--- title: "What should you do when two credible sources disagree? | Len P. van der Hof" description: "When credible sources disagree, check the claims before the names: population, measure, period and method. A worked founder example and a test for today." image: "https://lenvanderhof.com/media/generated/blog-hero-source-disagreement-vs-tiebreak.1c6cf2df697a.wide.webp" --- [Venture Building](https://lenvanderhof.com/en/blog/category/venture-building/) Research guide # When credible sources disagree: diagnose before choosing sides Before you back the source with the bigger name, check whether the claims actually conflict. Then name the check that could help resolve whatever disagreement remains. Len P. van der HofPublished 6 November 202610 min read Two weathered standing stones mark a fork in the track; a bicycle rests against the gate between the branches. Direct answer When two credible sources disagree, diagnose the disagreement before you pick the bigger name. Check whether both claims cover the same population, definition, measure, place and period. Read the exact passages with their qualifications, compare methods and uncertainty, and trace what the sources share. Separate different questions from genuine conflicts. Where a conflict remains, name the check that could help resolve it and record what you may do meanwhile. ## Key takeaways - Compare claims before publishers. Different populations, measures or periods can explain opposing headlines without making either finding false. - Preserve each claim's qualifications and each source's version and exact passage. An unknown field is not a match. - Let the cause of the disagreement choose the next check. A definition mismatch and conflicting measurements need different work. - A second analysis of shared data can expose a method sensitivity without providing a second measurement. Record the shared origin. - For a decision-critical claim that remains unresolved or in material conflict, the book's default is HOLD. Accepting risk changes the action, not the evidence. **When two credible sources disagree, compare their exact claims before you decide which source to trust.** A report on total spending and a report on median spending per company can point in opposite directions without contradicting each other. First establish whether they answer the same question. That is the difference between a tiebreak and a diagnosis. A **prestige tiebreak** picks the account whose publisher has the bigger name. **Disagreement diagnosis** first establishes whether both accounts measure the same thing, in the same scope, over the same period, and then names the check that could help resolve whatever difference remains. Some apparent disagreements dissolve once those conditions are separated; the rest become a defined open question. Diagnosis can still end with a defensible reason to prefer one account. Equal attention does not mean equal weight. Chapter 6 of [The Verified Research Loop](https://lenvanderhof.com/books/the-verified-research-loop/) gives this work an eight-step sequence. The five steps below adapt it for one founder decision and add the proposition-alignment check from chapter 10. This is the book’s authorial research practice; this article does not show that using it improves decisions. ## What should you do when two credible sources disagree? Name the decision that depends on the disputed claim: a hire, a new segment, a tool purchase. Write down what would change if the claim were wrong. Then work through five steps: 1. **Write each claim separately.** Copy the relevant sentence with its qualifications, and record the page, table or section so another reader can find it. Keep what the source states apart from the recommendation you want to draw. 2. **Check the documents and versions.** Open the relevant sections, tables and method notes. Confirm which product, entity and period they concern. Look for corrections or superseding versions, and record anything you could not obtain. 3. **Compare the conditions.** Set definitions, population, geography, dates, measures, comparisons, denominators and uncertainty side by side. A missing field stays unknown; it is not a match. 4. **Trace what the sources share and classify what differs.** Look for shared data, methods, assumptions, incentives, organizational control or reporting chains. Then name the cause of what remains: scope, measurement, interpretation or another class. 5. **Record the finding and the next step.** State whether the claims are compatible, remain in conflict or cannot yet be compared. Name the missing record, analysis or qualified judgment, with an owner and a deadline. Record separately what action is permitted meanwhile. Appraising each source against the claim it must support is a separate job, covered in [A source tier is a role, not a grade](https://lenvanderhof.com/en/blog/source-tier-is-a-role-not-a-grade/). Here the question is whether two appraised accounts actually contradict each other. ## What exactly are the sources claiming? Chapter 6 separates two objects that are easy to merge in a slide deck or memo: 1. **Claim capsule:** the proposition together with its material truth conditions. 