How to Extract Data from Screenshots & Images (Tables, Receipts, Invoices)

Data trapped in images is a surprisingly common productivity problem. A colleague screenshots a spreadsheet and sends it over WhatsApp. A client emails a scanned invoice. You photograph a table from a textbook or a whiteboard. In all these cases, the information is visually right in front of you — but completely inaccessible to any spreadsheet, database, or analysis tool until it's been converted to editable text.

Manual retyping is slow, error-prone, and completely unnecessary with the right approach. This guide explains the methods available, when to use each, and how to get the most accurate results.

The core challenge: images have no structure

To a computer, an image is just a grid of coloured pixels. There's no inherent meaning in those pixels — no indication that a certain row of text represents column headers, that a number is a price rather than a date, or that two values belong in the same table row. Extracting structured data from an image requires two separate steps:

1. Character recognition — identifying individual letters and numbers from the pixel grid (this is what OCR does)
2. Structure inference — interpreting the spatial arrangement of recognised text to reconstruct meaningful structure: rows, columns, cells, hierarchies

This is why simply copying text from a screenshot (when that's even possible) often gives you a single unbroken string of text rather than neatly organised data — the copy operation only captures characters, not structure.

What types of images can be converted?

Document type	Common sources	Ideal output
Data tables	Excel screenshots, website tables, database exports, whiteboards	CSV or Excel with proper column headers
Receipts	Retail receipts, restaurant bills, ATM slips, petrol receipts	Structured list: item, quantity, price, total
Invoices	Freelancer invoices, utility bills, purchase orders, tax receipts	Vendor, date, line items, subtotal, tax, total
Marks/grade sheets	School report cards, university transcripts, exam result sheets	Student name, subject, marks, grade, percentage
Timetables	Class schedules, work rosters, bus/train timetables	Day × time grid or structured list
Handwritten notes	Bullet lists, to-do lists, handwritten tables	Transcribed and structured text
Business cards	Physical cards, photographed cards	Name, title, company, phone, email, address

The four output formats and when to use each

CSV (Comma-Separated Values)

CSV is the universal data format — readable by Excel, Google Sheets, every major database, and virtually every data analysis tool. Use CSV when you want to import the extracted data into a spreadsheet or analysis workflow. It's the most portable option.

Excel (via CSV download and open)

There's no separate "Excel format" in browser-based tools — you download the CSV and open it in Excel, which recognises the comma-separated structure automatically. Double-click the .csv file in File Explorer/Finder, or in Excel go to File → Open and select it. Excel will correctly display the data in rows and columns.

JSON

JSON is the developer's choice — a structured text format that represents data as key-value pairs and arrays. Use JSON when you're integrating extracted data into an application, API, or database, or when you need to process the data with code (Python, JavaScript, etc.).

Plain text

Plain text output formats the extracted data as a monospaced ASCII table — rows and columns delimited by pipes and dashes. Use this when you need to paste the result into a document, email, or messaging platform where preserving the visual table structure matters.

How to get the best results: image quality tips

The single biggest factor affecting extraction accuracy is image quality. Here's a ranked list of practices for best results:

Tier 1 — always do these

• Use a screenshot rather than a photo of a screen. Screenshots are pixel-perfect; photos of screens suffer from moiré patterns, parallax distortion, glare, and angle issues. If the data is on a computer screen, always take a screenshot.
• Crop to the relevant data only. Remove navigation bars, headers, surrounding content, and whitespace. A tightly cropped image helps the recognition engine focus on the actual data without having to filter out noise.
• Use the highest available resolution. Higher resolution means more pixel information per character, which translates to higher character recognition accuracy.

Tier 2 — important for physical documents

• Ensure even, bright lighting. Shadows across text are one of the primary causes of misread characters. Use natural light or ensure even artificial lighting with no shadows.
• Keep the camera parallel to the page. Photographing at an angle distorts the apparent width of characters, making them harder to recognise and harder to cluster into columns correctly.
• Avoid reflective surfaces. Glossy paper reflects light in ways that can wash out individual characters — use matte paper whenever possible, or shoot at a slight angle to avoid direct reflection.

Tier 3 — helpful for handwritten content

• Print clearly in block capitals. Cursive and stylised handwriting is significantly harder for optical recognition than printed block letters.
• Use dark ink on white paper. High contrast between text and background is the most important factor for handwriting recognition specifically.
• One idea per line. Dense, multi-column handwritten notes are much harder to parse than clearly spaced single-column writing.

Step-by-step: extracting a table from a screenshot

1. Take the screenshot. On Windows: Win+Shift+S for a selection screenshot. On Mac: Cmd+Shift+4. On iOS: Side button + Volume Up. On Android: Power + Volume Down.
2. Crop tightly. Use your image editor or the built-in crop tool (most phones have this in the Photos app) to remove everything except the table data.
3. Select document type. For a data table, select "Table/Spreadsheet" mode — this tells the extraction engine to prioritise column structure over free-form text.
4. Upload and extract. The tool processes the image and displays a table preview.
5. Review the preview. Always verify the extracted data against the original image. Check column headers, check that multi-digit numbers aren't split or merged, and look for any characters that are commonly confused (0 vs O, 1 vs l vs I, 5 vs S).
6. Correct any errors in the preview. Most extraction tools allow you to edit the result directly before downloading.
7. Download in your preferred format. Click Download CSV (or JSON, or plain text) to save the structured data.

Common mistakes and how to fix them

Problem	Likely cause	Fix
Numbers are merged together	Columns too close, low resolution	Use a higher-res screenshot; ensure columns have visible separation
Rows run together	Insufficient line spacing in original	Zoom in on the original before screenshotting to increase visual line height
Characters misread (0 vs O, etc.)	Low contrast, compression artifacts	Use PNG format for screenshots instead of JPEG; increase contrast
Extra blank columns appear	Shadows or ruled lines misidentified as separators	Remove shadows; crop more tightly
Headers not detected	Headers visually similar to data rows	Select "Table" document type manually
Multi-line cells split into separate rows	Line breaks inside cells treated as row separators	Review and merge rows manually in the output

Privacy and security considerations

When using image-to-data tools, consider what's in your image before uploading it. Receipts contain purchase history; invoices contain vendor and pricing information; marks sheets contain personal academic records.

For browser-based tools that use client-side processing, your image never leaves your device — the extraction happens in your browser, and nothing is sent to any server. This is the most private approach and is suitable for sensitive documents. Always check a tool's privacy policy before uploading documents containing personal information, financial data, or confidential business information to cloud-based services.

Use cases across different industries

• Finance and accounting — extracting line items from photographed receipts for expense reporting; digitising paper invoices into accounting software
• Education — converting marks sheets and grade tables to spreadsheets for analysis; digitising handwritten assignment records
• Logistics and procurement — extracting delivery note data, purchase order line items, and stock counts from photographed forms
• Research — extracting data tables from published papers (screenshots of paywalled content) for independent analysis
• Small business — digitising supplier invoices, stocktake sheets, and handwritten order forms without dedicated scanning software
• Personal use — tracking personal expenses by photographing receipts; converting a screenshot of a weekly schedule into a digital calendar

The fundamental skill of turning image data into usable structured data is increasingly valuable as more information is shared as screenshots and photographs rather than as editable documents. With the right tool and good source image quality, what used to take an hour of manual retyping can now be done in under a minute.