A supported marketing claim says no more than the underlying record can establish for the reader who encounters it. Check the likely meaning, the measured outcome, the timeframe, the relevant conditions, and the team's contribution before choosing the strongest accurate wording.

In his guidance for this article, Mike Mangione, founder of The Mangione Group, recommends:

I’d narrow or remove claims built on incomplete records, misleading comparisons, unsupported causation, or guarantees the business cannot honor. An AI-generated summary isn’t a substitute for checking the source.

Mike Mangione, founder of The Mangione Group, recommends judging a claim by what a reasonable reader will believe it means, then comparing that interpretation with the evidence. His submitted guidance calls for narrowing or removing claims built on incomplete records, misleading comparisons, unsupported causation, or commitments the business cannot honor.

Key Facts at a Glance

  • A record of one project supports a statement about that project; it does not automatically establish a promise for every customer.
  • NIST distinguishes association from causation. A relationship in a chart does not prove cause and effect. NIST scatter-plot guidance.
  • Google Analytics defines attribution as assigning credit for important actions along a path. Attribution documentation.
  • Mean and median describe a dataset differently and can diverge in a skewed distribution. NIST measures of location.
  • A forecast should remain visibly different from an achieved result.
  • Team contribution, project outcome, and individual responsibility are separate statements that need separate support.
  • A headline, chart, caption, and qualification must communicate a compatible meaning.

Mike Mangione brings 18 years of marketing expertise to his role as founder of The Mangione Group. For this article, his contribution is a specific editorial recommendation: make the strength of a public claim depend on the strength and scope of the record behind it.

How do you determine what a reader will think a claim means?

Read the claim as a prospective buyer would, without relying on background knowledge held by the writer. Identify who appears to have achieved the result, which outcome is promised or reported, and how broadly the statement seems to apply. Then examine the surrounding headline, imagery, and context for meanings that extend beyond the sentence itself.

A claim is more than a number. “We transformed performance” implies an outcome and a contribution even though it contains no measurement. “Our process delivers predictable growth” implies repeatability and future performance. “A customer reported a smoother handoff” describes a particular account, provided the statement is genuine and permission is established. Each sentence asks the reader to believe something different.

Begin a review by writing that implied belief in ordinary language. If the headline says the team generated a result, ask whether the supporting record shows the team's work, the customer's outcome, and a credible connection between them. If it says customers typically achieve an outcome, ask which customers are represented and what “typically” means. Do not let an attractive phrase bypass a question you would immediately ask about a precise statistic.

Mike's recommendation puts the reader's interpretation ahead of the writer's private intention. A writer may intend a historical example while a page presents it as a current promise. The difference matters because the buyer makes a decision from the public page, not the internal briefing. A digital marketing agency should therefore review the whole message rather than approve each sentence in isolation.

Google's helpful-content guidance asks whether titles exaggerate and whether readers can see clear sourcing and evidence of expertise. Those questions support a sensible editorial review. They do not turn a marketing statement into a verified fact, and they do not establish that a more restrained claim will receive a particular ranking.

When two reviewers interpret the wording differently, investigate the disagreement rather than averaging their opinions. Ask a suitable reader what they expect after seeing the page. Keep the question open enough that the answer is not supplied by the interviewer. The purpose is to identify a plausible interpretation the evidence must support, or to change the wording so the unsupported interpretation is less likely.

What should an evidence record contain before a claim is published?

An evidence record should identify the proposed wording, its likely meaning, the original source, the measured outcome, the period, the relevant conditions, and the people responsible for the work. It should also record limitations and permission to publish. A polished summary is useful for navigation, but it cannot replace access to the underlying material.

Use one record per materially different claim. A case study may contain several: a customer had a particular problem, the team performed specific work, a measured outcome changed, and the work contributed to that change. Supporting one does not establish the others. A project brief may verify the planned scope without proving the work was completed. A completed deliverable may prove the work existed without establishing a commercial outcome.

Retain the source at a level that lets another reviewer inspect it. Record the report name, date, relevant section or cells, definitions, filters, and who can answer questions. A screenshot is sometimes useful for context, but it may omit the denominator or settings needed to understand the figure. A copied number without its definition becomes difficult to audit as soon as the original analyst is unavailable.

