Source library
Official references for checking frames, methods, and citations.
A good citation does more than decorate a claim. It lets a reader inspect the original record, the measurement definition, the denominator, the comparison class, and the limit on what the evidence can support.
This library groups official and primary references by the judgment a reader is trying to make. Use it before publishing a media-framing critique, accepting a vendor readout, citing a public dataset, or turning a measurement method into a budget recommendation.
Fast path: use the source and vendor evaluation desk to choose a workflow, the worked media claim example to see one claim move from source trail to supportable wording, the source triangulation checklist when several citations need independent confirmation, the evaluation worksheet to record a review, and the methodology page for Measurement Press's evidence grades.
Advertisement
Choose the right reference
| Reader job | Cite first | Then cite | Baseline workflow |
|---|---|---|---|
| Audit a public media claim | The original article, transcript, filing, dataset, or public record. | Newsroom standards only when they clarify sourcing, correction, or response expectations. | Media claim audit worksheet, worked media claim example, anecdote and exemplar checklist, and disconfirming evidence checklist |
| Inspect quote weight or response balance | The original article, full statements, response requests when disclosed, and any records used to support the central claim. | Newsroom standards only after the quote roles, placement, and evidence parity are visible. | Quote weight and response standard checklist |
| Triangulate several sources | The closest primary record, independent corroborating source, counter-source, denominator, comparison class, and method disclosure. | Secondary analysis only after same-origin citations are separated from independent confirmation. | Source triangulation checklist |
| Map source roles and incentives | The quoted source, cited report, dashboard, data owner, sponsor, vendor, or anonymous claim closest to the frame. | Official standards and secondary analysis only after the source role, incentive, and missing counter-source are visible. | Source role and incentive map |
| Check anecdote or exemplar framing | The case-selection path, source role, closest record, denominator, comparison class, counterexamples, and wording that marks whether the example is illustrative or representative. | Secondary analysis only after the example's source path and broader universe are visible. | Anecdote and exemplar framing checklist |
| Check case-study generalizability | The case selection note, eligible universe, baseline, outcome definition, comparison design, transfer conditions, and wording that marks what can travel beyond one setting. | Secondary analysis only after the case boundary, denominator, and comparison are visible. | Case study generalizability checklist |
| Check a base rate or denominator | The official statistical table, data release, record, or method note with date, population, and eligibility base. | Secondary analysis only after the underlying table, record, or denominator choice is visible. | Denominator framing examples and source quality scorecard |
| Check an average or composition claim | The subgroup values, subgroup weights, eligibility rules, and source table or report that can rebuild the blended average, rate, share, or index. | Trend, survey, campaign, or benchmark interpretation only after mix changes are separated from within-group movement. | Composition mix and average checklist and denominator framing examples |
| Trace a public-record claim | The record custodian, docket, filing, data portal, table, request response, or release closest to the claim. | Access rules, data documentation, revision notes, and citation guidance only after the record universe is clear. | Public records and denominator checklist |
| Check a before-and-after or trend claim | The original series, method note, denominator, full window, event timeline, and revision status. | Comparable prior periods, peer series, matched markets, holdouts, forecasts, or model baselines only after the plotted source is reproducible. | Before-and-after trend checklist |
| Check a timeline or event-order claim | The timestamped source trail, claim chronology, prior state, mechanism, lag window, concurrent events, and comparison timeline. | Primary records, earlier planning signals, peer timelines, prior periods, and disconfirming context only after the event order is documented. | Timeline and event-order framing checklist |
| Review a causal claim | The exact causal verb, observed fact, counterfactual, comparison timing, alternate explanations, denominator, and strongest supportable wording. | Method references, primary records, model notes, or survey disclosures only after the claim is separated from the evidence that supports it. | Causal claim review protocol |
