Book Metadata & Categorization Guide

How to write retail descriptions, choose categories, and pick keywords for an MMP title — including where IngramSpark’s “Assist (AI)” auto-generation tends to get things wrong and needs correcting.

Full Description (the retail synopsis)

• Field limit is stated in bytes, not characters (200–4000 bytes on IngramSpark). If diacritics are present (Ś, ū, etc.), each costs 2 UTF-8 bytes, not 1 — usually a small effect, but leave real margin: in practice the platform’s own byte count has landed noticeably higher than a straight UTF-8 byte count of the plain text, so don’t cut it right down to the 3900s. Target well under the cap (e.g. under 3000) if using HTML formatting, since markup tags also count.
• HTML is supported via the </> source-view toggle in the rich-text editor (<p>, <b>, <i>) — use it for a bold opening hook line and italicized title mentions, matching how professional retail listings are typically formatted. Minimal markup only — every tag adds to the byte count.
• Write what the book actually argues, not just its subject matter. For a book making a deliberately narrow/qualified claim (e.g., “not a claim of technological lineage, but a methodological comparison”), the description must preserve that same restraint — don’t let marketing copy overclaim what the book’s own front matter explicitly disclaims. Cross-check any drafted description against the book’s actual “What This Book Does—and Does Not Claim” framing if one exists.
• Case-study specificity is a judgment call, not a default. Early drafts should usually stay at the “teaser” level (what kind of case study, not which one) — only name specific real-world cases (a place, an institution, a named program) if the author explicitly wants them surfaced. Ask rather than assume either way.
• Consider adding a short “why this book exists” / origin-story paragraph near the end, before the author bio — grounded in the actual, specific research history behind the book (an article series, a prior project, a specific realization), never a vague “I’ve always been passionate about…” framing. Verify the concrete origin detail with the author rather than inventing one.

Short Description

• Much tighter limit (~250 bytes on IngramSpark) — one or two sentences, shown on retailer sites.
• Must represent the whole book, not just its most “hook-y” chapter or case study. An AI-generated short description narrowed an AI-and-history book down to “modern biometric systems” — accurate to one chapter, wrong as a summary of the whole argument. Always check an auto-generated short description against the full table of contents before accepting it.

Author Bio (Contributor Biography)

• Keep a short version (2–3 sentences, for back covers and short-bio fields) separate from the full version (for an inside “About the Author” page). Don’t reuse the long one where space is tight — it visibly crowds out the barcode/price/category block on a physical back cover.
• Every author-fact claim gets checked against the standing hard rules before it ships — a fixed private list of what can and can’t be said about Viveka Mohan Das in any bio or description text, covering specific institutions, credentials, and figures that must never be stated or must be phrased a particular way. That list isn’t reproduced here; check it directly before writing or approving any bio copy, for any book.

BISAC Subjects (up to 3)

• The single most common failure mode: an AI-suggested BISAC list that only reflects one subject area of a multi-subject book. On a book that was half AI/technology and half history/mathematics, IngramSpark Assist’s first pass picked three purely computer-science subcategories (including “Expert Systems” and “Natural Language Processing” — neither actually discussed in the book) and missed History and Mathematics entirely. Always manually check the BISAC list against the actual table of contents before accepting it.
• Prefer broad, accurate matches over narrow ones that sound impressive but don’t fit the content (e.g., plain “Artificial Intelligence / General” over a specific AI subfield the book doesn’t actually cover).

Thema Subjects & Qualifiers

• Thema is a separate international classification (used by UK/EU retailers, Amazon.co.uk, libraries abroad) — runs alongside BISAC, doesn’t replace it. Same rule applies: check it covers every major subject area, not just the most obvious one. Use “Find Themas” to search for missing angles (e.g., “History of mathematics,” “Epistemology & theory of knowledge”) if the AI-suggested set is AI/tech-only.
• Qualifiers are facets on top of a Thema subject (place, time period, language, interest level) — not new subjects. For a historical-claim-sensitive book, check any auto-suggested time-period qualifier against what the book itself actually claims. A qualifier narrower than the book’s own stated uncertainty (e.g., picking a precise 500-year window when the book deliberately only claims “around the first millennium BCE”) is a real, if minor, inconsistency worth catching.

Keywords

• Semicolon-separated list. Check for the single most obvious search term explicitly — an AI-generated keyword list omitted “artificial intelligence” entirely from a book whose own subtitle is ”…Trustworthy AI.” Don’t assume obvious terms made the cut; verify.
• Cut invented-sounding phrases an AI keyword generator sometimes produces that aren’t real search terms (e.g., “procedural truth,” “computational history” — ambiguous and not something a reader would type into a search bar). Prefer plain, high-traffic terms a real reader would search: the book’s core subject nouns, its field (AI, history, mathematics), and its geography if relevant (e.g., India).

Select Audience

• Default to “Trade/General (Adult)” for anything going into full retail/bookstore distribution. Reserve “Professional/Scholar (Adult)” only for a deliberately academic/library-only positioning — it’s a real categorization signal to retailers, not just a label, and can narrow how the book gets recommended/shelved.

Table of Contents field

• Paste the real chapter/part structure, matching what’s inside the book exactly. Plain indented text (no HTML) is the safest, lowest-overhead format if byte limits are a concern; a nested <ul><li> HTML list looks more “premium” if there’s room in the byte budget, but carries meaningfully more overhead per entry — check the actual byte count of whichever version you use, don’t assume.

Reusable checklist for the next title

• Full Description reflects what the book actually argues, respects any stated scope limitations, checked against real byte-limit margin
• Short Description independently checked against the full ToC — not just accepted from AI-Assist
• BISAC subjects manually verified to cover every major subject area, not just the most obvious one
• Thema subjects + qualifiers checked the same way, qualifiers checked against the book’s own stated claims
• Keywords include the single most obvious search term explicitly, invented-sounding phrases removed
• Audience set to Trade/General unless a deliberate academic-only positioning is wanted
• Author bio checked against the standing hard-rules list on every piece of author-facing copy, not just the main bio field