A credible case study starts with a claim system
“We transformed the client's marketing and delivered outstanding growth” sounds positive. A careful buyer still needs a starting point, defined result, measurement period and account of what else changed.
A useful client case study answers four questions early. What changed? Compared with what? Over which period? How does the business know?
The story still matters. Context helps a buyer recognise their situation and understand the work. Yet the story becomes credible only when its important statements connect to records, definitions, responsible people and agreed permission. This is the first-hand proof layer within a broader approach to citation-worthy website content.
Australian businesses also need to take care with the claims they publish. The ACCC says advertising claims should be true, accurate and based on reasonable grounds, and that the overall impression can be misleading even when individual statements are technically correct. Its guidance expressly includes websites, social media and testimonials. Read the current ACCC guidance on false or misleading claims when setting your review standard.
That makes a case study more than a copywriting exercise. It is a controlled set of commercial claims. This article offers general information for building that control. It is not legal advice.
Choose the engagement and define the proof job
Start with an engagement that has both relevance and records. A smaller project with a clean baseline, a documented intervention and specific permission may provide stronger proof than your best-known client.
Name the buyer objection the case study needs to address. A prospective client may wonder whether you understand a complex service, can improve enquiry quality, can deliver across several locations or can replace a fragile website without losing important content. Pick one primary proof job so the page does not become a catalogue of every service provided.
Then check whether the engagement can carry that job. Look for a recorded starting condition, dated work, consistent measurement and a client who can review the proposed disclosure. Set aside engagements with disputed results, unclear data ownership, unresolved confidentiality or measures that changed halfway through the comparison.
A single engagement supports claims about that engagement only. A finding intended to apply across a wider population needs an original research method with an appropriate sample and published methodology.
Build the evidence record before drafting
Create one evidence record for every material result. Do this before choosing a headline or requesting a quote. The record should let an authorised reviewer reproduce the number, understand the comparison and see the limits of the proposed wording.
| Field | What the reviewer needs to know |
|---|---|
| Context and starting condition | The relevant client situation, constraint and initial state |
| Intervention and ownership | What your team did, what the client did and what other suppliers changed |
| Metric definition | Exactly what was counted, included and excluded |
| Baseline and denominator | The starting value and, for a rate, the population it was calculated from |
| Comparison window | The dates compared and why those periods are comparable |
| Source and source owner | The system or record, plus the person accountable for its interpretation |
| Timing and contributors | When the work occurred and what else could have influenced the outcome |
| Limitations | Tracking gaps, small samples, changed definitions or unresolved uncertainty |
| Permission scope | Approved names, logo, quote, images, commercial details and channels |
| Approved wording and date | The exact claim accepted for publication and when it was approved |
| Review trigger | The change or date that requires the claim to be checked again |
Definitions prevent impressive numbers from becoming slippery. “Conversion rate” might mean form submissions divided by sessions, qualified enquiries divided by unique visitors or closed work divided by leads. Record the numerator, denominator and exclusions. If the measure is a count rather than a rate, say so.
Comparison windows need the same care. Comparing December with January may capture seasonality. Comparing the first month after launch with a twelve-month average may conceal normal variation. Use like-for-like periods where possible and explain material differences where they are not.
Keep links or references to the underlying records in the private evidence file. The published page can keep dashboards, contracts and customer data confidential while explaining the measurement well enough for a reader to understand the claim.
If a result depends on an external benchmark or platform fact, assess that borrowed evidence separately through the source evaluation workflow. First-party project records and external sources answer different questions.
Match the wording to what the result can prove
A change after your work establishes sequence. Claiming causal credit requires more because sales activity, pricing, seasonality, media spend, staffing, product availability, competitor behaviour and tracking changes may all contribute.
Off Piste uses four result descriptions to match wording to evidence strength.
- Observed change records a directly measured difference while making a limited claim about why it happened. “Qualified enquiries increased from the recorded baseline after launch” can fit here when the metric and windows are defined.
- Estimated impact uses a model, proxy or incomplete measurement. State the assumptions and use words such as estimated or modelled.
- Credible contribution connects the intervention, timing and mechanism while acknowledging other material influences. “The new enquiry journey contributed to the improvement” may be supportable when the records show that relationship.
- Causal result requires a defensible design that isolates the intervention, such as a suitable controlled experiment. Ordinary before-and-after reporting rarely meets that standard.
The two questions that matter most are whether the outcome was measured directly and how confidently the change can be attributed. Seeing them together stops a precise number from being mistaken for precise causation.
Evidence strength determines how firmly you can claim a result
Measurement directness increases
Observed change
Direct measurement with limited evidence about cause
Causal result
Direct measurement with a defensible causal design
Estimated impact
Modelled or indirect measurement with weak attribution
Credible contribution
Modelled or indirect measurement with documented timing and competing influences
Attribution confidence increases
The matrix is an editorial judgement tool, not a scoring model. An estimated impact can still help a buyer when its assumptions are visible. A directly measured change can still require cautious wording when several contributors moved at once.
