SEO Case Study Verification Checklist: Test the Claim Before You Buy

01

Preconditions

Before reviewing a provider’s case studies, define the claim you need evidence for:

• Technical remediation
• Local search visibility
• Ecommerce revenue
• Editorial traffic growth
• Lead generation
• Site migration preservation
• Reputation recovery
• International expansion
• Enterprise crawl management

A case study from a different business model, starting condition, market, budget, or implementation environment may demonstrate competence without predicting your outcome.

Google’s guidance for hiring an SEO advises buyers to ask for explanations, understand proposed changes, and reject ranking guarantees. It also reminds site owners that organic ranking cannot be purchased from Google. Do You Need an SEO?^[1]

02

Checklist

Claim identity

• Copy the exact claim rather than paraphrasing its most flattering implication.
• Identify whether the claim concerns rankings, impressions, clicks, traffic, leads, revenue, conversion rate, visibility, links, or technical health.
• Separate absolute change from percentage change.
• Record whether the claim is sitewide, page-level, query-level, location-level, or campaign-level.
• Identify whether the provider claims causation, contribution, correlation, or merely chronology.

Examples:

“Organic traffic increased 300%”
“Moved 40 keywords to page one”
“Generated $2 million in SEO revenue”
“Recovered from a Google update”

Each requires different evidence.

Client and market identity

• Confirm that the client is real or understand why identity is withheld.
• Confirm the industry, geography, business model, and target audience.
• Identify whether the domain was new, established, acquired, migrated, or previously penalized.
• Record brand demand, offline advertising, public relations, seasonality, and paid media that could affect search behavior.
• Determine whether the provider worked on the entire site or one bounded section.
• Confirm that the client authorized the use of confidential or performance information.

An anonymous case can still be useful, but anonymity reduces the evidence available for independent verification.

Baseline

• Record the starting date and ending date.
• Use comparable periods when seasonality matters.
• Show the full pre-intervention trend, not one unusually weak day or week.
• Identify migrations, outages, tracking breaks, manual actions, redesigns, or previous agency changes before the baseline.
• Confirm whether the site already had strong links, brand demand, content, or rankings.
• State the starting values in addition to percentages.

A rise from 10 to 40 visits is 300% growth. A rise from 100,000 to 110,000 is 10% growth. Neither percentage describes business consequence alone.

Denominator and population

• Define what “traffic” includes: users, sessions, clicks, entrances, pageviews, or another metric.
• Define which channels are included.
• Separate brand and nonbrand search where possible.
• Identify country, device, search type, and page group.
• State how many keywords were tracked and how they were selected.
• Report the denominator for success rates.
• Identify excluded pages, queries, customers, or months.

“Forty keywords reached page one” is materially different if the provider tracked 50 terms than if it tracked 50,000.

Measurement source

• Name the measurement platform.
• Distinguish Search Console clicks from analytics sessions.
• Record attribution model and conversion definitions.
• Confirm tracking continuity across the case-study period.
• Identify canonical changes that could reassign Search Console data.
• State whether rank tracking used a fixed location, device, language, and result type.
• Preserve exports or underlying tables rather than screenshots alone.

A screenshot can be authentic and still omit the filters, comparison dates, denominator, or tracking break that changes the interpretation.

Intervention record

• List the exact actions performed.
• Separate provider work from client engineering, editorial, PR, product, and sales work.
• Identify implementation dates rather than contract dates.
• Record which recommendations were rejected or delayed.
• Identify link acquisition, paid placements, sponsorships, or affiliate relationships.
• State whether old content was removed, consolidated, redirected, or rewritten.
• Record major sitewide changes that occurred during the window.

A credible case study should make it possible to distinguish “recommended” from “implemented.”

Timing and lag

• Show the full observation window.
• Identify the first implementation date.
• Avoid attributing immediate movement to work that had not been crawled or released.
• Note algorithm updates, seasonality, promotions, news events, and competitor disruptions.
• Continue the chart long enough to show whether the result persisted.
• Report reversals and plateaus, not only the peak.

Chronology is necessary for causation but not sufficient.

Attribution

• State what evidence connects the intervention to the result.
• Compare affected pages with unaffected or later-treated pages where practical.
• Use release annotations and URL-level evidence.
• Identify alternative explanations.
• Separate direct evidence from provider inference.
• Avoid attributing all site growth to one tactic when many changes shipped together.
• State what additional evidence would change the conclusion.

The strongest case studies are often less triumphant because they reveal the parts that cannot be isolated.

Typicality and selection

• Ask how many total clients or campaigns were eligible for selection.
• Ask why this case was chosen.
• Request the median or range of comparable results where available.
• Identify failed, flat, or negative cases.
• Determine whether the result depends on an unusually large budget, existing authority, celebrity demand, or press event.
• Check whether “results may vary” is used instead of describing expected results.

