Book metadata: title, author, categories and keywords

Metadata is the part of publishing that feels like paperwork and behaves like marketing. Nobody reads it, everything reads it: the store's search index, the library catalogue, the reader's device, the retailer's recommendation engine, and eventually whatever answers the question your book was written for.

It is also the last thing anyone does, at the end of a long project, in a hurry, in a form with twelve fields. That is why so many self-published books are shelved next to the wrong things.

Where metadata actually lives

There is no single place. The same facts are recorded three times, in three formats, and nothing keeps them in sync for you.

The EPUB's metadata is the one that has hard requirements: a title, a language code and a unique identifier are mandatory, and an EPUB missing any of them fails validation before it reaches a store. If you have run into that, it is one of the common EPUB validation errors and it is a two-minute fix.

The title field, and the four ways it goes wrong

• Subtitle in the title field. Most stores have a separate subtitle field. Cramming both into one produces a search result that reads as spam and a library record that is wrong.
• Series information in the title. "Book 2" belongs in the series fields, not in Title. Stores that understand series will order them for you; ones that do not will at least not display it twice.
• Keyword stuffing. "The Sourdough Book: Bread, Baking, Starter, Recipes, Beginners" is a title that gets a listing suppressed at some retailers and mistrusted at all of them.
• A title that differs from the cover. Even by a colon, an ampersand or a capital letter. Some retailers check, and a mismatch is a rejection or a manual review.

The subtitle is where the searchable words go, and it is doing real work in non-fiction. Slow Rise tells a store nothing. Slow Rise: A Beginner's Guide to Sourdough Without a Schedule tells it who the book is for, in the words those readers actually type.

The author name is an identifier

Treat your author name as a primary key, not as a piece of writing. Stores group books by exact string match, and a small variation creates a second author with one book on their shelf instead of a first author with two.

Pick one form, including the spaces between initials, write it down somewhere, and use it in the EPUB metadata, the PDF properties, every store, the copyright page and your website. The same discipline applies across a series, which is part of the wider argument for keeping a series consistent from the first book rather than the third.

Categories

Categories decide which shelf the book sits on and, more usefully, which bestseller lists it can appear in. Most retailers use BISAC codes — a standard subject taxonomy — or their own tree mapped onto it, and you usually get two or three choices.

1. Be as specific as the tree allows. Cooking > Methods > Baking beats Cooking. A broad category puts you behind every major publisher in it.
2. Pick categories where your book genuinely belongs. Miscategorising to reach an easier list produces returns and one-star reviews from readers who wanted the other thing.
3. Use your slots on different branches. Three neighbouring subcategories cover the same readers three times; two distant ones cover twice as many.
4. Check what is actually ranking there. If the top ten in a category are all traditionally published and five years old, it is a category with demand and no room.
5. Revisit after launch. Categories can be changed, and the right one at 200 sales is often not the right one at 20.

Keywords

Keyword fields are search queries, not tags. The mental model that works is: what would someone type if they wanted this book and did not know it existed? Not the words that describe your book — the words that describe their problem.

• Phrases, not single words. sourdough for beginners no starter is a query. sourdough is a category with forty thousand competitors.
• Do not repeat the title or the author. Those are already indexed. A keyword slot spent on them is a slot wasted.
• No other authors' names, no trademarks. It is against the terms at most retailers and it gets listings pulled.
• Include the format or context where it is genuinely distinguishing: workbook, printable, for teams, two-week plan.
• Write them in the reader's words, which are rarely the words a writer would choose. Look at the questions in reviews of similar books.

The description

The description is metadata that a human reads, which makes it the only field doing two jobs at once. It is indexed for search and it is the sales page, and the two pull in different directions — one wants the words readers type, the other wants a sentence that makes someone buy.

The resolution is order. The first two lines are for the reader, because that is all a store shows before the fold and it is where the decision is made. The rest can carry the practical detail — who it is for, what is inside, how long it is — in plain language that happens to contain the words people search for.

Identifiers

An EPUB needs a unique identifier and it does not have to be an ISBN — a UUID is valid and is what most tools generate. Whether you need a real ISBN depends on where you are selling and in what format, which is a question with more nuance than it first appears.

• One identifier per format. The EPUB and the paperback are different products and take different ISBNs, if you use them.
• Never reuse an identifier for a revised edition with substantial changes. A new edition is a new record.
• Keep the identifier stable across re-exports. A tool that generates a fresh UUID every time you export makes each build look like a different book to a library system.

A five-minute check before you upload

1. Title and subtitle match the cover exactly, in the right two fields.
2. Author name is the exact string you used last time.
3. Language code is set — this one is mandatory in EPUB and quietly wrong surprisingly often.
4. PDF properties carry the title and author, not Untitled or the name of your word processor.
5. Categories are as deep as the tree allows, on different branches.
6. Keywords are phrases someone would type, and none of them repeat the title.
7. Description opens with two lines aimed at a person.

It is worth doing this once, properly, and then keeping the answers in a file next to the manuscript. Every subsequent book, edition and format re-uses most of them, and the alternative is retyping your own name slightly differently at eleven at night — which is, genuinely, how most split author pages happen.