Case study · Reference content at scale
Everything already exists — reviews, rankings, listicles. What does not exist is any of it organized for the moment it is actually needed: a group in a vehicle, forty minutes out, deciding where to stop.
Everything already exists. Reviews, rankings, listicles, aggregator pages. What does not exist is any of it organized for the moment it is actually needed — a group in a vehicle, forty minutes out, deciding where to stop.
Two guides were built for exactly that moment.
The scale
Word counts from the finished documents. Both were also produced in phone-formatted versions, because the office copy is not the one anybody reads.
What the Maine guide holds
Restaurant totals drawn from the full listing set for the coverage area rather than a top-ten selection, then filtered and organized. Category counts are as compiled at the time of writing.
The Manhattan guide is organized differently, because the geography is. Five neighborhoods, each characterized before a single restaurant is named — what it is like, who eats there, what it is known for. Someone who does not know the city cannot use a list. They can use a character sketch and then a list.
Comprehensiveness first, then curation on top — not curation instead. A guide that only holds someone’s twenty favorites is useless the moment the reader wants something that is not on it. Hold the full set, then layer the shortlists over it.
| Decision | Why |
|---|---|
| Geography as the top level, always | The reader’s first constraint is where they are, not what they feel like eating |
| Category beneath geography | Fine dining, seafood, budget, late night — the second question, never the first |
| Every entry carries a reason | A name and an address is a directory. A reason to go is a guide |
| Neighborhood character before listings | Somebody unfamiliar with a city cannot choose from a list of names |
| Two formats, always | A document for planning, a phone version for the parking lot |
| Compilation date stated | Restaurants close. A guide that hides its own vintage is misleading |
They carry a compilation date, because a dining guide is perishable and pretending otherwise is the single most common failure in this kind of content. Anything with hours, prices or an address needs a date attached and a refresh plan, or it quietly becomes wrong while still looking authoritative.
| AI did this well | A person had to do this |
|---|---|
| Assembling and structuring the full listing set | Knowing which places are actually worth the detour |
| Drafting consistent entries at volume | Writing the neighborhood character sketches |
| Reformatting everything into a second, phone-ready version | Judging what a first-time visitor will misunderstand |
| Maintaining one voice across 58,000 words | Deciding what to leave out |
The structure and the volume came from the machine. The judgment came from a decade of driving people to these places and hearing what they said afterward. A guide written entirely from public sources reads like every other guide, because it is assembled from the same material as every other guide.
| The usual approach | What was done here |
|---|---|
| A top-ten list | The full set, with shortlists layered on top |
| Organized by category | Organized by where the reader physically is |
| Names and addresses | A reason to go, on every entry |
| One document | Two formats, because the reading context is different |
| Undated | Compilation date stated, because this content perishes |
Any organization sitting on knowledge that lives in someone’s head — a service manual, a territory guide, an onboarding pack, a product reference. The volume is not the hard part any more. Structuring it so it can be retrieved in the thirty seconds somebody actually has is the hard part, and it is almost always skipped.
Marketing Analytics Consultants
We structure it so it can be found in the moment it is needed — in the formats people actually use, with the shelf life stated rather than hidden.