2. **Decision translation:** the bounded inference or recommendation your team proposes to draw from that capsule. Give the capsule a stable identifier and version. Record who or what it concerns; the setting, period and jurisdiction; the method or governing authority; the comparison; the outcome, unit, denominator and result; the uncertainty; and where it may not transfer to your decision. Those details state when the sentence is true. Remove “among respondents,” “during the pilot” or “compared with last year,” and the sentence asserts something else. Keep the qualifiers in the capsule even when the decision translation is shorter; that shorter statement must not widen the group or period, turn an association into a cause, replace a range with a single number, or drop the place or rule that governs its use. Next, give each source body its own **claim-source relation** row instead of putting two sources in one cell. Chapter 6 uses five relation labels: `Supports`, `Contradicts`, `Complements`, `Contextualizes` and `Does not bear on claim`. Two sources that look opposed in a summary may complement each other by answering different parts of the question, or one may only contextualize the definitions while the other measures an outcome. If they estimate different quantities, keep separate capsules and note the gap. Filing both under a broad heading such as “market demand” does not make their numbers comparable. Keep publication date, observation period and version apart; a newer report may reuse an older dataset. Chapter 10 lets you read a difference between periods as change only when a sound model supports that step. If you reconcile two definitions or normalize units, treat that as a new analysis: record the assumptions, keep the original claims recoverable, and test how much the assumptions change the result when the decision could turn on it. ## Which kind of disagreement is it? Chapter 6 distinguishes eight reasons why sources can appear to conflict. Treat them as starting points for investigation; one dispute can involve more than one. KindWhat differs?Useful next checkIdentityEntity, product or instrumentRe-obtain the exact document; verify what it concerns.DefinitionOne word, different constructsCompare operational definitions; split the capsule if needed.ScopePopulation, geography, jurisdiction or outcomeSplit the capsule; check transfer to your intended use.TimeAge, period covered or versionCompare periods; check for corrections or later versions.MethodSamples, measures, models, assumptions or thresholdsCompare specifications, exclusions, sensitivity and uncertainty.ResultAligned scope and methods, materially different observationsPreserve both; seek replication or another relevant line where possible.InterpretationAligned observations, different conclusionsWrite out the reasoning, objections and alternative conclusions.AuthorityLegal or policy instruments clash in hierarchy, applicability or transitionEstablish facts and governing instrument; obtain qualified review. Chapter 10 puts it plainly: different conflicts need different repairs. Splitting a capsule addresses a definition gap but does nothing for two measurements that genuinely disagree; those call for auditing the instrument or finding another data-generating path. Check dependence alongside the diagnosis. Two publishers may analyze one dataset with different exclusions, definitions or models. Chapters 6 and 10 keep such a reanalysis inside the same data line while preserving its dispute about method: it can expose an analytical sensitivity, but it is not a second measurement. Conversely, two teams can collect separate data with the same flawed survey question. Record what is shared and what differs; [Corroboration is not a source count](https://lenvanderhof.com/en/blog/corroboration-is-not-a-count/) covers that provenance check in depth. ## Worked example: is spending rising or falling? **Hypothetical example.