Separate the record's state from the copy's state. “Drafted” does not mean verified. “Verified internally” does not mean cleared for public use. “Approved wording” should refer to the precise wording reviewed, not a stronger rewrite created later. This distinction is especially valuable when a claim moves from a detailed case study into a short advertisement or homepage headline.

GOV.UK's research-analysis guidance separates observations, findings, and actions, and recommends checking recordings when confirming observations. The same discipline can inform a claim file: retain what the record says, what you infer, and what you propose to publish as different fields. This is an editorial application of the guidance, not an assertion that customer research follows the same rules as every performance measurement.

Write down missing information directly. If the timeframe is unknown, say so internally and hold the time-dependent claim. If attribution is unresolved, describe the work without claiming the outcome as exclusively yours. Missing records are a reason to narrow the public statement or continue research. They are not a reason to add a broad qualifier that leaves the headline's unsupported meaning intact.

How do you check whether the measured outcome matches the words?

Match each important noun in the claim to an explicit measurement definition. Visits, inquiries, qualified opportunities, customers, revenue, and profit describe different things. Confirm what was counted, what was excluded, and whether the definition stayed consistent across the period. If the record measures an earlier step, do not rename it as a later business outcome.

Start with the unit. Does a row represent a person, a company, a form submission, a meeting, or a transaction? One person can produce several records. A count of submissions cannot automatically become a count of people. A booked meeting is not an attended meeting, and an attended meeting is not a customer. The language should preserve those differences even when a more impressive label would make a stronger headline.

Next, examine the denominator. “Half of inquiries were suitable” requires a definition of suitability and the relevant inquiry population. Were duplicate submissions excluded? Were unreviewed inquiries counted as unsuitable, excluded, or left unknown? Different choices can produce different percentages from the same raw activity. State the decision in the method rather than quietly using whichever interpretation yields the larger figure.

Product labels deserve the same attention. Google Analytics defines user engagement in terms of time with a web page in focus or an app screen in the foreground. That definition is useful for interpreting the report, but it does not establish that a person understood the offer or became interested in buying. A claim about buyer understanding needs evidence about understanding.

Google's default-channel documentation describes channels as rule-based definitions of traffic sources. Therefore a channel label is a classification under a system's rules. Before writing that a particular campaign produced an outcome, review the report's scope and classification rather than assuming the label explains the entire buying history.

A hypothetical example makes the distinction concrete. Suppose a report contains twenty completed forms, five repeat submissions, and three requests outside the service area. The record may support twenty submissions. It does not automatically support twenty distinct suitable prospects. This example is arithmetic for editorial review, not a client result or a recommended reporting benchmark. The important work is deciding the unit and exclusions before drafting the headline.

Our guide to testing offers and calls to action considers the related problem of evaluating outcomes beyond clicks. Claim review begins where measurement language becomes public language: preserve the meaning of the measure when you explain the result to a buyer.

Two people compare a document with open books at a wooden table.

How can a comparison be accurate and still create the wrong impression?

A comparison can use correct numbers while combining unlike periods, populations, definitions, or conditions. Check the baseline, the endpoint, the amount of exposure, and any changes in measurement. Explain whether the difference is absolute or relative, and avoid choosing a starting point solely because it makes the later result appear unusually strong.

Consider an illustrative change from two suitable inquiries out of one hundred eligible visitors to four out of one hundred. The rate moves from two percent to four percent. That is an increase of two percentage points and a relative increase of one hundred percent. Those expressions describe the same arithmetic but create different impressions when the starting values are omitted. Neither establishes why the change occurred.

Comparisons across time require another question: what else changed? A longer reporting window, a different audience, a seasonal peak, or a revised qualification rule can affect the result. You may still report the observed difference, but the wording should explain the relevant conditions. If the records cannot be reconciled, the honest action may be to report each period separately rather than force them into a growth claim.

For averages, inspect the distribution rather than treating “average” as a neutral shortcut. NIST's measures-of-location guidance defines mean and median and explains why skewed data can produce different values. If a few unusually large outcomes dominate a mean, a buyer may misunderstand that number as the experience of an ordinary engagement. Report the measure accurately and consider whether additional context is needed.

