Generate a client-ready PowerPoint from a spreadsheet in one command

The 7 slides every business review needs, the python-pptx call for native editable charts, and why pasted images get rejected.

Building a review deck by hand is three hours of copying numbers into shapes. The structure is identical every month, so it should be generated.

The seven slides that cover 90% of reviews

1. Cover — title, date, record count, total
2. Executive summary — four bullets with real numbers, not adjectives
3. Key figures — total / records / average / latest period
4. Trend — one line chart, one point per period
5. Breakdown — top 10 by the primary dimension
6. Details — top 10 table with share of total
7. Findings — momentum, concentration risk, data quality, next step

Native charts, not screenshots

A pasted PNG gets rejected because the first request is always "can you change the axis" or "my brand colour is different". Use real chart objects:

from pptx.chart.data import CategoryChartData
from pptx.enum.chart import XL_CHART_TYPE

cd = CategoryChartData()
cd.categories = list(by_month.index)
cd.add_series("Amount", list(by_month.values), number_format="#,##0")
gf = slide.shapes.add_chart(XL_CHART_TYPE.LINE_MARKERS,
                            Inches(0.9), Inches(1.8), Inches(11.5), Inches(4.9), cd)
gf.chart.series[0].format.line.color.rgb = RGBColor(0x27, 0xE0, 0xA0)

Now the recipient can click the chart and edit it in PowerPoint. That single decision is the difference between a deliverable and a rejected draft.

Write the findings from the data

The value of the deck is in slide 7, and it is arithmetic:

growth = (last - prev) / max(abs(prev), 1) * 100
share  = top_dimension_value / total * 100
flag   = "high concentration risk" if share > 40 else "healthy spread"

Every number in that sentence comes from the table. It is not analysis, but it is the analysis being asked for — and it takes zero minutes instead of an afternoon.

Make it degrade, not crash

A deck builder that dies because there is no date column is worse than useless at 9am on reporting day. Detect the columns you need, skip the slide you cannot build, keep going. Missing data should remove a slide, never the deliverable.

Packaged version

DeckBuilder is this generator: point it at an .xlsx or .csv, get a 16:9 deck with all seven slides, native charts, and auto-written findings. 29 USDT, pip install python-pptx pandas and go.