Explainer
Missing vs. zero: reading incomplete industry data
The short answer
Zero is a reported value. Missing is an absence of a usable value. Replacing missing entries with zero changes the question a calculation answers and can create a result the source never reported.
In this article
An empty table cell looks harmless until a spreadsheet treats it as a number. In an industry report, a blank may mean that a value was not collected, was not applicable or was withheld. None of those meanings is automatically equivalent to a measured zero.
One table, two different averages
Illustrative example
| Reporting unit | Documented count | Status |
|---|---|---|
| A | 10 | Reported |
| B | 20 | Reported |
| C | 0 | Reported zero |
| D | — | Not available |
| E | — | Not available |
The three available values sum to 30. Their mean is 30 ÷ 3 = 10. If the two missing entries are replaced with zero, the mean becomes 30 ÷ 5 = 6. The second calculation has introduced two observations that were never supplied. Neither result estimates the five-unit population mean without further assumptions about the missing values.
The accurately labeled descriptive result is “an average of 10 among the three units with reported values.” This wording does not solve the missing-data problem. It makes the scope of the calculation visible instead of concealing it in a denominator.
Keep status and value in separate fields
A robust worksheet can store the numerical value in one field and the reporting status in another. That preserves the distinction when sorting, importing or changing display formats. A dash used only for visual formatting can lose its meaning when the table is converted to a CSV file.
- Reported: a value is present, including a genuine zero.
- Unavailable: the source supplies no usable value.
- Not applicable: the measurement does not apply to this row.
- Suppressed: the source deliberately withholds the value.
- Below a stated threshold: the source gives a bound, not an exact zero.
These are suggested editorial categories, not a universal codebook. Read the legend of the actual document. The same symbol can mean different things in different releases, and a publisher may distinguish more cases than this small list.
Missing records can distort a trend
Suppose one month includes values from five units and another includes only three. A change in the total can reflect the reporting population as well as changes within units. A like-for-like comparison may be possible for units present in both periods, but that produces a narrower population and should be labeled accordingly.
For gaming industry coverage, this issue can arise when assembling information from separate public documents. Do not invent a value merely to produce a complete-looking chart. Preserve the source’s status and identify which comparisons remain possible with the available records.
Publish a small quality note
The W3C Data on the Web Best Practices includes guidance on data quality and provenance information. In an editorial table, that principle can be applied with a short note stating the number of available records, the missing-data treatment and any excluded rows.
