Charts That Sell · Field notes from the pitch room

The Friday-night data cleanup that saves your Monday pitch

Field note #5 · 6 min read

Monday morning, 8:40. The numbers the client sent at 6 p.m. Friday are open on your screen, the pitch is at ten, and the chart tool is doing that thing again: refusing to sum a column, sorting "10" before "9", and splitting your dates into three columns. Nobody in that meeting blames the data. They blame the person standing next to it.

Here's the rule I learned the hard way: spend fifteen minutes with the spreadsheet on Friday night and you buy back three hours on Monday morning. Charts are downstream of the table. A clean table makes itself into a chart in about ten seconds — paste, pick, export. A dirty table turns every chart tool into an interrogation room.

You don't have a chart problem. You have a table problem, one night early.

Run this checklist against the raw export before you close the laptop. Seven items, all mechanical, none of them require taste or judgment. That's why Friday night is the right time for them.

1. One header row, real names, no merged cells

The first row names the columns: "Region," "Q3 Revenue (€m)," "Growth (%)." If the source file has titles, logos, blank rows, or merged "Q3" headers above sub-columns, delete them. Every chart tool in existence reads row one as the schema; anything clever up there becomes confusing downstream. Unmerge everything — a cell either belongs to one column or it doesn't.

2. Numbers stored as numbers

The single most common Monday disaster. A column that looks numeric but is actually text — exports from finance systems love this — won't sum, won't sort, and quietly drops out of charts. In a spreadsheet, select the column and check: text is left-aligned, numbers right-aligned. Strip thousand separators and currency symbols that are typed as characters; keep the unit in the header instead. Convert the whole column in one pass.

3. One unit per column, stated in the header

"€m" mixed with raw euros in the same column is a chart that lies by a factor of a million. Pick one unit for the column and convert every cell to it. Percentages get their own column, and they stay percentages — never mix a rate and an absolute value on one axis ("revenue in millions" next to "growth in percent" sharing a single scale produces nonsense bars and, worse, nonsense conclusions). That's a combo chart's job: bars for the volume, line for the rate, two axes.

4. One date format, one grain

"Jan," "2025-01," "1/31/25," and "January 31st" are four strings to a spreadsheet and one thing to a human. Normalize the whole column to one machine-readable format (2025-01-31), let the chart display it however it likes. Same for grain: don't mix months and quarters in one column — aggregate one of them. Tools sort strings alphabetically, which is how you end up with "April, August, December" as your timeline.

5. Blanks mean missing — leave them blank

If the source says "N/A," "—," "pending," or a space, it isn't zero. Clear the cell so the chart shows a gap. This one deserves its own note — replacing a missing value with zero is the quietest way to fabricate data, and that's field note #6.

6. One row per thing; kill the totals row

Raw reports love a grand-total row at the bottom. It becomes a phantom category — a giant "Total" bar towering over every real segment — in the first chart you make. Delete it from the chart table (keep a separate summary sheet if finance needs it). The same logic: one row = one category or one period, no subtotals interleaved with data.

You can watch the phantom bar appear for yourself: paste a table that still has its total row into the bar chart maker and the last bar dwarfs every segment you actually care about.

7. Sanity-check the arithmetic before the tool gets blamed

Two checks, thirty seconds each. Eyeball the order of magnitude — is annual revenue plausibly in millions, not billions? And confirm the percent column is a fraction (0.112) or a full number (11.2) consistently, because 11.2 pasted where 0.112 is expected becomes a 1,120% growth rate, and that kind of screenshot circulates internally forever.

Why Friday, not Monday

Because judgment is the only thing that can't be rushed, and none of these seven items need it — but the thing they protect does. The job of Monday morning is deciding which conclusion the data supports and which chart sells it (note #1). Every minute spent converting text to numbers at 8:50 is a minute stolen from the actual argument. Clean the ammunition Friday; aim Monday.

A clean table also survives contact with automation: paste it into ChartThat, and types, units, percentages, and gaps are detected from the clean structure — the parser's whole job is easier when the table is honest.

— Written between pitches. Field note #5 of Charts That Sell. Next: never replace a missing number with zero.