How ZeroParse does on 83 real bank statements

Every converter in this category advertises an accuracy number. None of them publishes what it was measured on. This page is our attempt at the opposite: the corpus, the method, the per-file results, and the files we got wrong.

77 of 83

PDFs whose extracted rows add up against the balances the statement itself printed

4

transaction histories with no totals, where every row agrees with the balance printed beside it

2

printed nothing we could check against — reported as unverified, not as correct

0

where the arithmetic disagreed and we said so

What “reconciled” means, and what it does not

A statement usually prints facts about itself: an opening balance, a closing balance, sometimes column totals. Reconciled means we added up the rows we extracted and they match those printed figures. It is arithmetic, not an opinion, and you can redo it by hand on your own file.

It is not a claim that every field is perfect. A description could still be truncated or a date read in the wrong order without breaking the totals. Reconciliation catches the errors that move money — a dropped row, a missed digit, a debit read as a credit — which are the ones that matter most and the ones that are hardest to spot by eye.

Some files are not monthly statements but transaction histories downloaded from internet banking. They print no opening, closing or column totals, only a balance on every row. When every row agrees with that balance we say running balance, not reconciled: it checks every row after the first, but not the first row's amount or anything outside the dates the export covers.

A few files print nothing checkable at all. Those come back unverified. That is an honest “we cannot tell you”, and we would rather say it than show a green tick that means nothing.

Where this corpus came from, and how it is biased

These are 83 real PDF statements and 6 real CSV/Excel exports, collected from our own team's accounts and from family and friends who agreed to share theirs. They span 6 currencies, and 39 of the PDFs are credit-card statements rather than bank accounts.

This is not a random sample and we will not pretend otherwise. It is weighted towards Singapore and India, with statements from the UK, the United States, Australia and Brunei alongside them; it is mostly personal accounts, though a small retail business banks in there too; and it contains no statement from a bank that refused to work — because a file nobody could obtain never entered the set. A vendor's own corpus is the weakest form of evidence there is. It is simply more than anyone else in this category currently shows you, and it grows as real files arrive.

The files themselves are never published, and never will be. They belong to real people. What you see below is derived from them by a generator that reads only counts and verdicts and never reads a filename.

Per-file results: PDF

4,959 transactions across 83 files. 77 were read from the page's own text layer; 6 were scans that needed OCR.