| Read a poll, survey, or brand-study claim | The method disclosure with sponsor, field organization, population, sample source, field dates, question wording, weighting, and uncertainty. | Survey standards and methodology examples only after the specific study's own disclosure is visible. | Survey and poll claim checklist |
| Read a sponsored research or vendor report claim | The sponsor role, report producer, data owner, method note, data universe, denominator, comparison, and limitations. | Disclosure and survey standards only after the specific report's own source trail is visible. | Sponsored research and vendor report checklist |
| Review a vendor measurement result | The vendor's method note, assignment rule, inclusion rules, outcomes, and uncertainty statement. | Industry standards for definitions, quality controls, and disclosure expectations. | Source and vendor evaluation desk |
| Prepare a campaign reporting handoff | The campaign, placement, creative, destination, audience, test-cell, event, outcome, duplicate, change-log, context, traffic-quality, lead-status, matchback, attribution, and incrementality fields. | Data-quality and outcome references only after the campaign source trail, report structure, field dictionary, and term definitions are visible. | Advertiser intake worksheet, campaign readiness dashboard, campaign pacing and makegood register, private marketplace buying terms guide, private marketplace reporting field dictionary, campaign data-layer spec, campaign reporting sample, package performance readout, campaign reporting terms glossary, and attribution window checklist |
| Read attribution-window or conversion-lag claims | The touchpoint rule, lookback window, conversion window, pre-period intent check, data-lag allowance, reporting cutoff, and mature cohort definition. | Outcome and data-quality references only after the report separates credited outcomes from incremental outcomes. | Attribution window and conversion lag checklist |
| Read landing-page or lead-quality claims | The campaign source trail, landing-page funnel, lead filters, qualification definition, follow-up coverage, and comparison rule. | Traffic-quality and outcome references only after the report separates media delivery, page conversion, form quality, and sales status. | Landing page and lead quality checklist |
| Read identity matchback or clean-room outcomes | The identity unit, match method, eligible universe, outcome source, reporting windows, unmatched records, aggregation thresholds, and suppressed rows. | Outcome and data-quality references only after the specific report's matching and reporting rules are visible. | Identity matchback checklist and privacy-safe collaboration checklist |
| Read audience targeting results | The pre-exposure behavior, match rule, model score, retargeting trigger, CRM status, or retail-media eligibility that placed users in the segment. | Outcome and data-quality references only after the report separates audience quality from media-caused lift. | Audience selection bias checklist |
| Interpret an MMM result | Model specification, control variables, priors, diagnostics, calibration evidence, and model influence checks. | Official MMM documentation for what fit, baseline, causal assumptions, and calibration evidence can support. | MMM causal validity checklist and MMM calibration evidence checklist |
| Read attention, viewability, invalid-traffic, reach, frequency, or video exposure metrics | Measurement methodology, exposure definition, eligible impressions, filtration logic, identity unit, deduplication rule, and device coverage. | IAB and MRC references for attention, viewability, invalid traffic, and data quality language. | Viewability and invalid traffic checklist, reach and frequency checklist, attention measurement guide, and CTV/video checklist |
| Plan programmatic inventory | Placement map, seller authorization, seller records, SupplyChain path, creative sizes, package IDs, deal keys, floor inputs, refresh boundaries, reporting fields, and reporting dimensions. | Technical and reader-experience references that help buyers verify clean inventory. | Media kit and ad specs, private marketplace buying terms guide, advertiser intake worksheet, campaign readiness dashboard, campaign pacing and makegood register, contextual package proof checklist, supply path transparency checklist, ad yield and deal-readiness checklist, and private marketplace reporting field dictionary |
Citation examples by claim type
Use these examples to keep the citation as close as possible to the claim being made. The public reference should let a reader repeat the check: find the same record, inspect the same method, and see the same limit on interpretation.