Run the final result sentence back through your records. Evidence of an association supports “associated with” rather than “generated”. A measure covering organic enquiry forms supports that precise label rather than “all leads”. These boundaries make the claim more useful.
Set permission and confidentiality boundaries
Substantiation, quote accuracy, contractual confidentiality and privacy are separate checks. Client approval is important, though approval alone does not resolve every legal, contractual or privacy obligation.
The OAIC explains that publishing personal information online is a disclosure. Where consent is relied upon for a secondary use or disclosure, its APP 6 guidance says valid consent needs to be informed, voluntary, current and specific, and given by someone with capacity. Whether the Privacy Act and a particular Australian Privacy Principle apply depends on the organisation and circumstances. Seek appropriate advice for the case at hand.
Make the permission request concrete. Identify the company and people who will be named, the exact quote, metrics, screenshots, logo, links and channels. Explain whether the case study may be repurposed in proposals, social posts, presentations or advertising. Record who approved each item and the approval date.
Keep a client's words within the meaning they approved, and seek fresh approval for any substantive edit. The ACCC's review and testimonial sweep report describes concerns including fake or misleading reviews, the selective removal of negative feedback and inadequate disclosure of incentivised reviews. A case-study quote should be genuine, represented in context and traceable to the person who approved it.
Check the contract and any non-disclosure terms before sharing project material. Dashboards can reveal revenue, customer identities, staff details or commercially sensitive patterns beyond the metric you intend to show. Crop or recreate an approved extract only when its meaning stays accurate.
Structure the page around verifiable context
A credible page lets the reader follow the result without seeing the private audit file. Use headings that match the engagement rather than forcing every project into a dramatic challenge, solution and success formula.
Open with the client's relevant context and the decision they faced. Define the starting condition in observable terms. Explain the work and separate your responsibility from the client's work and other suppliers' contributions. Then present each material result with its definition, baseline, comparison window and limitations nearby.
Include the engagement or measurement dates. Name the author or responsible team. Add a client quote only when it contributes first-hand perspective that the records cannot provide on their own. Finish with a next step that fits the proof job, such as viewing the relevant service or reading the method behind a technical decision.
Google's people-first content guidance asks whether content provides original information, clear sourcing, evidence of expertise, accurate authorship and a satisfying experience. Its current guidance for generative AI features points publishers back to established SEO fundamentals and useful, unique content. A well-documented case study aligns with those quality tests. Rankings, inclusion in an AI feature and citation by an AI system still depend on factors beyond the case study.
If the CMS cannot keep evidence, authorship and dates consistent, the structured content guide explains the wider publishing model. Reusable fields for measurement period, metric definition, limitations, approval and review date can make the standard easier to maintain across many case studies.
Keep anonymised proof specific
An anonymised case study can still be specific. Keep the details that allow a buyer to interpret the work without identifying the client. Useful context may include the business category, approximate scale, market, relevant constraint, starting condition, work performed, metric definition and bounded result.
“A professional services firm with three offices” tells the reader more than “a leading client”. “Qualified website enquiries recorded in the CRM over the twelve weeks after launch” tells them more than “conversions improved”. Review combinations of details because several harmless facts can identify a client when placed together.
If you need an example but have no publishable engagement, label it clearly as constructed. Use it to demonstrate fields and reasoning, not to imply a commercial result. Never invent a client quote, metric or blended success story. A composite based on several clients also needs an explicit label and must respect each source client's permissions.
Approve the claim and preserve the audit trail
Approval works best when each reviewer has a defined job. The data owner verifies the metric and extraction. The delivery owner checks the intervention, timing and contributors. The editor tests whether the published wording stays within the evidence. The authorised client reviewer confirms factual accuracy and the agreed disclosure. A person accountable for publication resolves conflicts and gives final approval.
Preserve the evidence record, the exact approved wording, source references, permissions and approval dates together. Version the public copy when a material claim changes. This creates a route from the sentence a buyer sees back to the decision that allowed it.
Before publication, inspect every result through the website claim audit. Check that qualifications sit close to the relevant claim, not in a distant note. Confirm that a quote does not imply a result the metric record cannot support.
Keep published proof current
A case study can become inaccurate even when it was sound on publication day. Consent may change. Tracking may be rebuilt. A service may no longer include the work described. A client may merge, rebrand or ask for an identifiable detail to be removed.
Assign an internal owner and a review date. Also define event-based triggers, including a request to change permission, a revised metric definition, discovery of a tracking error, a correction from the client, a material service change or a comparison period that no longer represents current delivery.
Record corrections when they change the reader's understanding of the result. The full evidence maintenance method covers owners, source changes, correction records and refresh decisions after publication.
Turn one engagement into proof a buyer can assess
Choose one completed engagement with useful records and a clear buyer objection. Build the evidence record, agree on permission and draft the strongest claim the records support. Then ask someone outside the project to explain the metric, comparison and attribution back to you. A vague answer shows where the page needs more context.
When the evidence is sound but the claims and discovery path need shaping, SEO strategy can connect the case study to the questions qualified buyers ask. When the constraint is the page system itself, website design can provide reusable fields and templates for publishing proof consistently.