific accounts that consistently post irrelevant content matching your keywords.

When to ignore a user

• A bot that mirrors content and triggers many matches per day
• An account that uses your brand name in their username
• A spammer posting low-quality content
• A user who posts many times daily about a topic that includes your keyword but isn’t about your brand

How to ignore a user

1. Go to Keywords and edit the keyword
2. Open the Ignored Users section
3. Click Add Ignored User
4. Enter the platform and username
5. Save

Review the user’s events before ignoring them. Ignored users are configured on each keyword with a platform and username. You can remove a user from the keyword later if needed.

Solution 4: Per-platform adjustment

Best for: Keywords that are noisy on one platform but fine on others.

1. Go to Keywords, open the keyword, and use the Events page platform filter
2. Identify which platform generates the most false positives
3. Go to Keywords and click the keyword to edit
4. Uncheck the noisy platform
5. Save

Combining solutions

For best results, layer these approaches:

1. First, use compound keywords to eliminate the most obvious false positive patterns
2. Then, ignore users who are consistently noisy
3. Finally, set up a Lens to filter remaining false positives by meaning/intent
4. Adjust platform selection if one platform is much noisier

Monitoring your progress

After applying fixes, check your results:

1. Go to Keywords, open the keyword, and review the last 24 hours on its Events page
2. Compare the ratio of relevant to irrelevant events
3. If still too many false positives, tighten one of your filters
4. If now too few relevant events, loosen one of your filters

Where to go next

• Simple vs compound keywords: learn compound keyword syntax
• Writing effective Lens prompts: create AI filters
• Ignoring users: filter specific accounts