NIST separately describes skewness as a measure of asymmetry. You do not need a statistical term in every case study, but you do need to notice when the evidence is uneven. A range, distribution, or carefully scoped account may be more informative than a single impressive central value. The chosen summary should help the reader understand the record.

Check exclusions before publication. If incomplete projects are omitted, say what qualified for inclusion and why. If different customers had different follow-up periods, do not imply equal opportunity to achieve the outcome. A comparison is a designed explanation of evidence. Its selection rules belong to the meaning of the claim, not merely to a private spreadsheet the reader will never see.

How do you describe the team's contribution without claiming unsupported causation?

Describe what the team demonstrably did, then distinguish that work from the outcome observed during the engagement. Use causal wording only when the evidence can support the connection at the stated scope. When several contributors or changes affected the result, preserve that context instead of assigning the entire outcome to the most visible marketing intervention.

A project record can establish that the team researched an audience, wrote a landing page, or implemented a follow-up process. A separate outcome record may show more qualified conversations during the same period. Putting those facts next to one another is useful, but the connection still needs examination. Sales staffing, pricing, availability, customer demand, and other campaigns may also have changed.

NIST's scatter-plot guidance explicitly distinguishes association from causation. The editorial implication is straightforward: a chart showing two measures moving together does not, by itself, prove one caused the other. A case study can honestly describe the sequence and the team's interpretation while retaining the limitations of the record.

Google Analytics attribution documentation defines attribution as assigning credit and explains that models can use rules or data-driven algorithms. Attribution in a report is therefore not interchangeable with a controlled demonstration that an activity created an incremental outcome. When a claim relies on attributed results, describe the measure as attributed and retain the relevant model context.

If the business wants a stronger causal conclusion, plan measurement around that question. NIST's experimental-design introduction describes deliberately changing factors and planning analysis in advance. Its design-selection guidance identifies different designs for different objectives. These resources establish why an appropriate comparison should be planned; they do not imply that every small campaign has enough data for a reliable causal estimate.

Attribution to people requires equal care. Work completed by members of the assembled team in previous roles should not become a claim that the present firm alone produced every historical result. Mike's submitted guidance specifically calls for accurate attribution of team work. State the role, period, and contribution that the records support. Avoid changing “members of our team contributed” into “I delivered” merely because a first-person story sounds more direct.

The strongest accurate account may be modest in grammar and substantial in detail: what was changed, what was observed, which contributors mattered, and what cannot be isolated. A buyer can use that account to assess relevance. Unsupported certainty removes information the buyer needs to make that judgment.

How strong can the wording be when the evidence is limited?

Choose wording that matches the narrowest material limit in the evidence: the project, population, period, outcome definition, or contribution. A strong record about one engagement can support a detailed account of that engagement. Broader claims require broader support, and uncertainty should change the main statement rather than appear only in a distant footnote.

Claim-strength review: illustrative wording, not company results
Available recordPotentially supportable wordingUnjustified extension
Verified completed deliverableThe team produced the documented deliverableThe deliverable increased sales
One verified project outcomeThis project recorded the defined outcome during the stated periodCustomers can expect the same result
Customer's approved accountThe customer described the specified experienceThe experience is independently measured or universal
Clearly defined multi-project recordThe defined set recorded the stated distributionEvery future engagement will achieve the average
Scenario built from assumptionsUnder these illustrative assumptions, the calculation produces this resultThe business has achieved or will achieve that result

The table is a review aid, not an automatic approval system. Even the narrower sentence needs an authentic record, a correct calculation where relevant, and permission to disclose it. The point is to make the unsupported extension visible. A verified process statement does not become a verified performance statement because both appear on the same page.

Be careful with words that silently broaden scope: consistently, typically, predictable, proven, always, and guaranteed. They are not interchangeable stylistic flourishes. Each asks for additional evidence or an actual commitment the business can honor. If the record supports a limited observation, a detailed explanation of that observation is usually more useful than a broad adjective followed by several qualifications.

