dtypes decide what you can do. Object columns will not sum. Integers cannot hold NaN unless you use pandas’ nullable types. Convert early, after you have looked at a few values.

Goal

Turn text numbers into numbers, dates into datetimes, and labels into categories — without silently turning bad cells into NaN unless you asked.

See the problem

df = pd.DataFrame(
    {
        "product": ["A", "B", "A", "C"],
        "units": ["12", "7", "9", "n/a"],
        "price": [10.5, 22, 10.5, 31],
        "flag": [1, 0, 1, 0],
    }
)
print(df.dtypes)
print(df)
print()
print("sum of units column (strings concatenate):", df["units"].sum())

`to_numeric`

df = pd.DataFrame(
    {
        "product": ["A", "B", "A", "C"],
        "units": ["12", "7", "9", "n/a"],
        "price": [10.5, 22, 10.5, 31],
        "flag": [1, 0, 1, 0],
    }
)
df["units"] = pd.to_numeric(df["units"], errors="coerce")
print(df.dtypes)
print(df)
print("units NA after coerce:", int(df["units"].isna().sum()))
print(df.loc[df["units"].isna()])

errors="raise" (default) — explode on bad values
errors="coerce" — bad values become NaN
errors="ignore" — return the original (rarely what you want)

`astype`

df = pd.DataFrame(
    {
        "product": ["A", "B", "A", "C"],
        "flag": [1, 0, 1, 0],
    }
)
df["flag"] = df["flag"].astype(bool)
df["product"] = df["product"].astype("category")
print(df.dtypes)
print()
print(df["product"].cat.categories)
print(df)

Categories save memory on repeated labels and keep a stable order you can set with cat.set_categories.

`convert_dtypes`

Asks pandas to pick nullable dtypes (Int64, string, boolean).

df = pd.DataFrame(
    {
        "product": ["A", "B", "A", "C"],
        "units": ["12", "7", "9", "n/a"],
        "flag": [1, 0, 1, 0],
    }
)
print(df.convert_dtypes().dtypes)

Downcast (optional)

prices = pd.Series([10.5, 22.0, 31.0])
print(pd.to_numeric(prices, downcast="float"))

Useful on large frames; skip it on these toy tables.

Pitfall

"1,200" will not parse as 1200 until you strip commas: s.str.replace(",", "", regex=False) then to_numeric. The String chapter covers that.