Start with the claim inventory
Before drafting, list every intended claim: revenue lift, lead quality, conversion rate, cost reduction, time saved, customer satisfaction or ranking change. For each, record the source system, metric definition, date range, comparison period, population, exclusions and person responsible for verifying it. A screenshot of a dashboard without its definitions is not sufficient evidence.
Classify claims as observed fact, calculation, estimate, client quotation or interpretation. Do not quietly promote an estimate into an observed result. If the headline says 'doubled qualified leads,' the evidence must support the same qualification definition across both periods, not just show that total form submissions increased.
Reconstruct the baseline and denominator
Recalculate every percentage from raw numerator and denominator. A move from 10 to 20 qualified leads is a 100% increase, but its uncertainty and business significance differ from a move from 1,000 to 2,000. Include absolute values where disclosure is permitted and explain whether the baseline was typical or unusually weak.
Check seasonality, promotion timing, spend changes, tracking migrations, changes in lead definition and sales capacity. If advertising spend rose 70% while leads rose 40%, a headline celebrating lead growth alone is incomplete. Report efficiency and quality alongside volume, and distinguish correlation from causation unless a credible experiment supports a causal claim.
Trace every number to an approved source
Maintain an evidence ledger with claim ID, draft wording, calculation, raw export location, extraction date, filters, reviewer and approval status. Preserve read-only exports or documented snapshots so a future editor can reproduce the figure even after dashboards change. Use access-controlled storage; case-study working files often contain customer or commercial information.
Example: a claim of '32% lower cost per accepted opportunity' should link to campaign spend and CRM accepted-opportunity counts for both periods, with exclusions documented. If a client uses a different sales-stage definition, resolve the discrepancy before writing. The publishing deadline should never outrank the evidence review.
Get customer and legal approval correctly
Agree in writing on whether the client can be named, whether logos may be displayed, which performance numbers can be public and whether quotations require named attribution. Confirm that a client's permission to tell the story is not automatically permission to disclose individual customer data or confidential commercial details.
Anonymization needs more than removing the company name. Industry, geography, unique transaction values and project timing may identify a client when combined. Use ranges or contextual descriptions when necessary, but do not distort the claim. FTC guidance on endorsements and testimonials emphasizes truthful, non-misleading representations and appropriate support for implied claims.
Write the narrative with its limitations intact
Structure the story around context, business constraint, hypothesis, intervention, implementation, measurement and results. Include what was tried and rejected where it helps readers understand the decision. A credible case study can say 'we observed an improvement after launch, but did not run a randomized holdout' without undermining its usefulness.
Distinguish marketing outcomes from business outcomes. Higher click-through rate is not equivalent to greater profit. Explain how leads were accepted, how returns were treated or why a particular proxy was used. When results are exceptional, avoid implying every future client should expect the same magnitude of improvement.
Create a publication and refresh gate
Require a subject-matter reviewer, a data owner and a permissions owner to approve the final version. Verify every displayed number, quote, screenshot, source and internal link after layout. Record the approval date and any conditions, such as an expiry date for use of the client's logo or permission to reference a campaign.
Review published case studies when attribution methods, metrics or customer circumstances change. Link to the site's case-study production system and editorial evidence traceability guide. Use the SERP Preview tool to craft a specific, defensible search snippet rather than promising results the evidence cannot substantiate.