The short answer
Represent an unavailable observation as missing, not as a measured zero. Zero is a real value; missing means the evidence is not present. This distinction affects sorting, filtering, averages and the decisions readers make from a comparison table.
A worked example
A fictional imported publication has no traffic value, while another has a supplied traffic value of zero. Filling both cells with zero would make the first look measured. It would also distort a later average. The marketplace should show 'Unavailable' for the first and retain the explicit zero only when the dataset truly supplies it.
A practical checklist
- Preserve a nullable representation in the data model and a readable unavailable label in the interface. Do not repair empty cells with invented values.
- Decide how each filter treats missing records and explain that behavior. A minimum-score filter can exclude unavailable values without claiming they scored below the threshold.
- When aggregating, report the count of available observations and the number missing. Keep the denominator visible so a mean does not appear to cover the entire catalog.
What to avoid
Avoid describing a complete-looking table as fully measured when source cells were filled by a default. Imported metrics remain owner-supplied unless a provider and observation date are established. An active status controls visibility, not the reliability of the metric.
A useful follow-up
Should missing scores be displayed as zero to simplify the table?
No. Use an unavailable state and make filter behavior clear; zero and missing carry different meanings.