Optical character recognition (OCR) is the technology that converts an image of text, such as a scanned lease or a photographed rent roll, into machine-readable characters a computer can search and process. It reads pixels and returns text. In commercial real estate, OCR is the first step that turns a document scan into usable text.
How Does Optical Character Recognition Work?
Optical character recognition works by moving an image through a fixed sequence of stages: capture, preprocessing, text detection, segmentation, recognition, and post-processing. Preprocessing cleans the image with deskewing, noise removal, and contrast enhancement. Segmentation isolates lines, words, and individual characters, and recognition matches each character shape to a known symbol.
Recognition itself rests on feature extraction: OCR converts each segmented character into a vector of numerical measurements describing its edges, curves, and stroke patterns, then classifies that vector. The term was coined by IBM in 1959, and Ray Kurzweil commercialized omni-font OCR in the 1970s, which lifted the technology past reading one trained typeface at a time to reading almost any printed font.
Stage | What it does |
Preprocessing | Deskews, denoises, and enhances contrast; can lift accuracy 3 to 8% on degraded scans |
Segmentation | Separates lines, words, and characters |
Feature extraction | Converts each character into a numerical vector of its shape |
Recognition | Classifies the vector into a character |
Post-processing | Applies dictionaries and context to correct likely errors |
Why Optical Character Recognition Matters
Optical character recognition matters because most commercial real estate documents arrive as images, not data. Leases, offering memorandums, and rent rolls are scanned, faxed, or photographed, and none of their content is usable until OCR converts it to text. OCR is the gate: nothing downstream, no extraction and no analysis, runs on a document a machine cannot read.
Its reliability depends heavily on input quality and text type. OCR on clean printed text reaches 98 to 99% character accuracy, an industry standard, per 2025 benchmark reporting. Cursive or messy handwriting is far harder, with even strong systems landing around 60 to 85% character accuracy. A single point of character error is not trivial at document scale: at 99% accuracy, a 1,000-character page still carries about 10 wrong characters, which is why OCR output is a starting layer, not a finished record.
Example
Optical character recognition is easiest to grade with a worked number. Character error rate (CER) is the standard measure: the count of character insertions, deletions, and substitutions divided by the total characters in the verified ground truth.
Input | Value |
Total characters on the page (ground truth) | 1,000 |
Characters OCR read incorrectly | 8 |
Character error rate (8 / 1,000) | 0.8% |
Character accuracy (1 - CER) | 99.2% |
A clean scan of a lease page returns 8 misread characters out of 1,000, a 0.8% CER and 99.2% character accuracy. That reads as excellent, yet those 8 characters could sit inside a single dollar figure or date. This is why OCR accuracy and field-level accuracy differ: a field is correct only if every character in it matches, so a handful of scattered character errors can still corrupt a specific value that a downstream extraction step must then flag.
Variations and Edge Cases
Optical character recognition spans several related technologies that handle different inputs. The variants below change what kind of text the system can read and how reliably.
Variant | Handles |
Printed-text OCR | Typed and printed characters; 98 to 99% accuracy on clean input |
ICR (intelligent character recognition) | Handwriting; lower and more variable accuracy |
OMR (optical mark recognition) | Checkboxes and filled bubbles, not characters |
Zonal OCR | Reads only defined regions of a known layout |
Handwriting on forms | Neat print in form boxes reaches roughly 90 to 95% |
Optical Character Recognition vs Intelligent Document Processing
Optical character recognition is often confused with intelligent document processing, but OCR is one component of IDP, not a synonym for it. OCR converts an image into raw text and stops there. Intelligent document processing wraps OCR inside a wider pipeline that also classifies the document, extracts specific fields, validates them, and routes the structured result downstream.
The distinction matters in practice: OCR can read every word on a rent roll yet does not know which number is base rent and which is the security deposit. IDP adds the layer that assigns meaning. OCR gives you the text; IDP gives you the data.
Frequently Asked Questions
What does OCR stand for?OCR stands for optical character recognition. It is the technology that converts an image of text, such as a scanned or photographed document, into machine-readable characters a computer can search, edit, and process.
How accurate is OCR on commercial real estate documents?On clean printed text, OCR reaches roughly 98 to 99% character accuracy, per 2025 benchmark reporting. Accuracy drops on low-quality scans and on handwriting, where even strong systems reach only about 60 to 85% on cursive, so degraded leases and handwritten notes carry more error.
Is OCR the same as data extraction?No. OCR converts an image into raw text but does not identify what each value means. Data extraction, part of intelligent document processing, then locates and labels specific fields like base rent or lease commencement from that text.
Related Terms
Intelligent Document Processing
Confidence Score
Human-in-the-Loop
Rent Roll
Offering Memorandum