OmniDocHow to transfer scanned financial statements to Excel and check the amounts line by line
After the scanned document is converted into Excel, the amount is often wrong due to OCR recognition errors. This article provides preparations before conversion, conversion steps, and specific methods for checking totals, error-prone characters, and negative symbols line by line.
Key points of this article
- Before conversion, confirm whether the original is a scanned copy and note the amount writing conventions, such as negative number writing, millenths, and units.
- The verification order is to first check the row count and header structure, then recalculate the sum using the SUM formula, and finally sample and compare the original verbatim verbatimly.
- 0 and 6, 1 and 7, 5 and 6, 3 and 8, as well as decimals and commas, are common error positions in OCR and need to be paid special attention.
- Table recognition preserves formatting, which means preserving row-column structure, merging cells and numeric formats, rather than simply squeezing text into a single column.
- The results of machine conversion should be manually verified against the originals before submission or posting.
Finance staff often receive scanned reports, such as printed bank receipts, photographs of paper profit and loss statements, or account balance sheets stamped and sent back by partners. When you want to enter them into Excel for summary, your subconscious is usually to find an image to recognize as an Excel spreadsheet. The conversion itself can be completed in just a few minutes; what really takes time is confirming whether the results are reliable. OCR (Optical Character Recognition) easily misreads numbers, and decimal and minus signs may be lost. If a digit is misaligned, the subsequent sum and reconciliation relationships will all be off.
This article is intended for accountants, cashiers, and financial analysts who need to enter scanned reports into Excel. After reading, you will know: what to prepare before conversion, how to check amounts line by line after conversion, how to fix any issues found, and what exactly to keep in the "Table Recognition and Retention Format" policy.
After image recognition as an Excel spreadsheet, how do you check the amount line by line? In short, a conclusion
First, complete the conversion, then use the original to compare the amount line by line, and use Excel's own sum formula for a second check. After both passes are passed, enter the data into the system.
| Your situation | Recommended recipe | The reason |
|---|---|---|
| The scanned document is clear, the table lines are intact, and the number of lines is not long | After conversion, compare line by line against the original image, focusing on the amount column | Manual comparison is direct, allowing detection of misalignments during recognition |
| Many lines, long amounts listed | First, use the SUM formula to check the subtotal and total, then randomly check key rows | The formula can quickly locate the position where the total does not match |
| The image is skewed, shaded, or the stamp presses the numbers | Edit the page first, then recognize, and return the key numbers to the original for confirmation | The image quality before recognition directly determines the error rate |
| Reports can be merged cells or multi-level headers | First, check the meter header structure, then check the numbers | If the structure is wrong, matching the numbers makes no sense |
| The same report has two versions: before and after the revision | Use the PDF comparison feature to find discrepancies, only checking the row of differences | The workload is small, and it's not easy to miss any changes |
Before you start: Confirm these points first
- Whether the original is a scanned copy. Try to select a piece of text in the PDF. If you cannot select it, or only the entire picture is selected, it means it is a scanned document and you need to do OCR first. PDFs that can select text can usually be converted directly, with cleaner results.
- Whether the file is a PDF. If you have photos or image files, it is recommended to organize them into PDF first and then process them, for example, output them to PDF through the system's print function. After the pages are unified, subsequent OCR and page count checks will be more convenient.
- Whether the number of pages and lines are complete. When scanning, the edges are cropped or the table spans two pages, which will cause rows to be missing. Before converting, count the number of pages and write down approximately how many lines there are on each page of the original, so you can refer to them when checking.
- Written agreement on the amount. Whether you use a negative sign or parentheses for negative numbers, whether the amount has thousandths, and whether the unit is yuan or ten thousand yuan, all these must be written down first. If the agreement is not clear, it is impossible to determine whether a number is correct or not when checking.
Steps to perform OCR and export Excel files in the OmniDoc desktop client
- Open download.omnidoc.top and download the OmniDoc desktop client, which supports Windows, macOS, and Linux.
- Open the scanned PDF in the client. The text on the scanned document is in image form and usually cannot be directly selected after opening.
