What a case study should establish
A case study that only says the engagement went well proves nothing. Here, I show the initial problem, identify the signals corrected, give the observation period and specify the exact nature of the result.
In Entity SEO, evidence takes a specific form: an AI citation, a Knowledge Graph correction, restored Schema.org consistency or increased visibility across a defined set of queries. Vague claims do not belong on this page.
Three levels remain separate: what was delivered, what was measured and what remains outside our control. This is how results can be discussed without selling an impossible guarantee. A method that promises everything proves nothing.
Case 01 · Local AI search
Before. A local restaurant business ranked on Google's first page, with zero AI citations for the same local queries. Initial Entity Confidence score: 38/100. The problem was consistency: Google Business Profile, the website, social profiles, structured data and evidence pages each described a slightly different version of the same entity.
Structure. Rewrote LocalBusiness schema, previously missing from two-thirds of pages. Rebuilt heading hierarchies to place answers before supporting arguments.
Flow. Strictly aligned the name, address, opening hours and links across Google Business Profile, the website and social profiles. Connected evidence pages and product pages through internal links where none existed before.
Trust. Structured local answers as Q&A, supported by actual visitor figures and verified reviews. Aligned visible sources so an AI would not have to choose between contradictory versions of the same information.
Measured result. Entity Confidence score: 38/100 to 79/100 in 74 days. Entity Confidence is a secondary summary of the SFT profile. It must not be interpreted independently of the Structure, Flow and Trust scores. First confirmed Perplexity citation for a local industry query on day 74. Zero AI citations before; one verified citation afterwards, for a query selected for direct commercial relevance rather than ease.
Client quote, translated: “We knew we existed on Google. We didn't know we didn't exist elsewhere.”
Case 02 · A B2B SaaS struggling to explain its offer
Before. A B2B SaaS with 23 feature pages and no single page explaining who did what, for whom and with what evidence. The website described features without establishing a recognizable entity. Product-page bounce rate: 71%.
Structure. Mapped product entities in full. Created four pillar pages to replace fragmentation across 23 pages. Organized each page around an answer instead of a feature list.
Flow. Strengthened connections between resources, use cases and commercial pages. Replaced generic “learn more” links with meaningful anchor text.
Trust. Added visible FAQs to every pillar page. Standardized Organization/WebSite/WebPage schema. Aligned metadata across the website, LinkedIn and industry directory product profiles.
Measured result. Product-page bounce rate: 71% to 44% in 90 days. Average time spent on pillar pages increased 2.3 times. The priority was reducing ambiguity rather than adding content volume.
Client quote, translated: “We removed more pages than we created. And for the first time, a prospect told us they understood the offer in a single read.”
Case 03 · Personal brand and Knowledge Panel
Before. No verifiable entity. A scattered presence across LinkedIn, a website and several unconnected press mentions. No signal let a machine confidently establish who stood behind the name. The risk was indefinite dependence on one social profile without a central source of authority.
Structure. Created and completed the Wikidata entity. Structured the author profile separately from the marketing biography.
Flow. Consolidated sameAs links between Wikidata, LinkedIn, the website, social profiles and publications. Connected the author, books, method and tool, previously unrelated in structured data.
Trust. Linked Person and Organization JSON-LD through a unique @id. Strictly aligned biographical facts across every platform.
Measured result. No entity recognition before. Verified Wikidata entity afterwards, reference Q127330925. A Google Knowledge Panel obtained without a Wikipedia page in just over four months. This observation alone does not establish what caused the panel to appear. A Knowledge Panel obtained without Wikipedia is an observable signal of entity recognition by Google. In this case, it serves as an experimental setting to document relationships between Wikidata, structured data, cross-platform consistency and entity recognition.
Client quote, translated: “I was told I needed Wikipedia to get a Knowledge Panel. I got the panel before the page.”