GuideSynthetic examples included
Catching up starts with a folder of PDF statements. The risk is rarely one conversion: it is a month that is missing, a month that is in twice, or a file nobody checked. This is a procedure for turning that folder into one CSV you can stand behind.
It works with whatever produces your files — a converter, a bank export, hand entry. The examples are invented, with files to download and formulas to try.
| Statement | Opening | Closing | Gap |
|---|---|---|---|
| January 2026 | 4,320.15 | 5,512.05 | — |
| February 2026 | 5,512.05 | 6,764.75 | 0.00 |
| March 2026 | 6,764.75 | 6,887.90 | 0.00 |
Before converting
Write down which statements the catch-up needs, and three figures from each: the period it covers, its opening balance and its closing balance. They come from the PDF, so no converter has shaped them, and they are all it takes to find a missing month later.
A statement’s period is not always a calendar month. Copy the dates it prints, and note any statement that arrives as two files or runs to a different day each month.
| Statement | Period | Opening | Closing | Items | Gap | Next day |
|---|---|---|---|---|---|---|
| January 2026 | Jan 1 – Jan 31 | 4,320.15 | 5,512.05 | 8 | — | — |
| February 2026 | Feb 1 – Feb 28 | 5,512.05 | 6,764.75 | 8 | 0.00 | Follows |
| March 2026 | Mar 1 – Mar 31 | 6,764.75 | 6,887.90 | 7 | 0.00 | Follows |
With real dates in B and C (convert them first if they sit left-aligned as text), the two checks are one formula each, from row 3 down. A gap of 0.00 means the statement opens where the last one closed; TRUE means its period starts the day after the last ended.
G3 =ROUND(D3-E2,2)This statement’s opening balance less the closing balance above it, rounded to whole cents. It is not a tolerance: one cent still shows.
H3 =B3-C2=1This statement’s first day, less the last day above it, is exactly one.
The procedure
The order matters more than the tools. Checking a file before stacking it means a mistake costs one month’s review, not the whole sheet’s.
A problem in one month stays a problem in one month. Whatever makes the file, a conversion nobody has checked is a draft.
The opening balance plus every amount is the closing balance, and credits and debits each match their printed totals. The verification guide has the full method and formulas. Record the result in the tracker either way.
The same columns in the same order, ISO dates such as 2026-03-18, and one signed amount column with credits positive and debits negative. Reshape the layout only; do not correct values yet.
Add a statement column before stacking. It lets you filter one month, and it is how a row gets traced back to its PDF when someone asks.
For consecutive statements of one account, each should open at the previous closing balance and start the day after the previous one ended. A gap or overlap is a flag to settle against the PDFs, not a verdict: a bank can change its period dates or post an adjustment between statements. This is the one check no single file can run.
Every correction to a value, with the statement, the row and how you settled it. Import only after the join holds and the log is complete.
Joining the months
The same three statements, in three states a folder can really be in. In each, every file passes its own checks. Only the joins between them tell the folders apart.
Every statement is there and each opens where the last one closed.
| Join | Closing | Opening | Gap | Result |
|---|---|---|---|---|
| January to February | 5,512.05 | 5,512.05 | 0.00 | Holds |
| February to March | 6,764.75 | 6,764.75 | 0.00 | Holds |
Every link holds and every period follows the last: nothing is missing and nothing is doubled.
Two files are converted and each passes its own checks. The gap only shows where one closes and the next opens.
| Join | Closing | Opening | Gap | Result |
|---|---|---|---|---|
| January to March | 5,512.05 | 6,764.75 | 1,252.70 | Breaks |
The gap, 1,252.70, is February's net movement, and the periods skip a month. It tells you what the missing statement must add up to, not what is in it.
A re-download saved under a new name. Each copy passes its own checks, and the combined sheet would count February twice.
| Join | Closing | Opening | Gap | Result |
|---|---|---|---|---|
| January to February | 5,512.05 | 5,512.05 | 0.00 | Holds |
| February to February | 6,764.75 | 5,512.05 | −1,252.70 | Breaks |
| February to March | 6,764.75 | 6,764.75 | 0.00 | Holds |
The second copy opens −1,252.70 from where the first closed, and its period overlaps the first. Only that join breaks: the months either side still join up, so checking each file alone cannot see it.
Stacking
The statement column is the whole trick. It keeps January’s rows distinguishable from February’s, and it survives sorting.
Alike is not duplicate. Alder Row pays the same rent on the second of each month, and those rows are right. A real duplicate need not be a whole statement: one row can be repeated inside a single file. A statement counted twice breaks the join in the step above, but a repeated row does not, so two similar rows are a reason to read the PDF.
| Statement | Date | Description | Amount | Balance |
|---|---|---|---|---|
| 2026-01 | 2026-01-27 | Packaging supplies | −97.25 | 5,576.05 |
| 2026-01 | 2026-01-30 | Insurance premium | −64.00 | 5,512.05 |
| 2026-02 | 2026-02-02 | Studio rent | −950.00 | 4,562.05 |
| 2026-02 | 2026-02-04 | Online shop payout | 655.10 | 5,217.15 |
One more check once everything is stacked. Put the first statement's opening balance (4,320.15) in H1 and the last statement's closing balance (6,887.90) in H2: the amounts in column D must account for the whole difference.
H3 =ROUND(SUM(D2:D24),2)=ROUND(H2-H1,2)It reads rows 2 to 24, the 23 rows of alder-row-2026-q1-combined.csv; change the range to your own last row.
The exceptions log
A clean sheet with no history looks the same whether it was careful or lucky. Four columns are enough, and a month with nothing to record still gets a line.
| Statement | Row | What differed | How it was settled |
|---|---|---|---|
| 2026-03 | 2026-03-25 · Shipping labels | Description wrapped onto a second line in the PDF, and the file ran it into the next row. | Read against the PDF; restored the description. No amount changed. |
| 2026-02 | 2026-02-13 · Craft fair - stall fee | Date read day-first by the spreadsheet on opening the file. | Re-imported the column as text, then converted. The row was in order again. |
| 2026-01 | — | No change needed. | Passed its own checks and joined to February. Logged so the next reader knows it was looked at. |
Spreadsheet and converter
The public beta runs for a limited time, on a best-effort basis, and the upload on the homepage says whether it is open right now. It takes a text-based PDF statement of up to 10 MB and 8 pages; scanned statements are refused. The full converter remains a private, invite-only beta. No separate DocShape account is required.
The steps above are spreadsheet work and apply to a CSV from any source. The beta covers the conversion of one statement, and nothing that spans statements.
A year of monthly statements is twelve conversions. Each one gets checked on its own, and the rest of the procedure is yours. If the folder holds scans, the beta will not take them, and the procedure applies unchanged to whatever produces the CSVs. Why a match is not the same as correct explains what the checks leave unsaid.