The prompt
Score 10 of 10, Excellent
Use it when You saw a number in an article, deck or post and need to know where it came from and whether it still says what the original said.
- {{statistic}}
- {{where_seen}}
- {{intended_use}}
- {{source_text}}
Full prompt and why it works
Prompt
The statistic, exactly as I saw it: {{statistic}} Where I saw it (publication, date, link if any): {{where_seen}} How I want to use it: {{intended_use}} Source text I have found so far (the article, the report it cites, the original study), pasted in order from where I saw it back toward the original: """ {{source_text}} """ Task: Help me trace this statistic back to where it started and check whether the version I saw matches the original, so that I can decide whether to use it in {{intended_use}} and how to word it. Numbers drift as they are quoted: the year, the sample, the wording or the base changes. I need to see where. Format: 1. The chain so far: each source in the text above, in order, with what it says about the number (quoted) and whom it credits. 2. Differences: every change between versions, for example "global" becoming "US", "up to 40%" becoming "40%", a 2019 survey presented as current, or a share of respondents presented as a share of all people. 3. Still to find: what is missing from the chain, and where the original is most likely to be published (for example a statistics office, a survey company or a journal), with search terms to use. 4. Verdict: one of Traced and matches, Traced but changed, Not yet traced, with one sentence of reasoning. 5. Safe wording: one sentence I could use today, with the attribution it needs, such as "A 2021 survey of 1,000 US office workers by [organization] found..." That sentence shows the form only. Requirements: - Under 400 words. - Work only from the text above for what each source says. Do not name a specific report, page or author that is not in my text. If you recall the original yourself, you may mention it labeled "From memory, not checked", because I will have to confirm it before using it. - Do not judge whether the statistic is true; judge whether the chain supports the version I saw. If a source does not say where its number came from, mark that link "Not credited" and treat the chain as broken there. If I pasted nothing, skip steps 1 and 2 and give me the search plan, plus the questions to ask of each source I find: who measured it, when, how many people, and what exactly was asked.
Variables
- {{statistic}}
- The number and the words around it, copied exactly, for example "78% of employees say they would quit over a bad manager".
- {{where_seen}}
- Where you saw it, with the date and link if you have them.
- {{intended_use}}
- Where you want to use it, for example "a slide in our hiring proposal".
- {{source_text}}
- Text from each source you have found, in order, with its name and date. Write "none" if you have not looked yet.
Why it works
- What to make. Listing every change between versions, from the year to the base, finds the drift that turns a narrow survey result into a broad claim.
- If unsure. Anything recalled from memory is labeled "From memory, not checked" and the search plan names kinds of sources, never invented reports, so the trace cannot add a new error.
- What good looks like. A three-way verdict and one sentence of safe wording with its attribution give you something you can use or drop today.