** A founder is considering hiring a salesperson to sell customer-support software to small Dutch software companies. A large research firm’s report says spending increased; a specialist publisher’s report says it fell. The founder’s first instinct is to trust the larger firm’s name. The organizations, reports, companies, dates and figures here are invented. Opening the tables shows these claims: FieldLarge research firm’s reportSpecialist publisher’s reportReported findingTotal annual spending on customer-support software rose.Median annual spending on customer-support software fell.PopulationEuropean software companies, including large enterprisesDutch software companies with fewer than 50 employeesComparison periodCalendar year 2025 against 2024Calendar year 2025 against 2024MeasureTotal spending in euros across surveyed companiesMedian spending in euros per surveyed company The periods match. The populations and the measures do not. Total spending across a broad group can rise while median spending in a narrower group falls. A separate arithmetic illustration isolates the measure difference. Take three invented companies, each with complete records for both years: Company or measureAnnual spending, 2024Annual spending, 2025Company A€10,000€8,000Company B€20,000€18,000Company C€70,000€94,000Total€100,000€120,000Median€20,000€18,000 The total rises while the median, the middle value, falls. Opposite directions can coexist even within one group. These numbers illustrate the arithmetic; they are not data behind either fictional report. For the two reports, the first diagnosis is **scope and method**: different populations, and different measures of spending. Keep separate capsules for the European total and for the median among small Dutch respondents. The specialist’s population is closer to the founder’s intended customers, but its sampling, weighting, exclusions and uncertainty still need appraisal before the finding can be generalized. The repaired research note could read: > The reports measure different quantities. Over the same period, total spending among surveyed European software companies rose, while median spending among surveyed Dutch software companies with fewer than 50 employees fell. Neither establishes demand for our subscription among the prospects we intend to approach, let alone that a sales hire will pay off. Now suppose the larger firm supplies a breakdown for small Dutch software companies, using the same median measure and period. It shows an increase; the specialist still reports a decrease. Matching population and measure has not settled the question. Compare respondent selection, what counts as spending, exclusions, weighting and uncertainty before judging whether a material conflict remains. The open question is now precise: **what explains the remaining difference, and would the answer change the hiring decision?** ## What if the sources still conflict? Write down what could tell the remaining explanations apart. “Find another credible source” leaves the task too loose. In the hypothetical case, you could ask whether one report counts service fees bundled with subscriptions while the other excludes them. Obtain both definitions and, if the records allow, compare results under one definition. Record that as an analysis to perform, not an explanation already established. If the observations agree but the recommendations differ, inspect the reasoning between them; the split may sit in an assumption about future demand or the cost of waiting. Name those assumptions and the evidence each would need. For any proposed new observation, ask what each explanation would lead you to expect. [Diagnostic evidence vs fitting evidence](https://lenvanderhof.com/en/blog/diagnostic-evidence-vs-fitting-evidence/) develops that comparison. Use it to choose the next investigation, without inventing probabilities to make the disagreement look quantified. A diagnosis can justify preferring one account. You might establish that only one measures the relevant population, or find a corrected error in a later version. Record that reason and its evidence. Do not substitute “primary wins” for it: chapter 6 explicitly rejects that rule. Where material counterevidence survives proposition and version alignment, chapter 6 sets the relationship summary to `Conflicting`, even when other sources support the claim. The favorable sources stay in the ledger, but they are not averaged with the adverse one into a clean verdict. Keep the check state separate as well. A methods appendix nobody can obtain is `Inaccessible`; an issue that needs authority outside the team’s scope is `Specialist-owned`. `Inaccessible`, `Searched not found` and `Not yet checked` are different states, and none of them means the claim is false. Keep that evidence summary separate from the **decision disposition**, the owner’s choice about acting. For a decision-critical claim that is unresolved or in material conflict, the book’s default is `HOLD`. The owner may record another disposition, such as `Pilot`, `Specialist review` or `Accepted risk`. A risk acceptance names the affected claims, the reason, safeguards, expiry and review trigger. It changes the action record, not the evidence, and it does not license an unqualified `Proceed`. Assigning another check does not change the evidence either; update the record once the result is in and assessed. ## Try this today Take one disagreement you settled by trusting the bigger name, preferably one that