Feature · Source verification
An audit trail from every converted row back to the bank statement
A spreadsheet of transactions is only as trustworthy as your ability to check it. Every row here keeps a link to the page and line it was read from, so checking a figure means clicking it rather than searching a PDF.
- Every row linked to its page and line
- Digital and scanned statements
- Tells you when a highlight is an estimate
- Flagged rows reviewed first
The short answer
What source verification does
Source verification means every converted transaction can be traced to the exact line on the PDF it was read from. Select a row in the review grid and the statement page opens beside it with that line highlighted, so you can compare the extracted date, description and amount with what the bank printed without leaving the screen.
It covers the half of checking that arithmetic cannot. Reconciliation proves that amounts and balances agree, and says nothing about whether a date or a payee was read correctly. Looking at the source line does. With both, "I think this spreadsheet is right" becomes "I checked the rows that mattered against the statement, and here is where each one came from."
Step by step
How to review a converted statement against the PDF
A long statement does not need every row opened. This order puts your time where errors actually are.
- 1Read the reconciliation panel firstIt lists which checks passed, failed or were skipped. If the running balance broke, the break rows are already highlighted in the grid and are where you start.
- 2Click each flagged rowThe page opens at the right place with the line marked. Compare the printed amount and sign with what was read. A break whose difference is twice the amount is a direction problem; a round multiple is a misplaced digit or decimal.
- 3Look at low-confidence rowsRows the extraction was unsure of are marked "Low extraction confidence — worth a look". On a scan these are often the faint or smudged lines, and they deserve a glance even when the arithmetic passed.
- 4Spot-check dates and descriptionsArithmetic cannot see these, so sample them: the first and last row of each page, any row around a December to January rollover, and the handful of transactions that matter most to the job in hand.
- 5Fix what needs fixingFlip a row whose direction was read backwards, merge a transaction that the statement split across two lines, or exclude a stray row. Your changes are kept apart from the extraction, so the record of what the bank printed is never overwritten.
How it works on a digital statement
A PDF with a text layer records where every piece of text sits on the page. The parser keeps that position as it builds each transaction, so a row carries the coordinates of the line it was read from. Selecting the row scrolls to the page and draws a band over the line.
Where a row came through the model path rather than the geometric parser, the line is found by searching the page’s own text for the row’s words — the same way a person would, but instantly.
How it works on a scan, which is the hard case
A scanned page has no text to measure. The lines are found in the pixels instead: the page is rendered to a canvas, and rows of the image carrying text are dark where the gaps between them are not.
Two things printed on a ruled statement are not text and had to be handled before any of it worked. A horizontal rule reads as a very dark line and welds the rows above and below it into one band. A vertical rule — a table’s right-hand border — sits further right than every amount, so on every line it becomes "the rightmost mark", which quietly destroys the one content signal a scan offers. Both are found the same way: a printed rule runs unbroken, and text does not.
The estimate is then corrected against content. Amounts are right-aligned, so a statement’s transaction lines share a right edge that headings, addresses and totals do not — which is what keeps a row from being highlighted on the payment slip at the foot of the page, however far out the estimate was.
Why this is worth the engineering
It is the difference between a tool you check and a tool you trust. Reconciliation tells you row 47 is wrong; this shows you what row 47 says on the page. Without it, every flagged row costs a trip to a PDF viewer, and after three of those most people stop looking.
On a poor scan the band is an estimate
Why it works
Why a row-to-line link changes how you review
One click from row to line
No second window, no scrolling through twelve pages looking for 14 March. The row opens its page with the line marked.
Scans included
Scanned pages have no text to point at, so the line is located from the page image. The hard case is handled rather than left out.
Honest about precision
Each highlight says whether it is exact, located from the page image, or an estimate, so a confident-looking band is never quietly wrong.
Review in order of risk
Breaks, low-confidence rows and suspected duplicates at page joins are surfaced first, which is where nearly all real errors turn up.
Changes stay reversible
Merged rows are kept rather than deleted, so a merge can be undone. Excluded rows stay visible in the grid and are left out of every export.
Working papers you can defend
When someone asks where a figure came from, you can show them the line on the statement page, not just the spreadsheet cell.
Audit and evidence
Why tracing a figure to its source matters
Anyone who relies on a schedule built from bank statements, whether a reviewer, an auditor, a lender’s underwriter or the other side in a dispute, may ask the same question about any figure on it: where did this come from? The original statement is the authoritative record. The spreadsheet is useful only so long as each of its rows can be tied back to that record quickly.
Without a link, that tie is made by hand: open the PDF, work out which page the date falls on, scan down for the amount, and hope the same amount does not appear twice that week. It takes a minute or two per row, which is why most people stop after three. With the link, it is a click, so the rows that deserve checking actually get checked.
Keep the original PDF alongside the export. The converted file is a working copy for analysis and import; the statement remains the document you point to.