| Claim type | Citation should show | Careful wording | Weak wording to avoid |
|---|---|---|---|
| Public-record count | Record owner, release date, table or docket, population, exclusions, and access path. | "The record shows reported cases in this defined universe during this window." | "The record proves the problem is rising everywhere." |
| Disconfirming context | Central claim, closest primary source, alternate denominator, fair comparison, missing source role, scope limit, and the fact that would materially reduce confidence. Use the disconfirming evidence checklist when the weakener is not yet named. | "The frame is stronger if it survives this missing denominator, counter-source, or alternate comparison." | "No visible caveat means the claim is settled." |
| Before-and-after trend | Original series, chart window, denominator, method stability, event timing, seasonality, concurrent changes, and comparison rule. | "The metric changed after the event; the available comparison supports this bounded interpretation." | "The event caused the entire change shown in the chart." |
| Timeline or event-order claim | Timestamped records, prior state, claimed mechanism, expected lag, concurrent changes, and a comparison timeline. Use the timeline and event-order framing checklist when chronology is being used as causal evidence. | "The later action followed the event; the record should show whether it was already underway and what else changed in the same window." | "The first visible event caused everything that followed." |
| Quote-weight claim | Quote role, source incentive, placement near the claim, response request if disclosed, and whether each material claim has comparable evidence support. | "The article gives the central claim document support but summarizes the response in one sentence." | "The story is balanced because it quotes both sides." |
| Anecdote or exemplar claim | Example selection path, closest record, source role, denominator, comparison class, counterexamples, and whether the example is labeled as illustrative or representative. Use the anecdote and exemplar checklist when a vivid case may be doing broader evidence work. | "This documented case illustrates how the issue can appear; prevalence depends on the broader record." | "This example proves the issue is typical." |
| Rate or denominator claim | Numerator, denominator, eligibility rules, time period, geography, and whether the figure is revised or preliminary. Use the denominator framing examples when the base may change the frame. | "The rate changes when the denominator is limited to eligible cases." | "The percentage proves the headline comparison." |
| Average or composition claim | Subgroup values, subgroup weights, eligibility rules, comparison period, and whether the same groups are present before and after. Use the composition mix and average checklist when the aggregate may be hiding a mix shift. | "The blended average moved under this group mix; subgroup movement determines how broadly the claim applies." | "The average proves every group improved." |
| Poll or survey result | Sponsor, field organization, population, sample source, field dates, question wording, weighting, and uncertainty. | "Among the sampled population, respondents answered this wording at this time." | "The public believes this." |
| Sponsored research or vendor report claim | Sponsor role, report producer, data universe, sample selection, denominator, comparison class, method note, and limitations. | "The report describes this defined sample, customer base, or data set under the stated method." | "The report proves the category conclusion." |
| Brand-lift or survey-lift readout | Recruitment method, exposed/control definition, sample bases, field dates, questions, weighting, and confidence statement. | "The study reports a measured difference for surveyed respondents under this design." | "The campaign changed brand perception in the market." |
| Experiment or lift test | Assignment rule, holdout definition, suppression logs, control exposure, outcome window, leakage controls, exclusions, minimum detectable effect, and uncertainty. Use the minimum detectable effect planning checklist before launch, the comparison market and holdout planning guide before accepting the comparison, the holdout leakage and suppression QA checklist before trusting the control boundary, and the randomized lift test readout checklist before accepting the result. | "The test estimates incremental effect for the eligible population under the observed conditions." | "The channel always drives this much incremental revenue." |
| Campaign data-layer result | Campaign ID, placement ID, creative ID, destination ID, audience rule, test cell, outcome source, duplicate rule, and change log. Use the campaign data-layer spec before accepting a source trail as complete. | "The campaign preserved the fields needed to separate delivery, traffic, lead, and outcome evidence." | "The clean data layer proves the campaign created lift." |
| Campaign report sample | Package ID, deal key, context group, placement ID, creative theme, destination, qualified visit rule, lead status, comparison rule, and decision limit. Use the private marketplace reporting field dictionary for package and placement fields, the campaign reporting sample for structure, the private marketplace package performance readout for buyer-facing package language, and the campaign reporting terms glossary before turning a readout into a renewal claim. | "The report separates delivery, response, lead quality, and evidence-level language." | "The strongest table cell proves the best budget choice." |