FileCurrencyRead asRowsOut / InResult
F01 cardAUDText3432 / 2Reconciled
F02AUDText325301 / 24Reconciled
F03 cardBNDText111105 / 6Unverified
F04GBPText54 / 1Reconciled
F05GBPText6157 / 4Reconciled
F06GBPText205193 / 12Reconciled
F07GBPText242189 / 53Running balance
F08INRText32 / 1Reconciled
F09INRText163 / 13Reconciled
F10INRText1610 / 6Reconciled
F11 cardINRText1713 / 4Reconciled
F12INRText1914 / 5Running balance
F13INRText2115 / 6Reconciled
F14INRText4830 / 18Reconciled
F15INRText5341 / 12Reconciled
F16INRText5347 / 6Running balance
F17INRText10677 / 29Reconciled
F18—Text66 / 0Running balance
F19—Text126 / 6Reconciled
F20—Text126 / 6Reconciled
F21—Text1311 / 2Reconciled
F22—Text5422 / 32Reconciled
F23—Text7463 / 11Reconciled
F24—Text113106 / 7Reconciled
F25—Text123116 / 7Reconciled
F26 card—OCR1413 / 1Unverified
F27SGDText00 / 0Reconciled
F28 cardSGDText22 / 0Reconciled
F29 cardSGDText54 / 1Reconciled
F30 cardSGDText54 / 1Reconciled
F31 cardSGDText54 / 1Reconciled
F32 cardSGDText65 / 1Reconciled
F33 cardSGDText87 / 1Reconciled
F34 cardSGDText87 / 1Reconciled
F35SGDText95 / 4Reconciled
F36 cardSGDText98 / 1Reconciled
F37 cardSGDText98 / 1Reconciled
F38 cardSGDText105 / 5Reconciled
F39 cardSGDText119 / 2Reconciled
F40 cardSGDText1110 / 1Reconciled
F41 cardSGDText1211 / 1Reconciled
F42SGDText1310 / 3Reconciled
F43 cardSGDText1614 / 2Reconciled
F44SGDText325 / 27Reconciled
F45SGDText3728 / 9Reconciled
F46 cardSGDText3731 / 6Reconciled
F47SGDText3828 / 10Reconciled
F48SGDText3931 / 8Reconciled
F49SGDText4431 / 13Reconciled
F50SGDText4532 / 13Reconciled
F51SGDText4536 / 9Reconciled
F52SGDText4631 / 15Reconciled
F53SGDText4833 / 15Reconciled
F54SGDText5040 / 10Reconciled
F55SGDText5131 / 20Reconciled
F56 cardSGDText5436 / 18Reconciled
F57SGDText5636 / 20Reconciled
F58SGDText5642 / 14Reconciled
F59 cardSGDText5653 / 3Reconciled
F60 cardSGDText5655 / 1Reconciled
F61SGDText5844 / 14Reconciled
F62 cardSGDText6158 / 3Reconciled
F63 cardSGDText6563 / 2Reconciled
F64SGDText7756 / 21Reconciled
F65 cardSGDText8584 / 1Reconciled
F66 cardSGDText8584 / 1Reconciled
F67 cardSGDText109108 / 1Reconciled
F68 cardSGDText109108 / 1Reconciled
F69 cardSGDText118112 / 6Reconciled
F70 cardSGDText121119 / 2Reconciled
F71 cardSGDText122119 / 3Reconciled
F72 cardSGDText124118 / 6Reconciled
F73 cardSGDText124120 / 4Reconciled
F74 cardSGDText127124 / 3Reconciled
F75 cardSGDText128121 / 7Reconciled
F76 cardSGDText135132 / 3Reconciled
F77SGDText24733 / 214Reconciled
F78 cardSGDOCR87 / 1Reconciled
F79SGDOCR4233 / 9Reconciled
F80SGDOCR7170 / 1Reconciled
F81SGDOCR133127 / 6Reconciled
F82 cardUSDText10197 / 4Reconciled
F83 cardUSDOCR2413 / 11Reconciled

File ids are positions in a sorted list, not stable identifiers — they shift when the corpus grows. “card” marks a credit-card statement, where a purchase is exported as a debit so it lands the right way round in QuickBooks and Xero.

Per-file results: CSV and Excel

1,350 of 1,361 source rows mapped into transactions across 6 files. These are parsed entirely in your browser and never uploaded, which is also why this path is weaker at self-checking: it reads column headers, and a header can be missing or meaningless.

FileSource rowsMappedSkippedDate orderResult
C0112120MDYUnverified
C0218180MDYUnverified
C0345450MDYUnverified
C0454486DMYRunning balance
C056116065MDYReconciled
C066216210MDYSuspect

The ones we got wrong

Publishing only the wins would make this page marketing. These are the open problems this corpus currently shows.

  • 2 PDFs can't be checked. They print no totals and no running balance, only transactions, so nothing on the page can confirm the rows. You check them against your own balance in the app.
  • One CSV doesn't match its own figures. Every row maps, but the balance printed on some rows doesn't move by the amount beside them. That usually means rows were lost before the file reached us, in a PDF-to-CSV conversion, and we flag it rather than pass it.

Check us on your own file

You do not have to take this page's word for anything. Drop one of your own statements in — the same reconciliation check runs on it, and tells you which of the results above your file got, before you export anything. CSV, Excel and OFX files never leave your browser at all.

Method

Every file in the corpus is re-run through the real extraction pipeline — not a test double — on every change that touches it. The run records each file's row count, debit/credit split, zero-amount rows and reconciliation verdict, and compares them against a stored baseline, so an unexplained change in what a real file extracts is itself treated as a defect. This page is generated from those baselines, so it cannot drift from what the checks actually measured: if the corpus grows or the numbers move, the page is stale until it is regenerated, and our build checks fail until it is.

Figures on this page were last regenerated from a corpus of 83 PDFs and 6 CSV/Excel files.