At each step from data to facts to interpretations, what can go wrong?
Bad observation corrupts the data; biased selection and over-condensing corrupt the facts; unwarranted assumptions or missing context corrupt the interpretation.
Errors compound down the chain, so a flaw early on poisons everything built on top:
| Stage | How it goes wrong | Result |
|---|---|---|
| Data | Observe wrongly, measure imprecisely, record only partially | Faulty data |
| Facts | Select in a distorting way, connect illogically, condense too aggressively | Faulty facts |
| Interpretation | Rest on unproven premises, or lay things out stripped of context | Faulty interpretations |
Notice that none of these requires an outright lie — each is a quiet slant in an ordinary-looking step. A statistic can be perfectly "factual" yet rest on cherry-picked data, then get an interpretation ("X is the new smoking") with no real basis. The lesson: interrogate every stage, not just the final claim.
Tip: "Over-condensing" is a favourite trick — squeezing a nuanced dataset into one dramatic number ("93%!") that technically came from the data but no longer represents it faithfully.