Pandas value_counts Explained

value_counts() counts how many times each unique value appears in a Series and returns the result sorted by frequency, most common first. It's the fastest way to answer "what's in this column?"

How Do You Get Percentages with normalize=True?

Pass normalize=True to get relative frequencies (proportions that sum to 1) instead of raw counts; multiply by 100 for percentages.

import pandas as pd

s = pd.Series(["chrome", "safari", "chrome", "firefox", "chrome"])

s.value_counts()
# chrome     3
# safari     1
# firefox    1

s.value_counts(normalize=True).round(2)
# chrome     0.6
# safari     0.2
# firefox    0.2

Does value_counts Include NaN?

No — missing values are silently excluded by default, which can hide a data quality problem. Pass dropna=False to count NaN as its own row.

s = pd.Series(["chrome", None, "chrome", None])

s.value_counts()               # chrome  2   (looks fine!)
s.value_counts(dropna=False)
# chrome    2
# NaN       2   <- half your data is missing

How Do You Bin Numeric Values?

For continuous columns, bins=N splits the range into N equal-width intervals and counts values per bin — a quick text histogram without matplotlib.

ages = pd.Series([22, 25, 31, 35, 41, 58, 62])

ages.value_counts(bins=4, sort=False)
# (21.959, 32.0]    3
# (32.0, 42.0]      2
# (42.0, 52.0]      0
# (52.0, 62.0]      2

Use sort=False so bins stay in numeric order instead of frequency order. For custom edges, use pd.cut then value_counts().

value_counts vs groupby + size — Which One?

value_counts handles one Series; for counting combinations across columns, call it on a DataFrame subset or use groupby(...).size(). The DataFrame version of value_counts (pandas 1.1+) counts unique rows.

df = pd.DataFrame({
    "browser": ["chrome", "chrome", "safari", "chrome"],
    "os":      ["mac", "windows", "mac", "mac"],
})

df[["browser", "os"]].value_counts()
# browser  os
# chrome   mac        2
#          windows    1
# safari   mac        1

df.groupby(["browser", "os"]).size()   # same numbers, sorted by key

The practical difference: value_counts sorts by count and drops NaN groups by default, while groupby(...).size() sorts by key and feeds naturally into further aggregation — more on that in pandas groupby.

Common Pitfalls

• Hidden NaN: default dropna=True means the counts can silently exclude a large share of rows. Check with dropna=False first.
• Case and whitespace split categories: "Chrome", "chrome", and "chrome " count separately. Normalize with s.str.strip().str.lower() — see Python string methods.
• It returns a Series: the values become the index. Use .reset_index() to get a two-column DataFrame for plotting or export.
• normalize + dropna interact: with the default dropna=True, proportions are computed over non-null values only.