ValueError: Length of values does not match length of index (pandas)

The error#

ValueError: Length of values (900) does not match length of index (1000)

What it means#

A DataFrame column must have exactly one value per row. You assigned a sequence of 900 values to a frame with 1000 rows. Pandas cannot decide which rows should get values and which should not, so it refuses.

Why it happens#

Almost always, the values were computed from a different version of the frame. A typical sequence:

df = pd.read_csv("orders.csv")            # 1000 rows
clean = df.dropna(subset=["price"])       # 900 rows
scores = model.predict(clean[features])   # 900 predictions
df["score"] = scores                      # boom: 900 into 1000

Other routes: predictions from a train/test split assigned to the full frame, a list built in a loop that skipped some rows, or .unique() output (deduplicated, so shorter) assigned back as a column.

How to fix it#

1. Assign to the frame the values actually came from.

clean["score"] = scores                   # 900 into 900

If clean was a slice, create it with .copy() first to avoid the SettingWithCopyWarning.

2. To place partial results into the full frame, assign a Series with the right index. Pandas aligns Series assignments by index label; rows without a match get NaN.

df["score"] = pd.Series(scores, index=clean.index)

This works because clean kept the original row labels. It fails silently after a reset_index, so check which index the slice carries.

3. For lookups, use map or merge instead of positional assignment.

df["city_tier"] = df["city"].map(tier_lookup)

map matches by value and cannot go out of step with row counts.

4. When lengths should match but do not, find the dropped rows.

print(len(df), len(clean))
print(df.index.difference(clean.index))

Then decide: filter the frame first, or stop filtering.

How to prevent it#

Keep one rule: compute new columns from the same frame you assign them to, in the line directly above. When you filter, give the filtered frame a clear name and keep using it. Note that assigning a plain NumPy array skips index alignment entirely — lengths must match exactly — while assigning a Series aligns on labels. Choose deliberately.

Related errors#

• Found input variables with inconsistent numbers of samples — the scikit-learn version of the same drift
• SettingWithCopyWarning
• KeyError: column not in index
• You are trying to merge on object and int64 columns