Social listening Guide
Evaluate sentiment analysis before acting on its labels
Test sentiment labels against human-reviewed comments in your brand's languages. Use an annotation rubric, disagreement log and worked confusion matrix.
Blog Tag
Data quality
11 articles are tagged Data quality: 2 guides and 9 reference notes. Start with "Evaluate sentiment analysis before acting on its labels".
11 articles
Start here
All articles
- Email finderGuide What an email finder can and cannot verify A found email does not prove it reaches the creator. Separate source, identity, mailbox checks and consent before treating a contact as ready for outreach.
- Campaign measurementReference Build an early campaign warning dashboard Track overdue creator posts, blocked approvals, budget exposure and missing evidence with a dashboard specification and worked hypothetical warning thresholds.
- Creator discoveryReference Compare creator databases with a fixed test brief Evaluate creator discovery software with the same known profiles, search filters and export tasks. Use a trial scorecard to separate evidence from vendor claims.
- Social listeningReference Compare social listening tools with a known query set Evaluate social listening tools with known relevant and irrelevant examples. Compare retrieval, delays, languages and exports without assuming complete coverage.
- Email finderReference Deduplicate creator contacts across several profiles Use a deduplication worksheet to connect creator profiles, keep shared agency inboxes separate from people, and avoid duplicate campaign outreach.
- Social listeningReference Explain listening coverage gaps to stakeholders Explain why visible posts can be missing from listening reports. Use a coverage statement and investigation form to check access, timing and query limits.
- Influencer seedingReference Record gifted mentions without inflating posting rates Calculate a product seeding posting rate with a fixed delivery cohort, equal observation windows and visible missing-data counts. Includes a worked example.
- Social listeningReference Remove duplicate posts from a listening sample Separate duplicate rows, reposts, quoted commentary and repeat authors with a worked decision table that preserves useful discussion in a listening sample.
- Campaign measurementReference Report missing campaign data without filling it with guesses Use a report legend and completeness table to separate zeros from missing creator metrics, show coverage, and keep partial campaign totals clearly labelled.
- Social listeningReference Write brand listening queries that exclude irrelevant mentions Build brand listening queries for common-word names. Use a three-brand worksheet to test context, spellings and exclusions without hiding missed mentions.