- Perform OCR on scanned PDFs, turning the text and numbers on the page into optional, searchable text.
- Quickly browse the recognized pages page by page, confirming that the number of pages matches the original, with no missing pages or gaps between pages.
- Export the results as an Excel file. If you still need text archiving, you can also export it separately to Word.
- Open the export file in Excel, check if the columns are aligned and if the numbers are treated as text.
After completion, check each item one by one
It's recommended to check the structure first, then the numbers, and don't reverse-order them.
| Inspection items | How to see it | What should you do if there's a problem? |
|---|---|---|
| Number of lines and pages | Compare the original page count with the number of rows per page, and count the rows in Excel | If a line is missing, it should be returned to the recognition results to supplement and check if it is missing at the page gap |
| The meter header corresponds to the column | Check if the column names match the originals and if the merged cells have been split | Manually adjust column structure; do not directly compare numbers when columns are misaligned |
| Combined with minor calculations | Recalculate using the SUM formula in Excel and compare it with the original sum | If the difference is exactly an error in a certain digit, first return to the original image to check that line |
| Characters prone to errors | Pay special attention to 0 and 6, 1 and 7, 5 and 6, 3 and 8, as well as decimals and commas | Correct each one by comparing with the original picture; do not judge based on impressions |
| Negative numbers and signs | Check whether the parentheses indicate negative numbers or regular parentheses, and check if the minus sign and horizontal line have been misidentified | After unifying the writing method, then sum up to avoid negative offsetting |
| Sampling controls | Randomly draw a few rows from each amount column and compare them verbatim with the original | Once errors are found, the scope of random inspections is expanded until no errors appear in the samples |
When verifying, you can add a column of "Original Amount" in Excel, manually enter the numbers on the original, and then use a formula to check if the two columns are equal. The difference will be obvious at a glance. Machine recognition may be error; important data should be based on the original, and manual verification should be performed before submission or posting.
📊 Core Comparison
There are three common ways to verify amounts, each suitable for different scenarios:
| Method | Advantages | Limitations |
|---|---|---|
| Manually compare the original images | Direct and reliable, capable of detecting and recognizing misalignments | Too many lines are time-consuming, making it easy to get tired and miss reading |
| Excel formula recalculation total | Quickly identify the uneven points of the total | It can only detect summary errors, but cannot detect cases where a line's numbers themselves are misread but do not affect the total |
| Compared to the pre-revision version | Just check the changes made | The premise is that there is a comparable older version |
❓ Frequently Asked Questions (FAQ)
1. What does it mean to keep format for table recognition?
Refers to retaining the column structure, merging cells, numeric formats (decimal, thousands, negative number notation), and header hierarchy after recognition, rather than cramming all content into a single column of text. Results in preserved formats can be directly calculated and filtered, saving a lot of organizing work.
2. Can I use free online tools for image recognition to Excel spreadsheets? How should you choose?
When choosing free online tools, first check file size limits, whether exported results have watermarks, and whether formatting and numbers are preserved. Financial data is sensitive information; when involving client or company internal information, it is important to first confirm how the documents will be stored and handled.
3. If the recognition does not match the total, is it OCR error, or is there a problem with the original itself?
First, check if the sum on the original is consistent, then look at the recognition results. If the original is consistent but the Excel is uneven, it is mostly due to identification errors; If the original document itself is uneven, you must return to the original voucher for confirmation.
4. Can the converted Excel be used directly for reimbursement or tax filing?
It is not recommended to use them directly. Machine conversion errors may occur; before submission, manual verification should be made against the originals, and important figures should be verified by the original voucher.
5. Is it really necessary to check so meticulously with just a few lines of numbers?
When the number of lines is small, manually compare line by line to the original image. The key points of sum, error-prone characters, and negative number symbols still apply, but some formula checks can be omitted.
📝 Summary
Scan the report and convert it to Excel. Recognizing the image as an Excel table is only the first step; only by checking the amount line by line can the data truly be usable. First, check the number of rows and the header structure, then use the sum formula and sample comparison to find errors, and finally complete the review based on the original.
