forwardfill

Forward fill, often abbreviated as ffill, is a data imputation technique used to replace missing values in a sequence by propagating the most recent non-missing value forward through subsequent gaps. It is also known as last observation carried forward (LOCF) and is widely used in time series, data cleaning, and analytics workflows.

Applications and context

Forward fill is appropriate when missing values are assumed to be the same as the last observed value over a period with no new information. It is common in time-stamped data, sensor readings, financial records, and spreadsheet workflows where gaps represent temporary pauses rather than new observations. It helps maintain continuity of the series and enables downstream analyses that require complete data.

How it works

The method relies on the order of observations. Starting from the beginning, keep track of the last seen non-missing value. If the current entry is missing, substitute it with that last_seen value; if the current entry is present, update last_seen to this value. If no prior non-missing value exists, the initial missing values remain missing. In practice, many data tools implement this as a simple forward-fill operation (for example, in certain data libraries, ffill or LOCF).

Variants and considerations

Forward fill can apply to numeric and categorical data, carrying forward the previous category or value. It assumes stability of the value over the gap and should not be used when the last observation is not representative. Initial gaps cannot be filled unless supplemented by another imputation strategy.

Relation to other methods

Related techniques include backward fill (filling with the next valid observation), interpolation (estimating values between known points), and more advanced imputation methods. Forward fill is a simple, fast option when its assumptions hold.