Precision
How each row finds its line, and how precise the highlight is
Not every row can be located with the same certainty, and the review screen does not pretend otherwise. How the line is found depends on how the statement was read.
| How the statement was read | How the line is found | How sure the highlight is |
|---|---|---|
| Digital PDF, read in your browser | Position recorded from the text layer as each row is built | Exact |
| Digital PDF, read by the processing layer (unusual layouts) | The row’s own words searched for in the page’s text | Exact when the text is found |
| Scanned page, reasonable quality | Model’s position estimate corrected against the text lines detected in the page image | Located from the page image |
| Scanned page, faint, skewed or annotated | Model’s position estimate only | Labelled as an estimate |
Speed
Reviewing a 300-row statement in a few minutes
A long statement that reconciles cleanly does not need every row opened. The arithmetic has already vouched for the amounts; what is left to check is concentrated in predictable places.
- Every break row, and the row directly above it. A misread balance shows up as two adjacent breaks, and the cause is on the first.
- Rows marked as low confidence, especially on scanned pages.
- The first and last rows on each page. Page joins are where rows are dropped, duplicated, or confused with a carried-forward balance.
- Rows flagged as possible duplicates across a page break. Decide each one; nothing is removed for you.
- Transactions the job turns on: the large deposit a lender asked about, the transfers in a disclosure schedule, the payments to one counterparty in a forensic review.
- Dates near a year boundary on statements that print day and month without a year, where the year is taken from the statement period.
Review edits
What you can change, and what stays as the bank printed it
Review is for correcting how a row was read, not for rewriting the statement. You can flip a row whose debit and credit were read the wrong way round, merge a transaction the bank printed across two lines (a card payment and its foreign exchange fee, say) into one row, exclude a row that should not be there, and label rows with a category.
None of it overwrites the extraction. Changes are stored as a separate layer on top of what was read, the grid shows how many rows you have edited, and merges can be undone because the absorbed rows are kept. The reconciliation result is also left alone: it describes the extraction against what the bank printed, so deliberately excluding a row does not turn a balanced statement red.
Third-party statements
When the statements come from someone else
Accountants, lenders, family-law teams and forensic accountants often work from statements they did not download themselves: emailed PDFs, scans of printouts, photocopies of photocopies. Those are the documents where a misread digit is most likely and where being able to show your working matters most.
Source linking does not authenticate a statement and does not detect alteration. What it gives you is a clear, fast path from every number you rely on to the line it was taken from, so a reviewer can follow your schedule back to the document as easily as you built it.
Side by side
Checking a converted statement, with and without source links
| Spreadsheet only | Spreadsheet and a separate PDF viewer | Row linked to its source line | |
|---|---|---|---|
| Find the line behind a flagged row | Not possible | Search by date and amount, page by page | One click |
| Check a date or description | Not possible | Possible, slowly | Side by side |
| Scanned pages | Not possible | Possible, by eye | Line located from the page image |
| Know which rows to check first | No | No | Breaks, low confidence and duplicates flagged |
| Record of what you changed | Overwritten in the cells | Overwritten in the cells | Kept separately; merges reversible |
Who uses it
Who uses source verification
Bookkeepers signing off a conversion
Check the flagged rows and a sample of dates against the statement before importing, so the month-end reconciliation in your software has nothing to explain.
Accountants building review files
Tie a client’s transactions to their statements quickly, and show a reviewer exactly where any figure came from.
Family-law and disclosure work
Schedules of transfers or withdrawals built from the other party’s statements, with every entry traceable to its page.
Forensic accounting
Tracing payments to a counterparty across years of statements, where each row in the analysis needs to point back at its source.
FAQ
Every row points back at the line it came from: questions
How do I check that a PDF to Excel conversion is accurate?
Check the arithmetic first: here every conversion is reconciled against the running balance and printed totals, and breaks are flagged. Then check what arithmetic cannot see, dates and descriptions, by clicking rows and comparing them with the highlighted source line.
Can I see where each transaction came from in the PDF?
Yes. Every row carries the page and position it was read from. Select it in the review grid and the page opens with the line highlighted.
Does it work on scanned statements?
Yes. Scanned pages have no text layer, so the line is located from the page image. On a clean scan that is close; on a faint or skewed page the highlight is labelled as an estimate.
What if the highlight is on the wrong line?
On scans it can be a line or two out, and the interface says when a position is an estimate. The page itself is still the right one, and the row’s date and amount let you find the line by eye.
Do I need to check every row?
Usually not. If the statement reconciles row by row, the amounts are consistent. Check the flagged rows, low-confidence rows, page joins and a sample of dates and descriptions.
What can I change during review?
You can flip a row’s debit or credit direction, merge rows the bank printed as one transaction across two lines, exclude rows, and apply a category to selected rows. Changes are stored separately from what was read, and merges can be undone.
Is a converted spreadsheet acceptable as evidence?
That is a question for whoever will rely on it; we do not give legal advice. The original statement remains the primary document. What the source link gives you is a quick way to show where each figure in your schedule came from.
Is the PDF kept so I can check it later?
That is a setting you control, from keeping nothing to keeping files for up to a year, with expired data deleted daily. The security page sets out what is stored and for how long.
Keep going
Related guides and converters
Related features
Work that depends on traceable rows
Related
Other features
See the line behind every row
Create an account, convert a statement and click any transaction to see where it came from. Plans start at $15 a month for 300 pages.