still affects a decision. Open both source documents and write their exact claims side by side, with the scope, population and date beside each. Note the report versions and the page or table references. Copy this record, with a separate column or note for each source: ``` Decision affected / owner: Source A / version / exact passage: Source B / version / exact passage: For each claim: Exact wording and qualifications: Population, setting and geography: Observation period / publication date / version: Definition, measure, unit and denominator: Method, comparison and uncertainty: What differs: What data, assumptions or reporting do the sources share? Can the two findings coexist? Why, or what is missing? What conflict remains after alignment? Next discriminating check / owner / deadline: Decision disposition / conditions / review trigger: ``` Mark each difference before you decide which account deserves more weight. If a difference explains the apparent conflict, rewrite your memo to keep both findings and their limits. If not, state the open question precisely, name what could answer it, and mark missing information as unknown. Finish with a handoff test. Ask a colleague to reopen both passages from your notes alone and explain why the claims can coexist or exactly what still conflicts, then name the next check, its owner and what you may do meanwhile. If they cannot, the comparison is unfinished; record the gap. Passing makes your diagnosis inspectable. It does not show that either source is right. Cite this:When credible sources disagree: diagnose before choosing sides.Len P. van der Hof. [https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/](https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/) · Published 6 November 2026. ## Sources 1. [The Verified Research Loop](https://lenvanderhof.com/books/the-verified-research-loop/) 2. [A source tier is a role, not a grade](https://lenvanderhof.com/en/blog/source-tier-is-a-role-not-a-grade/) 3. [Corroboration is not a source count](https://lenvanderhof.com/en/blog/corroboration-is-not-a-count/) 4. [Diagnostic evidence vs fitting evidence: what changes a belief?](https://lenvanderhof.com/en/blog/diagnostic-evidence-vs-fitting-evidence/) ## Further reading - [A source tier is a role, not a grade](https://lenvanderhof.com/en/blog/source-tier-is-a-role-not-a-grade/) - [Corroboration is not a source count](https://lenvanderhof.com/en/blog/corroboration-is-not-a-count/) - [Diagnostic evidence vs fitting evidence: what changes a belief?](https://lenvanderhof.com/en/blog/diagnostic-evidence-vs-fitting-evidence/) - [The Verified Research Loop](https://lenvanderhof.com/books/the-verified-research-loop/) About the author ## [Len P. van der Hof](https://lenvanderhof.com/en/authors/len-p-van-der-hof/) Entrepreneur, AI Innovator and Venture Builder Len P. van der Hof builds practical AI systems, digital ventures and evidence-informed tools for founders. ```json { "@context": "https://schema.org", "@graph": [ { "@type": "Person", "@id": "https://lenvanderhof.com/#person", "name": "Len P. van der Hof", "alternateName": [ "Len van der Hof", "L.P. van der Hof", "Leendert Pieter van der Hof" ], "honorificSuffix": "MSc", "url": "https://lenvanderhof.com/", "image": [ "https://lenvanderhof.com/photos/len-portrait-1.jpg", "https://lenvanderhof.com/photos/len-portrait-2.jpg", "https://lenvanderhof.com/photos/len-portrait-3.jpg", "https://lenvanderhof.com/photos/len-portrait-4.jpg", "https://lenvanderhof.com/photos/len-speaking.jpg", "https://lenvanderhof.com/photos/len-hero.jpg" ], "jobTitle": "Entrepreneur, AI Innovator and Venture Builder", "description": "Len P. van der Hof, MSc, is a Dutch entrepreneur and AI innovator in Zwijndrecht. He builds ReasonKit, MindSesh, Undominated.ai, books under his name, the fiction imprint LPH98.lifestyle, and technology ventures through LPH98.ventures. Eleven titles in Systems for the Strategic Self are available now, in English and Dutch.", "address": { "@type": "PostalAddress", "addressLocality": "Zwijndrecht", "addressCountry": "NL" }, "alumniOf": { "@type": "CollegeOrUniversity", "name": "Rotterdam School of Management, Erasmus University" }, "knowsAbout": [ "Artificial intelligence", "AI agents", "Agentic AI systems", "LLM routing", "SEO", "Generative engine optimization", "Answer engine optimization", "Venture building", "Founder performance", "Founder psychology", "Evidence-based decision-making" ], "sameAs": [ "https://www.linkedin.com/in/lenvanderhof/", "https://x.com/LenvanderHof", "https://www.youtube.com/channel/UCTG20buKqYYbitqqf7l3zJA", "https://www.instagram.com/Lenvanderhof/", "https://www.threads.com/@lenvanderhof", "https://github.com/Lenvanderhof", "https://huggingface.co/LPH98", "https://www.npmjs.com/~lenvanderhof", "https://www.goodreads.com/author/show/70983905.Len_P_van_der_Hof", "https://www.amazon.com/author/lenvanderhof", "https://www.bol.com/nl/nl/b/len-p-van-der-hof-msc/609879394/", "https://bsky.app/profile/lenvanderhof.com", "https://mastodon.social/@Lenvanderhof", "https://crates.io/users/Lenvanderhof", "https://cursor.com/@Lenvanderhof", "https://medium.com/@Lenvanderhof", "https://gitlab.com/Lenvanderhof", "https://hub.docker.com/u/lenvanderhof/", "https://dev.to/lenvanderhof", "https://www.facebook.com/Lenvanderhof", "https://soundcloud.com/Lenvanderhof" ], "affiliation": [ { "@id": "https://lenvanderhof.com/#publisher" }, { "@id": "https://lenvanderhof.com/#mindsesh" }, { "@id": "https://lenvanderhof.com/#lifestyle" } ] }, { "@type": "WebSite", "@id": "https://lenvanderhof.com/#website", "url": "https://lenvanderhof.com/", "name": "Len P. van der Hof", "description": "Len P. van der Hof, MSc, is a Dutch entrepreneur and AI innovator in Zwijndrecht. He builds ReasonKit, MindSesh, Undominated.ai, books under his name, the fiction imprint LPH98.lifestyle, and technology ventures through LPH98.ventures. Eleven titles in Systems for the Strategic Self are available now, in English and Dutch.", "inLanguage": [ "en", "nl" ], "publisher": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "Organization", "@id": "https://lenvanderhof.com/#publisher", "name": "LPH98.ventures", "url": "https://lph98.ventures", "founder": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "Organization", "@id": "https://lenvanderhof.com/#mindsesh", "name": "MindSesh", "url": "https://mindsesh.net", "founder": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "SoftwareApplication", "@id": "https://lenvanderhof.com/#reasonkit", "name": "ReasonKit", "url": "https://reasonkit.sh", "creator": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "SoftwareApplication", "@id": "https://lenvanderhof.com/#undominated", "name": "Undominated.ai", "url": "https://undominated.ai", "creator": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "Organization", "@id": "https://lenvanderhof.com/#lifestyle", "name": "LPH98.lifestyle", "url": "https://lph98.lifestyle", "founder": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "ImageObject", "@id": "https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/#primaryimage", "url": "https://lenvanderhof.com/media/generated/blog-hero-source-disagreement-vs-tiebreak.1c6cf2df697a.wide.webp", "contentUrl": "https://lenvanderhof.com/media/generated/blog-hero-source-disagreement-vs-tiebreak.1c6cf2df697a.wide.webp", "representativeOfPage": true }, { "@type": "BreadcrumbList", "@id": "https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/#breadcrumb", "itemListElement": [ { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://lenvanderhof.com/" }, { "@type": "ListItem", "position": 2, "name": "Blog", "item": "https://lenvanderhof.com/en/blog/" }, { "@type": "ListItem", "position": 3, "name": "Venture Building", "item": "https://lenvanderhof.com/en/blog/category/venture-building/" }, { "@type": "ListItem", "position": 4, "name": "When credible sources disagree: diagnose before choosing sides", "item": "https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/" } ] }, { "@type": "WebPage", "@id": "https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/#webpage", "url": "https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/", "name": "When credible sources disagree: diagnose before choosing sides", "description": "When credible sources disagree, check the claims before the names: population, measure, period and method. A worked founder example and a test for today.", "isPartOf": { "@id": "https://lenvanderhof.com/#website" }, "primaryImageOfPage": { "@id": "https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/#primaryimage" }, "breadcrumb": { "@id": "https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/#breadcrumb" }, "inLanguage": "en-GB" }, { "@type": "BlogPosting", "@id": "https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/#article", "mainEntityOfPage": { "@id": "https://lenvanderhof.com/en/blog/source-disagreement-vs-tiebreak/#webpage" }, "headline": "When credible sources disagree: diagnose before choosing sides", "description": "When credible sources disagree, check the claims before the names: population, measure, period and method. A worked founder example and a test for today.", "datePublished": "2026-11-06T20:00:00.000Z", "author": { "@id": "https://lenvanderhof.com/#person" }, "publisher": { "@id": "https://lenvanderhof.com/#person" }, "image": [ "https://lenvanderhof.com/media/generated/blog-hero-source-disagreement-vs-tiebreak.1c6cf2df697a.square.webp", "https://lenvanderhof.com/media/generated/blog-hero-source-disagreement-vs-tiebreak.1c6cf2df697a.landscape.webp", "https://lenvanderhof.com/media/generated/blog-hero-source-disagreement-vs-tiebreak.1c6cf2df697a.wide.webp" ], "articleSection": "Venture Building", "inLanguage": "en-GB" } ] } ```