Source Quality Ladder

List the claims your argument stands on, check what others say about each source, rank it, replace the weakest and log the ones you keep, with page and citation, before you write conclusions.

Mixed research and practice: lateral reading has supporting classroom studies, while its packaged form SIFT is more thinly tested (Caulfield, 2019; Caulfield & Wineburg, 2023). Checklists such as CRAAP are practitioner tools (Kapoun, 1998; Blakeslee, 2004). The ladder's rungs and the target source mix are rules of thumb. The reference log is working craft; the citation format is a style convention (Pears & Shields, 2022).

Use it when:

• A report, plan or proposal rests on published claims: articles, industry reports, statistics or vendor figures.
• Someone shares a striking number in a meeting and the team is about to act on it.
• You review or inherit a draft and want to see which claims hold weight, which are borrowed and which you cannot trace.

Skip it when:

• The evidence is your own (interviews, order data, a test): judge it with 2.9 Evidence Quality Judgment.
• Your reader requires another citation style (APA, Chicago): keep the log, but write the entries in that style.

Source evaluation has two generations, and this page uses both. The older one is the checklist: authority, accuracy, objectivity, currency and coverage, written for library teaching (Kapoun, 1998) and later packaged as the CRAAP test (Blakeslee, 2004). These are practitioner tools with a known weak spot: every criterion is judged from inside the page, so a polished but misleading source can pass. The newer generation is lateral reading. You leave the page and see what independent sources say about it, with SIFT as the packaged version (Caulfield, 2019; Caulfield & Wineburg, 2023). The support is mixed: lateral reading has classroom studies behind it, SIFT as a package less so, and in narrow fields with little independent coverage the move can mislead.

So the lateral check comes first. A source that fails it never reaches the ladder, whatever its format suggests. The ladder then gives a default rank by type, which a specific source can rise above or fall below. The checklist questions come last, as depth checks on sources that survived.

Finally, a source you keep needs a citation a reader can follow. Citation format is a convention, not evidence (Pears & Shields, 2022), so the log records where you found each source and what it shows, with the page, while it is open.

Write the topic and the decision. Then list the three claims that would sink the argument if they were wrong, and where each one comes from.

Strong answer: "Claim: small cafes in our region close within a few years of opening. Source: a trade magazine article that cites 'industry figures'." Not: "Claim: cafes are struggling. Source: the internet."

Topic and the decision it feeds:

Illustrative example: the business, the roles and all sources are made up.

The Gothenburg bakery from 2.1 is preparing its board report on the small-cafe route. The sales lead checks three claims in the draft.

Step 1: Claim 1: small cafes close within a few years (trade magazine). Claim 2: cafe numbers in the region are growing (statistics office). Claim 3: cafes that buy from a wholesaler pay less per loaf (a consultancy report).

Step 2: The trade magazine belongs to an equipment supplier and names no source: fail. The statistics office passes, rung 7. The consultancy report drops from rung 6 to rung 4 because its method is not published.

Step 3: The consultancy report also fails objectivity (the firm sells purchasing software) and currency.

Step 4: Claim 1 will come from the business register or be dropped. Claim 3 is replaced by the bakery's own question in the cafe interviews: "What do you pay for bread today?"

Step 5: The statistics office is logged with URL, accessed date and "(Regional Statistics Office, 2025, p. 12)". The report now rests on one strong published source and the bakery's own data, and the answer to "where does that figure come from?" takes ten seconds.