Accuracy Overview
Monitor system correctness at field level, manual editing effort, and matching accuracy.
Time range
Customer
Showing data from Oct 30 – Nov 29, 2025
Overall field accuracy
Percentage of AI-extracted field values that match ground truth.
94.1%
across all fields +5.9 pp
vs period startPerfect order rate
Percentage of orders that required zero manual corrections.
80.2%
orders with no edits +4.2 pp
vs previous 30 daysAvg. fields changed
Average number of fields users edit per order after AI extraction.
1.4
per order -0.3
Lower is betterClassification accuracy
Accuracy of is_order detection and category classification.
96.8%
is_order & category +1.2 pp
vs previous 30 daysCustomer match accuracy
Percentage of messages with correctly identified customer.
89.2%
correct customer identified +6.7 pp
vs period startField accuracy over time
Is field-level accuracy improving or degrading? Track overall and by field group.
Overall
Address Fields
Date Fields
Quantity Fields
Order editing effort over time
How does the distribution of Perfect / Minor / Major orders evolve?
Perfect
Minor edits
Major edits
Customer match accuracy
Is customer matching becoming more reliable over time?
Match Accuracy
Initial Match Rate