A number can be accurate and still tell an incomplete story.

Tesla's FSD safety claim — stating that its vehicles drive more than seven times farther between crashes than average — made headlines this month when Reuters found those figures were self-produced and submitted to European regulators without independent verification. A Norwegian official noted they're nearly impossible to cross-reference with official accident statistics. Key methodology details remain redacted even in filings to U.S. regulators.

The number may very well be real. But the story that it's being used to tell is not the full one.

This pattern shows up everywhere data is used to make a public case — not just in autonomous vehicles or government surveys. It shows up in NPS scores that only reached satisfied customers. Employee engagement surveys administered during a good quarter. Conversion rates that don't account for who never made it to the funnel.

The question worth asking before you publish a number isn't just "is this accurate?" It's:

➡️ Who is and isn't in this sample? ➡️ What happened after this data was collected that changes how we read it? ➡️ What is this number being used to argue — and does the methodology actually support that argument?

Technical accuracy isn't the finish line. Context is.