Marketing data claims, testimonials and case studies without overstating results

That "94% customer satisfactioncustomer satisfactionCustomer Satisfaction Score, a direct measure of satisfaction captured right after a specific interaction or experience, usually on a short rating scale.View full definition →" figure on the pricing page came from somewhere else: a survey of 18 accounts, all hand-picked from the top tier, fielded 14 months ago by a CS manager who has since left. Marketing wrote the sentence. Marketing did not generate the evidence, cannot reconstruct the sample, and is the party that answers for it. Most claims trouble in SaaS starts exactly here, with second-hand evidence nobody re-checked before it went into a paid campaign.

Testimonials, review scores, ROIROIReturn on Investment: the ratio of net profit to the cost of an investment. A 300% ROI means each dollar invested returns $3.View full definition → statistics and customer case studies carry disclosure duties of their own, sitting on top of the reasonable-basis standard the advertising-claims lesson sets out. This lesson covers those duties and the vetting that makes borrowed evidence survivable.

What the endorsement rules add

The FTC Endorsement Guides, revised in June 2023, govern anything presented as another person's experience of your product. Two obligations sit above ordinary truthfulness.

The endorsement has to reflect the endorser's honest opinion or actual experience. Ghost-writing a quote and getting a signature on it fails that test if the customer would not have said it unprompted.

Then typicality. If the featured result is not what a typical customer gets, you disclose the generally expected performance. The fix most teams reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → for stopped working in 2009, when the FTC withdrew the safe harbour for "results not typical" disclaimers. Saying results vary is no longer sufficient. If your median account cuts onboarding time by 20% and the case study headline says 70%, the 20% has to appear near the 70%. The FTC's endorsement Q&A works through the edge cases in plain language.

The 2024 rule on fake and deceptive reviews goes harder than the Guides: it carries per-violation civil penalties (the statutory maximum sits above $50,000 and is adjusted for inflation each year), which the Guides alone do not.

Where the evidence breaks

1. The case study built on your best account

"How [Customer] cut onboarding time by 70%" converts well. If that account is the single best outcome from 500 customers, the ad's implied claim is that a buyer might reasonably expect something like it. Add the cohort average next to the headline number, and record which cohort you mean: "average reduction across the 2025 cohort was 22%" is defensible, "results vary by customer" is not.

2. ROI numbers pulled from your own analytics

Your own telemetry feels safer than a customer's word. It usually is not. Segment (a customer data platformcustomer data platformSoftware that unifies customer data from every source into one persistent profile that marketing, sales and service teams can act on.View full definition →, so a vendor of the very instrumentation under discussion) exists because event definitions drift: an onboarding_completed event gets redefined twice in a year, and the 40% improvement you measured in Q1 is not measuring the same thing in Q4. When a regulator or competitor asks how the number was computed, "our dashboard said so" is not an answer.

The quieter problem is survivorship. An ROI figure computed across accounts still active today excludes everyone who churned. If a third of the cohort left before renewal, "3x ROI in 90 days" describes a population partly defined by having got value in the first place. Recompute on the full entry cohort, including churned accounts, before the number goes into an ad. The same discipline applies to any payback figure borrowed from the LTVLTVLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → work the modelling lesson handles.

3. Review platforms and gating

In January 2022 the FTC settled with Fashion Nova for $4.2 million over review suppression: the retailer used a review management tool that held back submissions below four stars, so its product pages showed a filtered picture. Not a SaaS company, but the mechanism is the one B2B teams use every week, dressed differently: the "would you leave us a review?" email that fires only for NPSNPSNet Promoter Score (NPS) measures customer loyalty by asking how likely customers are to recommend a brand, then subtracting detractors from promoters.View full definition → promoters. That is review-gating, and the 2024 rule names it.

Trustpilot, which sells review collection software and so has a commercial stake in the pipepipeAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.View full definition → staying credible, publishes annual transparency reporting showing it removes millions of fake or non-genuine reviews a year, a meaningful percentage of everything submitted. Assume the platforms are looking. Incentives are allowed on most of them, including gift cards; the incentive has to be disclosed on the review itself, and it cannot be conditional on sentiment.

