Spreadsheets vs Code

You have already heard the high-level argument: spreadsheets are manual, code is reproducible. This chapter makes the comparison concrete with side-by-side scenarios so you can feel, in your bones, where each tool wins and where each one loses.

Two tools, one job

Pretend you are an analyst at a coffee company. Every Monday morning a CSV file lands in your inbox with last week's orders. You need to produce:

• Total revenue.
• Top-selling product.
• Revenue by region.
• A chart for the weekly team meeting.

Let us do this both ways and compare the feel.

The spreadsheet way

1. Open Excel. Double-click the CSV.
2. Excel asks how to import. Pick "general." Click through.
3. Click the column headers to confirm types. Notice that the phone column got auto-converted to scientific notation (5.55E+11). Sigh, fix.
4. Select the revenue column. Read the sum off the status bar. Copy it into a separate "Summary" sheet.
5. Sort the table by product, eyeball the top one. Type it into the summary sheet.
6. Sort by region. Use =SUMIF(B:B, "West", D:D) for each region. Copy values into summary sheet.
7. Highlight summary cells, Insert > Chart, pick bar.
8. Save as weekly_report_2024_03_18.xlsx. Email it.
9. Next Monday: do it all again.

The code way

import pandas as pd

df = pd.read_csv("orders.csv")

print("Total revenue:", df["revenue"].sum())
print("Top product:", df.groupby("product")["revenue"].sum().idxmax())
print(df.groupby("region")["revenue"].sum())
df.groupby("region")["revenue"].sum().plot.bar()

Save the file as weekly_report.py. Next Monday: re-run.

What the comparison actually proves

The "code" version takes more upfront learning. There is no pretending otherwise. But once the learning is paid for, every repeated run is essentially free — and the script is itself a record of how the analysis was done.

That is the central trade-off:

There is no winner in the abstract — only winners for a given problem.

Five common spreadsheet pains, in Pandas

1. Accidental edits

In Excel, every interaction modifies the file. There is no "diff" of what you changed; no git log of why. In Pandas, your script is the file. Diffs are line-by-line. Reviewers can ask "why did you switch from mean to median on line 42?"

2. Copy-paste from another sheet

In Excel, this is invisible after the fact. Was that column copied from Q4_sales.xlsx or Q4_sales_FINAL_v2.xlsx? In Pandas, the source filename or URL is right there in your pd.read_csv(...) call.

3. The 65k-row wall and slowness

.xls capped at 65,536 rows. .xlsx raised it but became slow. Pandas handles tens of millions of rows on a laptop, and millions interactively.

4. Manually-tracked transformations

In a spreadsheet, "I filtered out test users, dropped Saturday, and converted everything to USD" lives in someone's head. In Pandas:

Anyone reading this code knows exactly what you did. There is no oral history.

5. "Refresh the report"

In Excel, a refresh means an analyst manually re-doing the work, or building a fragile chain of Power Query / VBA / OFFSET formulas that one person on the team understands. In Pandas, refreshing is python report.py — or scheduling that command to run automatically every morning.

Where spreadsheets still win

It is important not to overstate the case. Pandas can be overkill for:

• Quick exploratory totals on a tiny file. Excel is faster for "how much did I spend at the grocery store last month?"
• Sharing a working model with a non-technical stakeholder. The CFO wants to play with the model — change assumptions, see results. A spreadsheet is the right vehicle.
• Visual layout. Excel is also a layout tool — borders, shading, merged cells, embedded charts. For documents whose appearance matters, it is still hard to beat.

A mature analyst uses both. The pipeline often looks like: raw data → Pandas (clean, summarize, validate) → Excel (final presentation for humans).

A hybrid example

Pandas can write .xlsx files directly. So you can do the intelligence in code and the presentation in Excel — best of both worlds. (We will return to this in the Exporting Cleaned Data chapter.)

In a real workflow you would write to a real .xlsx file and email it (or upload it to a shared drive). The analysis lives in Python so it is reproducible; the delivery lives in Excel so the audience can read it.

Conceptual comparison: manual vs reproducible

That right-hand pipeline — same inputs, same outputs, audit trail — is the whole game. Pandas is the most popular tool for building it in the Python world.

What to take away

• This course is not anti-spreadsheet. It is pro- reproducibility.
• The single biggest reason analysts move from spreadsheets to code is not speed; it is the ability to re-run an analysis reliably.
• The second biggest reason is transparency: code is the record of the analysis. Spreadsheets are only the result.
• The third is scale: code does not care whether the dataset has fifty rows or fifty million.

Check your understanding