Uncertainty also belongs in estimates. NIST's confidence-limits guidance explains why an interval can communicate uncertainty around an estimated mean. The broader editorial lesson is to avoid presenting an estimate as if it were exact. Do not add a confidence interval without a valid method, or assume a statistical interval resolves missing records, biased selection, or a poorly defined outcome.

Scope can make a claim more useful, not merely safer. A buyer considering a similar situation may care more about the conditions of a well-described project than an unbounded success statement. Relevant detail allows comparison. It also helps unsuitable buyers recognize that a result from another setting may not transfer to their own.

A dated example comes from the UK Committee of Advertising Practice. Its October 22, 2026 guidance on plug-in solar advertising explains that savings claims need evidence matching the benefit consumers are likely to achieve, while price comparisons need clear assumptions and a clear basis. This is UK guidance about a particular product category, not a statement of US requirements. It reinforces the editorial question here: does the evidence support the promise a reader will actually take away? CAP guidance on plug-in solar claims

A hand holds a pencil above paper beside a magnifying glass and photographs.

How should forecasts and hypothetical examples be labeled?

Label forecasts as estimates and hypothetical examples as illustrations before the reader encounters the result. Show the inputs, assumptions, calculation, and important limits. Keep those materials visually and verbally separate from achieved outcomes, so a scenario used to explain a decision cannot be mistaken for a result the business has already produced.

A scenario answers what follows if certain inputs are assumed. A forecast makes an estimate about what may happen, usually using information and a method. An achieved result describes something that has occurred and been recorded. Mixing those categories can happen through small wording changes. “Would produce” becomes “produces,” or an illustrative example loses its label when copied into a presentation.

For a hypothetical calculation, suppose a campaign receives one thousand eligible visits and five percent become inquiries. The arithmetic produces fifty inquiries under those assumptions. It does not establish that five percent is achievable, typical, or appropriate for a particular business. It also says nothing about suitability, meetings, customers, revenue, or delivery capacity unless those additional steps are explicitly modeled.

Make the assumptions inspectable near the output. If a scenario depends on customer behavior, capacity, or a future implementation, describe that dependency in ordinary language. Avoid introducing unjustified precision through a calculator that produces several decimal places. The format should reflect the uncertainty in the inputs rather than making an estimate appear more certain because it comes from software.

AI-generated summaries need the same source discipline. Mike's guidance recommends checking the underlying record rather than treating a generated summary as evidence. A summary can help locate a passage or organize a review, but the writer must confirm the factual meaning against the source. If the source cannot be opened or the number cannot be located, the claim remains unverified.

When comparing a forecast with a later result, preserve the forecast as it existed at the time. Replacing assumptions after the outcome is known obscures what the forecast actually predicted. Record the difference, investigate the causes, and describe the learning without pretending the original estimate was more accurate than it was. That history can be useful evidence about the planning process, provided its limitations remain clear.

How should claims, charts, and qualifications work together on the page?

Present the central claim and its material conditions close enough that readers can understand them together. Label chart units, periods, and populations, and make references accessible. Review the page at realistic sizes, because an accurate qualification can lose its practical meaning if it is separated from the promise or hidden in an unreadable presentation.

Begin with a plain-language claim above the supporting detail. Name the outcome and scope before adding decorative language. A case title that says “A documented change in qualified inquiries during a defined campaign” may need refinement for readability, but its facts should survive the refinement. The final headline should not remove the distinction between a recorded observation and a promised result.

Charts should identify what is being compared. Include units, time periods, and any material exclusions. If one bar represents a shorter period, a different audience, or a changed definition, the comparison needs explanation or redesign. Do not use a graphic to suggest a like-for-like comparison the data cannot support. A clear table may be the better format when exact definitions matter more than visual impact.

W3C's table guidance explains how captions help readers identify a table and understand its subject. Use meaningful headers and a caption that describes the actual evidence. A table labeled “Results” leaves too much work to the reader if it mixes forecasts, historical observations, and customer impressions.

W3C's link-purpose guidance also supports descriptive source links. “Read the measurement method” is more useful than an unexplained symbol when the reader needs to inspect how a number was defined. Keep evidence links recognizable and usable with keyboard navigation. A source should function as an opportunity to verify the statement, not a visual signal of authority alone.

