A small table lists the number of birds seen during five visits. Four cells contain numbers; one is blank. Replacing the blank with zero makes the sheet look complete, but it may also invent an observation that never happened.
The first question is what the empty cell means. The answer belongs in the dataset’s documentation, not in a guess based on appearance.
Separate absence of a value from a measured absence
An observed zero might mean a visit took place and no birds were recorded under the stated method. A blank might mean the visit was cancelled, the note was lost or the value was not applicable. These situations cannot be treated as interchangeable without justification.
If the documentation uses special codes, check them before calculating. A number such as 999 might be a missing-value code rather than a real count.
Watch what changes in an average
Consider the invented values 2, 4, blank, 6 and 8. The mean of the four observed values is 5. Replacing the blank with zero and dividing by five gives 4. The different answer comes from a different assumption, not a new observation.
Neither calculation settles how the missing visit should be handled in a real study. That depends on the question, collection process and appropriate analysis.
Keep uncertainty visible
Record the number of available observations and describe how missing entries were handled. If you create a cleaned copy for analysis, retain enough provenance to distinguish recorded data from your transformations.
Do not silently replace every blank just because a charting tool prefers a complete column.
Our dataset-reading guide helps locate definitions and collection notes. A tidy table is useful only when the tidying preserves the meaning of its contents.