A blank cell can mean several things
Imagine a survey asking employees how many days they worked from home last month. Some respondents enter 0; others leave the answer blank. Those responses are not interchangeable. Zero says that a respondent reported no work-from-home days. A blank might mean the question was skipped, the answer was unavailable, or the person did not understand what to enter. Treating every blank as zero would change the result without evidence that zero was intended.
Missing values can appear in less obvious forms. A dataset might use an empty cell, a dash, “N/A,” or a code such as 999 to represent an unanswered question. It may also contain a value that was never collected because a question did not apply to that person. Before calculating anything, find out what each placeholder means. A code that looks like a number should not accidentally enter an average.
Check how much is missing and where
Start by counting missing entries for each variable you plan to use. Then ask how many records have all the information needed for your particular calculation. A dataset can have few blanks in each column but still lose many usable records when an analysis requires several columns at once.
Next, look for patterns. Are answers to one question more often missing for a particular group, location, or time period? Suppose a training survey has fewer salary responses from senior employees than from junior employees. An average based only on reported salaries may not describe the whole workforce well. You cannot infer the missing salaries from that pattern alone, but you can see why the reported average needs a qualification.
Also check the collection process. If a form was updated halfway through a project, a newly added question will be blank for earlier records by design. That is different from a question people chose not to answer. A short note about when and how the data was collected can help explain a pattern that the spreadsheet alone cannot.
Why the reason for a gap matters
The effect of missing data depends partly on why values are absent. Sometimes a gap is plausibly unrelated to the value being studied: for example, a few responses may be lost during data entry. In other cases, the likelihood of a blank may be connected to the answer itself. People with unusually long commutes, for instance, might be less inclined to report their travel time. Calculating an average from those who replied could then understate the experience of the full group.
Usually, you cannot prove the reason for missingness just by looking at the available cells. What you can do is compare the records with and without a response using other information you have. If the groups differ noticeably, that is a reason to be more cautious. Even if they look similar on recorded characteristics, they may differ in ways your dataset does not capture. Missing-data checks help you recognize uncertainty; they do not make it disappear.
Choose a response that fits the question
For a simple descriptive summary, it may be reasonable to use the available answers and state how many people answered that question. If 80 people were asked and 62 replied, label the resulting average as based on 62 responses, not 80. Avoid filling blanks with zero unless zero is genuinely what the blank represents.
Other approaches include excluding incomplete records from a specific calculation or estimating missing values from information in other columns. Each choice has costs. Exclusion reduces the amount of data used and may skew the result if the excluded records differ from those retained. Filling every blank with the column average can make the data appear more uniform than it is. More elaborate estimation methods require assumptions that should be made explicit rather than treated as observed facts.
When the choice could affect a conclusion, try a modest sensitivity check. For example, calculate a summary using the available answers, then consider whether a plausible range of missing answers would change your main takeaway. This does not recover the unknown values, but it helps distinguish a conclusion that is fairly stable from one that rests heavily on the gaps.
Make the limitation visible to readers
A useful report tells readers what was missing, how missing entries were handled, and which conclusions might be affected. You might write: “Commute time was reported by 62 of 80 respondents; the average uses those 62 answers.” If you excluded records or estimated values, say so in plain language. Keep the denominator alongside a chart or table when different measures rely on different numbers of responses.
The goal is not to make every dataset complete. It is to avoid giving an incomplete dataset more authority than it deserves. By separating blanks from genuine values, checking for patterns, and describing your choices, you give readers a clearer view of both the findings and their limits.