4. Material connections nobody disclosed

A discount, a free tier upgrade, a co-marketing fee or a services credit given in exchange for a case study is a material connection, and it goes near the claim rather than in a footer.

Two versions get missed. First, the connection created afterwards: the customer gave the quote for nothing, then received a $500 credit as thanks. If a reader would weigh the endorsement differently knowing that, it is disclosable. Second, the reciprocal one, where you are a paying customer of the company endorsing you. That relationship almost never appears in the case study and is exactly what a competitor's counsel will find.

Vetting evidence before it ships

Before a testimonial, statistic or logo enters a campaign asset:

1. Signed permission on file (logo release, quote approval), with an expiry date and a named approver at the customer.
2. Representativeness: what does the median account get, and is that figure sitting next to the featured one?
3. Recency: under 12 to 18 months, or visibly dated.
4. Material connection: anything of value received, in either direction, disclosed near the claim.
5. Reproducibility: can someone re-run the query or produce the interview notes six months from now, after the analytics schemaschemaA schema is the formal blueprint that defines how data is structured, named, typed, and related within a database, file, or message.View full definition → changed?
6. Badges: earned under the review platform's own terms, not bought outside an approved collection programme.

A claim register keeps this usable:

claim_id | claim_text | source_type | sample_size | date_collected | disclosure_required | disclosure_text | approved_by | permission_expires

This is the artefact the launch sign-off sequence draws on. Without it, legal review turns into archaeology two days before a campaign date.

Who actually comes after you

The FTC does not audit B2B SaaS advertising at volume. The realistic threats are closer in.

A competitor can sue under the Lanham Act for false advertising, with damages and injunctions attached, and they do not need to prove you intended to mislead. Faster and cheaper for them: a challenge at the National Advertising Division of BBB National Programs, which resolves in months, publishes its decisions, and refers non-participants to the FTC. Enterprise procurement is the third route, where a prospect's legal team asks for the substantiation behind a number on your own site and gets nothing.

Then the failure mode that costs nothing to avoid and keeps recurring: the case study that outlives the relationship. The customer churns, gets acquired, or the champion leaves, and the page stays live for two more years with a logo you no longer have rights to and a result nobody can defend. Public-company customers add another layer, since their comms team may never have cleared a metric that touches reported figures. Put an expiry date on every permission and review the library quarterly.

Sector sweeps are the exception to all of the above. The FTC's Operation AI Comply, announced in September 2024, targets inflated AI performance claims specifically. "Our AI cuts support tickets by half" needs your own current data, not a slide from the model vendor.

For a plain-English primer on what counts as a defensible claim, the FTC's business guidance hub is free, regularly updated and better than secondhand legal blog summaries.

🎬 [VIDEO: "FTC Endorsement Guides Explained" - youtube.com/@FTC - official FTC channel content walking through testimonial and endorsement disclosure requirements in plain language]

Europe: same spirit, sharper teeth

The UK stopped being the soft option in 2025. The Digital Markets, Competition and Consumers Act 2024 lets the Competition and Markets Authority decide consumer law breaches and fine directly, up to 10% of global turnover, without going to court first. Fake and incentivised reviews are named in the banned practices, and the duty extends to businesses that commission or host them without taking reasonable steps to check. The ASA route stays fast and public: a complaint about a UK ROI claim can be ruled on in weeks, and the ruling is indexed and searchable by your prospects. In the EU, the Unfair Commercial Practices Directive holds the same line. Substantiate before publication, not after the complaint.

Key Takeaways

• Evidence you did not generate is still yours to defend: sample, date, method and permission, on file before the claim runs.
• Typicality disclosure means publishing the generally expected result, not the words "results vary"; that disclaimer lost its safe harbour in 2009.
• Recompute ROI figures on the full entry cohort. Numbers drawn only from surviving accounts overstate by construction.
• Review-gating is the common SaaS version of what cost Fashion Nova $4.2 million, and the 2024 FTC rule attaches civil penalties to it.
• Competitors and NAD move faster than the FTC, and the CMA can now fine up to 10% of global turnover without a court.