Confidence System

Every AI suggestion comes with a confidence score from 0 to 100%. Unlike many tools that inflate scores to look impressive, Shoptera calibrates confidence honestly — if a score says 90%, the agent is wrong roughly 10% of the time.

How confidence is calculated

Each agent type has its own calibration method, tuned for accuracy:

• Rules-based agents (Condition, Bundle): Higher confidence (85-95%) — they use deterministic patterns with known accuracy
• AI-generated agents (Title, Description, Category): Variable confidence (55-90%) — depends on how much data is available to work with
• Lookup agents (GTIN): High confidence when an exact match is found in product databases, lower when the match is inferred

Using confidence effectively

Start with auto-approve at 95% or higher — only near-certain improvements go through automatically. Review lower-confidence suggestions manually in Activity. As you build trust in specific agents for your product catalog, gradually lower the threshold to save more review time.