| Landing-page or lead-quality result | Campaign source, destination, form funnel, duplicate rule, disqualification reasons, qualified definition, follow-up coverage, and comparison rule. Use the landing page and lead quality checklist before treating form fills as demand. | "The campaign produced observed qualified leads under this source trail, page, filter, and follow-up process." | "The campaign created incremental pipeline." |
| Identity matchback result | Identity unit, match rate, join rule, unmatched universe, outcome source, reporting window, deduplication, and comparison rule. Use the identity matchback checklist before accepting matched outcomes as lift. | "Matched records produced observed outcomes under this join rule and window." | "The matched conversions prove incremental impact." |
| Attribution-window result | Touchpoint rule, lookback window, conversion window, conversion-lag curve, pre-period intent check, data-lag allowance, mature cohorts, and comparison rule. Use the attribution window and conversion lag checklist before accepting credited outcomes as lift. | "The campaign received credit for observed outcomes inside the stated window." | "The credited conversions prove incremental impact." |
| MMM result | Model period, controls, priors, saturation assumptions, diagnostics, calibration evidence, and uncertainty interval. | "The model estimates contribution under its assumptions and available controls." | "The model proves the budget caused the outcome." |
| Attention, viewability, or invalid traffic metric | Eligible impressions, measurable impressions, exposure definition, filtered counts, device coverage, vendor method, and reporting window. | "The metric describes exposure quality for measurable inventory." | "Attention caused sales lift." |
| Programmatic inventory claim | Ad slot, seller authorization, seller record, SupplyChain path when available, size set, placement context, refresh rule, reporting key, package proof, floor rationale, sell-through field, and package boundary. Use the contextual package proof checklist before turning context fit into a performance claim, the supply path transparency checklist before accepting seller-path language, and the ad yield and deal-readiness checklist before treating scarcity, floors, or sell-through as renewal evidence. | "The placement is eligible for this package and can be audited with this reporting key and seller-path status." | "The inventory quality guarantees performance." |
| Reach or frequency claim | Reach unit, eligible universe, measured universe, deduplication method, frequency bands, cap rule, overlap check, and unmatched records. | "The campaign reached this measured universe under this identity definition and frequency distribution." | "The campaign reached this many people and caused the observed outcomes." |
Claim-language templates
For a public data point:Cite the table, release, or record closest to the number. State the population, time period, denominator, and revision status before explaining the frame.
For an average or composition claim:Cite the source that shows subgroup values and subgroup weights. State whether the aggregate moved because group values changed, group mix changed, or both.
For disconfirming context:State the missing fact, counter-source, alternate denominator, or fairer comparison that would materially weaken the frame. If it has not been checked, keep the claim tentative.
For a timeline claim:Cite the timestamped record for each material event. State what was already underway, what mechanism is claimed, what lag is plausible, and what concurrent event could explain the same outcome.
For a survey result:Cite the method disclosure and quote the exact question only when it materially affects interpretation. Keep the claim inside the sampled population and field dates.
For a measurement readout:Cite the method note, assignment or matching rule, outcome window, exclusion rules, and uncertainty. Separate observed performance, modeled contribution, and incremental effect.
For a standards reference:Use the standard to define terms, disclosure expectations, or quality controls. Do not use it as proof that a specific report, vendor, or placement met the standard.
Measurement references
| Category | Official reference | Use when | Do not use as |
|---|---|---|---|
| MMM and causal interpretation | Google Meridian model-fit guidance | A model result needs clearer language around fit, controls, baselines, causal assumptions, and uncertainty. | Proof that any one MMM output is causally correct. |
| MMM method context | Google Meridian introduction | A reader needs an official overview of what a modern marketing mix model is built to estimate. | A substitute for the model owner's own data, code, diagnostics, or calibration evidence. |
| Outcome and data quality | MRC Outcomes and Data Quality Standards | A campaign readout depends on outcome data, matching, attribution, data controls, or quality disclosures. | A guarantee that a platform's reported outcome is incremental. |
| Attention measurement | IAB and MRC attention measurement framework | Attention is being used as an exposure-quality signal, planning input, or optimization diagnostic. | Evidence that attention alone caused sales, brand lift, or business impact. |
| Viewable impressions | MRC viewable impression guidelines | A report treats served impressions, measurable impressions, viewable impressions, and attention as interchangeable. | Proof that a viewable ad changed behavior. |
