# OCR Accuracy for Business Cards: Field-Level Test | EximAgent

*Source: https://eximagent.ai/blog/business-card-scanning-workflow-benchmark · Published: September 21, 2026 · Category: Technology*

> How to measure OCR accuracy on business cards at character, field and record level, with a repeatable test protocol and an example scorecard you can copy.

# OCR Accuracy: A Business Card Field-Level Test and Example Scorecard

“OCR accuracy” is too broad to guide a business-card workflow. A useful test separates character, field and record accuracy, then shows which fields needed review before export.

## Three useful levels of accuracy

Character accuracy asks whether individual characters were transcribed correctly. Field accuracy asks whether a complete field, such as an email address or phone number, is usable. Record accuracy asks whether the contact is ready for handoff after the required fields are checked.

A record can have high character accuracy and still fail a field if one digit breaks a phone number. Define the rule before you test.

## Test rules

Use a fixed set of cards that represents the layouts your team actually receives. Keep the reference transcription hidden from the person reviewing OCR. Define required fields, what counts as a correction and what makes a record ready. Do not turn an illustrative score into a product claim.

## Eight-step protocol

1. Select a representative sample and create a reference transcription.

2. Scan every card with the same capture conditions.

3. Export the extracted fields without manual rewriting.

4. Compare each field with the reference.

5. Record character errors and field failures separately.

6. Mark whether the record needs a second review.

7. Calculate the agreed metrics.

8. Save the scorecard with the sample date, device and rules.

## Illustrative scorecard

This fictional 40-card example shows the format, not EximAgent performance:

Field | Fields tested | Fields correct | Field accuracy | Notes

Full name | 40 | 37 | 92.5% | stylized fonts needed review

Company | 40 | 38 | 95.0% | abbreviations varied

Email | 40 | 39 | 97.5% | one punctuation correction

Phone | 40 | 35 | 87.5% | country codes required review

Website | 40 | 36 | 90.0% | small print caused misses

Ready records | 40 | 33 | 82.5% | all required fields usable after rules

## How to read the result

A low field score tells you where a review rule is needed. A low ready-record score means the handoff still depends on manual work. Compare the same fields over time and keep the test conditions stable so changes are meaningful.

## Fit the scanner to the workflow

If phone digits or email addresses often need correction, make those fields mandatory in review. Keep event and conversation notes with the contact. EximAgent Business Card Scanner supports editable fields and export; see [/business-card-scanner](/business-card-scanner).

For a workflow comparison, see [/blog/ocr-vs-manual-data-entry](/blog/ocr-vs-manual-data-entry).

The purpose of a benchmark is a repeatable decision rule: know which records are ready, which need review and what to improve next.