[ Case study ]
The brand's glossary explained its topics better than any competitor, but answer engines quoted competitors' definitions instead. The content was good; the machine-readable facts around it were not.
CLIENT a consumer-fintech education brand — FOCUS Make the facts extractable
Representative examplesEvery case study in this library is an illustrative composite of the kind of engagement we deliver — written to show our method and standards, not to name clients.
Explaining money plainly is the craft of this consumer-fintech education brand: a 140-article glossary written by people who know the material, reviewed by compliance before anything ships. Human readers who found it rated it above every competitor. The problem lived downstream of quality — the machine-readable facts around the articles barely existed, so answer engines quoting 'what is X' questions pulled definitions from competitors whose pages were easier to extract, not better to read. No one on the small content team owned that layer.
The brand's glossary explained its topics better than any competitor, but answer engines quoted competitors' definitions instead. The content was good; the machine-readable facts around it were not.
We proposed making the facts extractable without touching the approved claims. Definitions would move into question-and-answer structures with schema stating the entity, the claim, and the source — written for extraction, still readable for humans, and never reworded beyond structure. The brand's entity needed resolving: organization and author schema, sameAs links, and consistent naming so systems could attribute a quote to somebody. Measurement would be built, not bought — a fixed weekly prompt set run against the major answer surfaces, logged, and honestly framed as a panel rather than a guarantee.
Just as important is what we ruled out, and why:
Definitions moved into question-and-answer structures with schema that states the entity, the claim, and the source — written for extraction, still readable for humans.
Organization and author schema, sameAs links, and consistent naming were aligned so systems resolve who is making the claim.
A weekly prompt set ran against the major answer surfaces, recording whether and how the brand was cited — a panel, not a guarantee.
Delivered by the growth pod — content strategist + technical SEO specialist over 10 weeks, with working increments reviewed with the client every week.
Obstacle
The first restructure batch dropped the attributions: definitions grounded in 'per the regulator' language shipped as bare claims, and the extracted answers read as the brand's own rulings rather than sourced facts.
Handled: Attribution became a required field in the restructure pattern — no definition ships without its source named, and batches missing one bounce back — so grounding is enforced by the field, not by wording rules.
Obstacle
The 'before' citation counts had been estimated from memory because the panel started after content work began — a measurement gap we were unwilling to paper over.
Handled: We re-ran the full prompt set against frozen pre-change snapshots of the articles on staging, reconstructing a defensible baseline and restarting the clock honestly.
The headline: weekly prompt-set answers citing or summarizing the brand's content, weeks 1 to 10 — 4 → 27, read from Weekly answer-panel tracking. A second check: organic entrances to the restructured articles at +21%.
The brand's name now appears in answers its own authors could have written, and the team writes definitions with the claim-and-source shape by habit — the structure outlived the engagement because it made review faster, not harder. Compliance trusts the format now that meaning has never moved. The weekly citation email replaced a question nobody could answer: where do we stand, and can we trust the number?
The result was read from Weekly answer-panel tracking against the pre-engagement baseline over the stated window, with a guardrail check on organic entrances to the restructured articles. Where platform-reported numbers and business outcomes differ, this record says which layer it is quoting.
What we would do differently
We would have built the answer-panel baseline before touching content — the 'before' citations were estimated from memory, which we now consider a measurement gap.
[ Related service ]
[ Related builds ]
14 31Qualified inquiries per month attributed to service pages, over the 8 weeks after rollout versus the 8 before
3 43Referring domains from non-sponsored coverage, over 12 weeks
[ Next step ]
Next case study