Before a claim moves into a chart or headline, record who owns that placement and what would require it to change. For a hypothetical delivery-time claim, a revised service scope or customer input requirement should trigger review. For a campaign result, a corrected report or changed attribution finding should do the same. Keep the approved wording, source version, publication locations, and review trigger together. A historical observation may remain accurate while its use as a description of today's service needs to expire.

Our landing-page copywriting guide addresses how a campaign promise connects with the page's next step. Check that connection during claim review. A careful explanation on the destination page cannot fully clarify an incompatible promise in the message that brought the buyer there.

How can you review a claim before handing it to a writer?

Use the worksheet to identify missing support before polishing the sentence. Record the proposed claim and check the meaning, source, measure, period, contribution, limits, and publication permission. The output lists unresolved review tasks. Completing the checklist is evidence of a review process, not certification that a claim is true or appropriate for every placement.

Use nonconfidential wording. This worksheet does not send or save entries.

Which checks have supporting records?

Enter a claim and check only items supported by records.

Imagine a hypothetical draft stating that a new page doubled sales. The source verifies an increase in inquiries, but sales records are unavailable. The outcome-definition check cannot be treated as complete for the draft's present wording. The next action is to obtain sales evidence or revise the sentence to the measured inquiry outcome, while separately reviewing whether the page caused the change.

In another hypothetical review, the outcome is verified but members of the team contributed while working alongside other organizations. The contribution check should capture that history. The claim may remain valuable when accurately attributed. The worksheet helps prevent the act of shortening a sentence from quietly changing who appears responsible for the result.

A digital marketing company using this tool should attach the relevant records in its own approved system and name a reviewer for unresolved items. The page does not retain your entries. Copy the review decisions into your working document if you need them later. Never interpret the count of checked boxes as a probability that the claim is accurate.

Two people discuss a sheet of paper at a wooden desk.

How do you keep approved claims accurate as the business changes?

Give each approved claim an owner, a source record, a version, and a reason to review it again. Recheck claims when the service, measurement method, supporting evidence, or attribution changes. Track where the wording appears so a correction reaches the website, advertisements, presentations, and other materials using the same statement.

Approval should attach to wording and context. A detailed case-study paragraph may be accurate because its dates and conditions are visible. A shorter version used in a headline may omit those facts and imply a broader result. Treat that new placement as a new review, even if most of the words came from an approved source.

Maintain a simple register of active statements and their destinations. Include the canonical wording, permitted variations if any, relevant source, reviewer, and review trigger. If the supporting record changes, the register makes it possible to locate affected materials. Without that map, a corrected website can coexist with an old sales presentation carrying the same unsupported promise.

Historical results do not automatically become false with age, but their relevance can change. A record from an earlier service model may remain an accurate historical account while offering limited guidance about the current offer. Retain the date and conditions. Do not refresh a timestamp in a way that makes an old observation appear newly measured.

When a correction is needed, fix the claim at its source and inspect derivative versions. Record what changed and why so a later writer does not restore the old wording from an archived brochure. If the issue is only an unclear definition, clarify it. If the underlying outcome cannot be verified, narrow or remove the statement. The scale of the correction should match the evidence problem.

The website copywriting guide treats claims as part of helping buyers decide. Maintaining them is therefore a continuing communication responsibility. A claim file is useful only if the public words still match its contents and the business still knows who can explain the record behind them.

What else should you check when reviewing marketing claims?

Claim review often turns on practical distinctions that a headline cannot show by itself. The following questions cover limited evidence, customer statements, unnamed projects, AI summaries, historical team work, and comparative language. In each case, the aim is to preserve what the record establishes while avoiding an unsupported extension of its meaning.

Can we publish a case study without a numerical result?

Yes, a case study can explain a verified problem, the work performed, and a documented qualitative outcome. Keep those statements specific and accurately attributed. Do not invent a number to make the story appear stronger, or use a customer's impression as if it were an independently measured performance result.

Does a customer testimonial prove a performance claim?

A genuine, approved testimonial can document what that customer said about their experience. It does not automatically establish a measured result, explain causation, or support a promise for other buyers. Verify the statement, its context, and permission, then make sure surrounding copy does not give the testimony a broader meaning.

Can we anonymize a project and still make a useful claim?

