The expensive time to discover a book-production problem is after every chapter has been laid out. A better test is one representative chapter in two forms: a print-size PDF and a genuinely reflowable EPUB.
These are not two exports of the same page design. Print fixes the page. A reflowable e-book lets the reader change the type size and allows the reading system to repaginate the text. A chapter that works in both will expose most of the decisions that need to be made before the rest of the manuscript follows.
Table of Contents
Choose the chapter that is least ordinary
Do not begin with the cleanest chapter. Choose one that contains the features most likely to break later: several heading levels, a long paragraph, an image and caption, a footnote, a quotation, a list, an unusual character, or a change of language.
For a bilingual book, include a passage where the two languages do not divide neatly into matching sentences. For a history or reference book, include a name, date, citation, and source note. For a technical book, include an equation, code block, or table if those are part of the real manuscript.
The sample should be small enough to revise, but complicated enough to answer an honest question: if this chapter works, is there any major kind of page left untested?
Freeze the source before styling it
Keep one source file from which both outputs are built. Give each chapter, section, paragraph, image, and note a stable identifier. Record where supplied text and images came from, what rights apply, and which version was approved.
This does not require a large publishing system. A short structured record is enough:
{
"id": "ch03-p014",
"source": "chapter-03.md#L88",
"language": "en",
"text": "The paragraph as approved by the author.",
"review": "approved"
}The important rule is that a correction returns to this source. If the PDF and EPUB are edited separately, they soon become two slightly different books. Stable identifiers make it possible to prove that the same paragraph reached both outputs and to trace a bad line break or missing sentence back to its origin.
Test the print page at its real size
Choose the intended trim size before tuning typography. Amazon KDP’s current trim, bleed, and margin guidance makes the dependency clear: margins depend on page count and bleed, while the gutter changes as the book becomes thicker. A beautiful sample set on the wrong page geometry does not answer the production question.
Inspect the PDF at 100 percent and, if possible, print a few facing pages. Check:
- whether the body type is comfortable rather than merely elegant on a monitor;
- whether headings leave a stranded line or an almost-empty page;
- whether inner margins remain usable near the binding;
- whether images, captions, notes, and page numbers have a consistent rhythm;
- whether all required fonts are embedded and every page has the agreed dimensions.
Bleed is a production choice, not a decorative afterthought. If even one interior element must reach the trimmed edge, the file dimensions and image placement need to account for it from the start.
Let the EPUB behave like an EPUB
An EPUB should not be judged by whether it resembles the print PDF at one screen size. EPUB 3.3 defines an ordered reading spine and a navigation document; its default layout is reflowable. The related reading-systems specification describes how a reader dynamically lays out that content.
Open the chapter with small and large type, narrow and wide screens, and light and dark themes if the design uses colour. Confirm that:
- the table of contents reaches the correct heading;
- reading order remains sensible without the print page positions;
- images fit the viewport and retain useful alternative text;
- footnotes and internal links return to the right place;
- language changes and unusual characters use suitable fonts and metadata;
- no heading, caption, or paragraph has been converted into a picture merely to preserve its shape.
Kindle Previewer can show a reflowable EPUB across Kindle device classes, but it is still worth opening the file in another standards-based reader. Run EPUBCheck as well. Passing validation does not prove a pleasant reading experience, but it catches packaging and conformance failures that visual browsing can miss.
Make one correction travel through both outputs
Now introduce a real correction: fix a name, replace a sentence, change a caption, or move a note. Apply it once in the source, rebuild both formats, and inspect the exact location in each.
This is the most revealing part of the specimen. It tests whether the production method is repeatable or whether the first result depended on quiet handwork that will have to be repeated across hundreds of pages. A useful delivery should also include a small manifest, source ledger, style sheet, and file hashes, so the next round begins from known inputs rather than a folder of vaguely named finals.
For multilingual work, trace the correction across the aligned passage as well. A changed source sentence should not leave an obsolete translation, reading aid, glossary entry, or index term behind.
Decide with an acceptance checklist
Before committing the full book, write down what the chapter has actually proved:
- the approved text appears completely in both formats;
- headings, notes, images, and language order follow the agreed style;
- print dimensions, margins, bleed assumptions, and embedded fonts are known;
- EPUB navigation, reading order, links, metadata, and reflow have been inspected;
- a source correction can regenerate both outputs without divergence;
- remaining exceptions are listed rather than hidden in “final” files.
The sample does not need to settle the cover, ISBN, marketplace metadata, printer choice, or every rare appendix. It should settle the repeatable interior rules and reveal which parts still need separate decisions.
Inspect a working specimen
I built a short source-aware book specimen from project-owned English and Traditional Chinese text. The public packet contains four 6 × 9 PDF variants, one reflowable EPUB, structured source with stable paragraph IDs, a style sheet, source ledger, validation report, manifest, and hashes. PocketPolyglot shows the broader multilingual publishing pipeline behind the experiment.
It is a process specimen, not a customer result or a claim that one layout suits every printer. Its value is that the files can be opened, compared, rebuilt, and traced before anyone has to trust a promise.
If you have final, rights-cleared copy and want to test one representative chapter before commissioning the whole book, the USD 250 Book Specimen Sprint covers one chapter up to 5,000 words, one agreed 6 × 9 print profile, one reflowable EPUB, and the source-and-validation packet. The page shows the exact limits and starts with a fit check before any manuscript upload or payment.
One difficult chapter is usually enough to replace a vague production estimate with a concrete decision.