| Difference to investigate | How to check it | What not to conclude |
|---|---|---|
| Reporting period | Find the table’s data year, not just the article’s update stamp | That the later-looking webpage contains newer data |
| Animal or country | Read the row label and currency | That a cat or Canadian figure applies to a US dog |
| Product definition | Check accident-only, accident-and-illness or embedded wellness | That unlike benefits are contradictory estimates |
| Data method | Identify market data, a quote sample or one insurer’s customers | That each number estimates the same population |
Do not average two incompatible published numbers to create a compromise. First determine whether they are answering the same question. A true correction may simply be relabeling an old benchmark honestly or separating two coverage categories.
If the source cannot be identified, treat the amount as unverified. A number repeated by many websites does not become a separate set of measurements. Following the citation back to the original table is more informative than counting repetitions.