FTC endorsement guidance says endorsements must be truthful and not misleading. It warns that unrepresentative testimonials may mislead unless the advertiser explains what consumers can generally expect. Advertisement Endorsements^[2]

FTC guidance also states that testimonials are not themselves substantiation for objective claims. Myths and Half-Truths About Deceptive Advertising^[3]

Material connections and disclosure

• Identify whether the quoted client received a discount, payment, free service, referral fee, equity, or other benefit.
• Identify whether the provider owns or controls the featured site.
• Disclose relationships that could affect how the endorsement is evaluated.
• Confirm that employees or contractors are not presented as independent clients.
• Identify whether a research or measurement vendor was sponsored by the provider.

The FTC’s updated endorsement guidance emphasizes disclosure of unexpected material connections and truthful representations. FTC Endorsement Guides: What People Are Asking^[4]

Ranking claims

• Record the exact query.
• Record location, device, language, date, and search engine.
• Identify whether the result is organic, local, image, shopping, video, or another search feature.
• Distinguish a manual screenshot from a controlled tracker.
• Show ranking history rather than one observation.
• Report tracked-query selection and denominator.
• Check whether brand terms are mixed with nonbrand terms.
• Reject guarantees of future ranking.

Traffic claims

• Show absolute clicks or sessions alongside percentages.
• Separate new pages from existing-page growth.
• Separate brand and nonbrand demand.
• Identify tracking changes, cookie-consent changes, or analytics migrations.
• Compare year over year when seasonality is material.
• Show landing-page and query distribution.
• Report whether traffic converted or merely expanded at low relevance.

Lead and revenue claims

• Define a lead and a qualified lead.
• Identify duplicate, spam, test, and existing-customer submissions.
• State the attribution model and lookback window.
• Separate assisted from last-click conversions.
• Identify offline close-rate verification.
• Show gross revenue versus contribution margin where relevant.
• Exclude paid-search or direct demand unless the claim includes them.
• State whether price, sales staffing, inventory, or conversion flow changed.

Technical claims

• Require affected URL examples.
• Record before-and-after status, directives, canonicals, rendering, or performance evidence.
• Distinguish crawler inventory from Google indexing outcomes.
• Identify whether the fix shipped globally or to a sample.
• Verify that the issue did not recur after later releases.
• Confirm rollback and monitoring.

Current reproducibility

• Check whether the cited pages and redirects still exist.
• Confirm that the provider still offers the described service.
• Verify that the case predates or postdates major platform changes.
• Ask whether the client retained the result after the provider left.
• Identify corrections or updates to the case study.
• Archive the reviewed version and access date.

03

Completion criteria

A case study is complete as purchasing evidence only when another reviewer can reconstruct the claim from the underlying population, dates, measurements, and implementation record. The provider should distinguish what the evidence directly shows from what it infers, explain why the featured case was selected, and disclose whether comparable clients usually achieved similar results. The buyer must also be able to inspect material connections and verify that the result persisted beyond the selected peak. Passing this standard does not make the outcome transferable; it makes the limits visible enough for a responsible decision.

A case study is decision-useful when it provides enough evidence to answer:

1. What exactly changed?
2. From what baseline and denominator?
3. Over what period?
4. For which pages, queries, markets, and users?
5. What work was actually implemented?
6. What other events could explain the result?
7. Was the result typical, exceptional, or unknown?
8. Can the evidence be reproduced or inspected?
9. Are material connections disclosed?
10. What part of the conclusion remains uncertain?

A case study can pass as evidence of a provider’s process while failing as a forecast for your business. Those are separate judgments.

04

Failure cases

Stop treating the case study as decision evidence when the baseline, denominator, measurement source, observation window, or implemented actions cannot be established. A missing field is not a harmless presentation choice when it changes the apparent size or cause of the result. Also stop when the provider offers only cropped screenshots, refuses to disclose selection criteria, combines paid and organic outcomes without attribution, or presents an exceptional testimonial as typical. The case may remain marketing material, but it cannot support a forecast or provider comparison until the missing conditions are documented and independently reviewable.

warning

Stop relying on the case study when the provider cannot identify the baseline, denominator, time window, measurement source, or implemented actions.

Additional red flags:

• Cropped screenshots without filters or dates
• Percentages without absolute values
• Rankings without query and location
• Revenue without attribution method
• Testimonials presented as typical evidence
• A case selected from an undisclosed client population
• Unexplained tracking or domain changes
• No mention of algorithm updates, seasonality, PR, or paid campaigns
• Guaranteed replication
• Client identity or data used without permission
• Material relationships hidden
• A chart ending at the highest point

References

Sources behind this record

1. Do You Need an SEO? — Google (accessed August 2, 2026)
2. Advertisement Endorsements — Federal Trade Commission (accessed August 2, 2026)
3. Myths and Half-Truths About Deceptive Advertising — Federal Trade Commission (accessed August 2, 2026)
4. FTC’s Endorsement Guides: What People Are Asking — Federal Trade Commission (accessed August 2, 2026)

Corrections

Correction history

No corrections recorded.

To report an error, use the public corrections path.