Potentially, if public use is cleared and removing identifying details does not erase essential context. Preserve the outcome definition, period, conditions, and contribution at an appropriate level. Anonymity does not solve missing evidence or permission. If the remaining detail could still identify a restricted project, the material needs further review.

Can AI check whether the evidence supports the wording?

AI can help organize a claim list, locate possible inconsistencies, or draft questions for a reviewer. The underlying records still need human verification, especially where definitions, context, permission, or attribution are ambiguous. A confident generated assessment cannot supply a missing source or establish that a result occurred as described.

How should we describe work completed before the current firm existed?

Describe the relevant team members' contribution and historical context accurately, using only cleared information. Avoid wording that makes the present firm appear to have completed work it did not perform. A useful account can preserve prior experience without turning a multi-contributor history into a single company or personal achievement.

What evidence supports saying our service is better?

First define better for the buyer's decision: scope, suitability, responsiveness, a specific feature, or a measured outcome. Then verify the comparison under comparable conditions and at a stated date. If the evidence supports only a difference, describe that difference. A preference or broad aspiration should not be presented as an established comparative fact.

How can The Mangione Group help make your claims clearer?

The Mangione Group can help turn verified capabilities and documented experience into copy buyers can understand and judge. The work involves more than editing adjectives. Claims must fit the offer, preserve the meaning of the evidence, and remain consistent across the website, campaigns, and sales materials that introduce the business.

Our marketing copywriting service includes websites, landing pages, emails, ads, and sales materials. For a claim review, the useful starting point is the current wording and the records behind it. A marketing company should make clear which statements can be strengthened through better explanation, which need narrower wording, and which need additional evidence before public use.

If nobody can locate the underlying records or confirm what the service currently delivers, organize those facts before commissioning a broad rewrite. Outside writing cannot replace missing evidence. If the records exist but the public story overreaches, undersells useful detail, or leaves attribution unclear, a 30-minute Marketing Review can help identify the first practical change. Bring the page or campaign, its strongest claims, and the supporting material you are able to share.

Book My Marketing Review

Michael Mangione

Michael Mangione, founder of The Mangione Group.

Founder, The Mangione Group

Michael Mangione brings 18 years of marketing expertise. He's assembled a leadership team whose track record includes more than $400 million in advertising spend managed, over 600,000 leads generated, and more than $1 billion in client outcomes.

He built The Mangione Group to help businesses compete with greater precision, win more customers, and grow revenue. His approach brings innovative technology and AI together with sharp strategy, persuasive copywriting, and compelling design to make businesses stand out, make their value unmistakable, and give customers a reason to choose them.

About Mike Mangione | Mike Mangione on LinkedIn

References and Sources

Primary documents checked October 25, 2026. Statistical and product documentation supports the specific definitions cited; the review workflow is editorial analysis.

  1. Scatter Plot. NIST/SEMATECH e-Handbook. Checked October 25, 2026.
  2. Get started with attribution. Google Analytics Help. Checked October 25, 2026.
  3. Measures of Location. NIST/SEMATECH e-Handbook. Checked October 25, 2026.
  4. Creating helpful, reliable, people-first content. Google Search Central. Checked October 25, 2026.
  5. Analyse a research session. GOV.UK. Checked October 25, 2026.
  6. User engagement. Google Analytics Help. Checked October 25, 2026.
  7. Default channel group. Google Analytics Help. Checked October 25, 2026.
  8. Measures of Skewness and Kurtosis. NIST/SEMATECH e-Handbook. Checked October 25, 2026.
  9. What is experimental design?. NIST/SEMATECH e-Handbook. Checked October 25, 2026.
  10. Choosing an experimental design. NIST/SEMATECH e-Handbook. Checked October 25, 2026.
  11. Confidence Limits for the Mean. NIST/SEMATECH e-Handbook. Checked October 25, 2026.
  12. Caption & Summary. W3C WAI. Checked October 25, 2026.
  13. Understanding SC 2.4.4: Link Purpose (In Context). W3C WAI. Checked October 25, 2026.
  14. Plug-in solar: new product, familiar advertising rules. ASA/CAP. Published October 22, 2026. Checked October 26, 2026.