Time series and windows Updated November 2026

shift and diff

s.shift(1) moves values down one row so each row sees the previous value (NaN in the first row); s.diff() is s - s.shift(1), the change from the previous row, and s.pct_change() is the relative change. Negative periods look forward instead.

Syntax

Series.shift(periods=1)   |   Series.diff(periods=1)   |   Series.pct_change(periods=1)

Part	What it is
`periods`	How many rows to shift or compare across; negative looks ahead
`fill_value`	shift only: value used instead of NaN for the vacated rows

Examples

import pandas as pd

df = pd.DataFrame({"month": ["Jan", "Feb", "Mar", "Apr"], "sales": [100, 120, 90, 135]})
df["prev"] = df["sales"].shift(1)
df["change"] = df["sales"].diff()
df

Result: month sales prev change 0 Jan 100 NaN NaN 1 Feb 120 100.0 20.0 2 Mar 90 120.0 -30.0 3 Apr 135 90.0 45.0

df["growth"] = df["sales"].pct_change().round(3)
df["growth"].tolist()

Result: [nan, 0.2, -0.25, 0.5]

Feb is 20% above Jan, Mar is 25% below Feb, Apr is 50% above Mar.

df["sales"].shift(-1).tolist()

Result: [120.0, 90.0, 135.0, nan]

Negative periods bring the next row up.

Common mistakes

Integer column becomes float after shift: NaN forces float; pass fill_value=0 if that is acceptable.
Lags cross customer or product boundaries: use groupby(...).shift() instead of a plain shift.
Wrong results because the rows are not in time order: sort by date before shifting.

Practice shift and diff

These drills check your answer on live data, free, no account needed.

Month over month growthReturn the month over month change in `revenue` as a fraction (0.25 means 25 percent growth). The first month has no prior month and is NaN.
Change from the previous periodReturn the change in `revenue` from the previous month (this month minus last month), using `shift`. The first month is NaN.

Questions

How do I shift by time instead of rows?

On a DatetimeIndex use s.shift(freq="D") to move the index by one day, keeping values attached to their new timestamps.

How do I compute year-over-year change on monthly data?

s.pct_change(12) compares each month with the same month a year earlier, assuming one row per month with no gaps.

rollingHow to compute a moving average in pandas with rolling: window size, min_periods to avoid leading NaN, centered windows, rolling sum and std, time windows.
cumsumHow to compute a running total in pandas with cumsum, cumulative sums per group with groupby, cumulative max and count, cumulative percentages, and resets.
resampleHow to resample a time series in pandas: monthly, weekly or daily buckets with sum or mean, upsample with ffill, use a date column with on, and new aliases.
groupbyHow groupby works in pandas: group by one or more columns, apply sum, mean, count or size, get a flat table with as_index=False, and iterate over groups.