| Invalid traffic | MRC Invalid Traffic Detection and Filtration Guidelines Addendum | A report depends on filtered impressions, traffic quality, denominator changes, or post-filter outcome claims. | A guarantee that every invalid impression was removed or that remaining traffic caused outcomes. |
| Authorized sellers | IAB Tech Lab ads.txt guidance | Programmatic buyers or publishers need a public seller-authorization reference. | Evidence of campaign effectiveness or audience quality. |
| Seller transparency | IAB Tech Lab sellers.json and SupplyChain reference | A campaign or package needs seller disclosure and supply-path visibility checks before activation or renewal. | Proof that a specific impression was valuable, viewable, or incremental. |
| Ad experience | Coalition for Better Ads standards | A placement plan needs a reader-experience check across desktop, mobile web, video, or app contexts. | A replacement for editorial judgment about page usefulness. |
Survey and polling references
| Category | Official reference | Use when | Do not use as |
|---|---|---|---|
| Survey disclosure | AAPOR disclosure standards | A public poll, survey, or brand-study result needs a method checklist for sponsor, conductor, sample, wording, dates, weighting, and uncertainty. | Proof that a specific survey result is accurate or representative. |
| Transparency norms | AAPOR Transparency Initiative | A reader needs a durable reference for why public survey reports should make methods inspectable. | A replacement for inspecting the actual questionnaire, sample source, and field dates. |
| Total survey error example | Pew Research Center U.S. survey methodology | A page needs a reader-friendly example of coverage, sampling, nonresponse, measurement, processing, adjustment, and question-wording disclosure. | A general validation of unrelated polls or customer surveys. |
Media-framing references
| Category | Official reference | Use when | Do not use as |
|---|---|---|---|
| Source identification and response | Associated Press news values and principles | A critique turns on source naming, anonymous-source limits, response standards, correction practices, or separation of editorial and commercial work. | Evidence that a specific non-AP story is fair or unfair. |
| Accuracy, independence, and corrections | Reuters journalistic standards | A story needs a reference point for accuracy, independence, conflicts, corrections, and source handling. | A shortcut for assessing the actual source trail in the story being reviewed. |
| Ethics language | Society of Professional Journalists Code of Ethics | A page needs a professional journalism reference for truth seeking, transparency, accountability, and independence. | A legal rule or a substitute for specific evidence. |
| Public records | FOIA.gov request guide | A missing primary record could materially change a public claim and the relevant record is held by a federal agency. | Evidence that a record exists, will be released, or supports a claim. |
| Official data citation | U.S. Census Bureau citation guidance | A page relies on Census tables, tools, technical documents, or research and needs a replicable citation path. | A general source for claims outside the cited Census product. |
| Labor and price data | BLS data retrieval tools and BLS information guide | A claim depends on labor, price, productivity, occupational, or economic time-series data. | A substitute for citing the exact series, table, release date, access date, and definition notes. |
| Economic accounts | BEA data pages | A claim depends on national, regional, industry, international, income, spending, or investment data. | A guarantee that an article's chosen denominator or comparison is the right one. |
| Search-quality framing | Google Search Central helpful content guidance | A page needs a public reference for useful, original, reliable, people-first content expectations. | A ranking guarantee or a substitute for serving readers well. |
Citation workflow
1. Start with the original record.When a claim depends on a document, dataset, transcript, filing, public table, method note, or report, link the closest available source before linking commentary about it.
2. Separate method from interpretation.A standard can define a method or disclosure expectation. It does not prove that a particular campaign, story, or vendor result met that standard.
3. Name the denominator and date.Include the population, time period, table name, field dates, sample base, campaign window, or access date when it affects how a reader should interpret the number.
4. Match the source to the claim strength.Use official references for definitions and primary records for facts. Use secondary analysis only when it adds method, context, or a documented synthesis that the primary source does not provide.
5. Make limits visible.A reliable citation trail should also show what the source cannot answer: omitted data, alternative comparisons, possible confounders, weak response standards, or untested causal claims.
When to add a source
A source belongs in this library when it is an official standard, a primary-reference workflow, or a durable method guide that readers can use across many claims. It should clarify definitions, data access, source quality, measurement quality, citation practice, or reader experience. A source does not belong here only because it supports one article's conclusion.