Not the job you came for? Tell me what you need
Every column, one at a time: what it holds, how much of it is missing, and what's odd about it. Plus the totals, where a total means anything.
A sales ledger, read by the same profiler your file goes through — not a recording. Four columns that all look like numbers, and only one of them is a quantity.
Invoice,Postcode,Amount,Balance INV-1001,0800,240.00,240.00 INV-1002,2500,485.00,725.00 INV-1003,0800,180.00,905.00 INV-1004,2519,320.50,1225.50 INV-1005,3000,90.00,1315.50 INV-1006,2500,75.00,1390.50 INV-1007,0870,610.00,2000.50 INV-1008,2519,145.00,2145.50 INV-1009,2500,260.00,2405.50
| Column | Read as | What it says |
|---|---|---|
Invoice |
Reference |
Different values: 9 ·
Shortest: 8 characters ·
Longest: 8 characters
Every value is different, so there is nothing to group by
|
Postcode |
Reference |
Different values: 5 ·
Shortest: 4 characters ·
Longest: 4 characters
Kept as text so the leading zeros survive — as a number, 0412 becomes 412 and cannot be got back
|
Amount |
Amount |
Smallest: 75.00 ·
Median: 240.00 ·
Largest: 610.00 ·
Total: 2,405.50
Unusually far from the rest: 1 above 583.75
|
Balance |
Running balance |
Latest: 2,405.50 ·
Smallest: 240.00 ·
Largest: 2,405.50
Each row already includes the ones above it, so we never add these up — thirty balances summed counts the same money thirty times
|
Invoice numbers are references. Adding them up would
produce a figure, and it would mean nothing.
Postcode stayed text, so 0800 is still
Darwin rather than the number 800.
Balance already contains every row above it, so it gets
its latest value and never a sum — which happens to be the same
2,405.50 the amounts add to, so you can check both.
| Every column, listed | What we read it as, how many rows are filled, and how many different values are in it — including the columns we did nothing to. One missing from that list would read as one that was handled. |
| The shape of each one | Smallest, median and largest for numbers. Earliest and latest for dates. The commonest values, with their share, for categories. |
| What's odd about it | Blanks counted per column, values sitting unusually far from the rest, and columns where every value is different — which means the column names a row rather than measuring one. |
| What we had to decide | A code kept as text so its leading zeros survive. A balance we refused to add up, and why. A rate we left out of the totals. These are the calls a spreadsheet doesn't record and nobody else tells you about. |
| Totals, and a breakdown | Where they mean something — with a table under every chart, so no value is ever chart-only. |
Why anything happened. Only arithmetic you could check by hand — totals, change between months, share by category, unusually large values. A confident sentence under a chart is more persuasive than we can justify, so there isn't one.
A total of things that don't add. A balance or a running total already contains every row above it, so summing thirty of them counts the same money thirty times. Those get their latest value instead. Invoice numbers, postcodes, years and interest rates all parse as numbers and are meaningless summed, so they're described rather than added.
When nothing in a file can honestly be added up, you get no total and an explanation — rather than a figure that looks authoritative and isn't.
Whether your data is right. We can tell you a column is empty on 84 rows, that one value sits a long way from the others, or that a reference appears twice. We can't tell you which of those is a problem — that depends on what the file is for, and only you know that. Every line here is a fact about the file, not a verdict on it.