Writing with evidence

Writing with evidence

Assertion tells; evidence shows. "Our approach is faster" is a claim the reader must take on faith; "it cut the median response from 2.1s to 600ms" is a fact they can weigh. Specific, verifiable support is what separates persuasive, credible writing from generic filler.

Method

1. Replace adjectives with evidence. Every "significant", "powerful", "seamless", "robust" is a claim dressed as a fact. Cut it and show the thing: the number, the example, the before-and-after. If you cannot substantiate the adjective, the reader should not believe it, and you should not write it (see the AI-tell words to avoid in adjust-tone, proofread).
2. Prefer the specific to the general. "Many users struggled" is weak; "of 50 new users, 34 could not find the export button" is strong. Real numbers, named examples, and concrete detail carry weight that vague quantifiers ("often", "most", "a lot") cannot.
3. Choose evidence that fits the claim's weight. A big claim needs strong evidence (data, multiple examples, a source); a small one needs only an illustration. Match the support to what you are asking the reader to believe, and do not over-claim past what your evidence shows.
4. Show your sources where it matters. For factual or contested claims, cite where the number or fact came from, so the reader can trust it and check it (see fact-checking, source-evaluation). Unsourced statistics read as invented, because often they are.
5. Use examples to make the abstract concrete. A principle followed by "for example..." lands where the principle alone slides past. One well-chosen, specific example teaches more than three sentences of generalization (see clear-writing's concrete-nouns rule).
6. Be honest about the evidence's limits. Note the sample size, the caveat, the case where it does not hold. Acknowledged limits build credibility; hidden ones destroy it when found. Overstated evidence is worse than modest evidence honestly framed (see persuasive-writing's concede-what-is-true).

Boundaries

• Evidence must be real: fabricated numbers, invented quotes, or cherry-picked data are the fastest way to lose trust permanently, and AI writing is prone to confident invention, so verify before asserting (see fact-checking).
• Not every sentence needs a citation; over-hedged, over-sourced prose is its own failure. Support the load-bearing claims; state the obvious plainly.
• Evidence supports a point; it does not replace one. A pile of facts without a thesis is a data dump, not writing (see structure-and-flow, data-to-insights).