KDP royalties and delivery costs: the parentheses that change the calculation

A percentage is only the beginning

A royalty rate can make a publishing calculation look easier than it is. Seventy per cent is twice thirty-five per cent, but the two KDP options in this model do not apply their percentages to the same amount. The 70% calculation first subtracts a delivery charge. The 35% calculation does not subtract that charge. Understanding the base of each calculation is therefore more useful than comparing the two headline percentages on their own.

This study makes that difference visible in a reproducible set of hypothetical transactions. It uses Amazon.com, US dollars, zero applicable VAT and an assumption that every non-price eligibility condition has been met. Its 5,500 scenarios are mathematical combinations selected by Alisson Bento, not observations of authors, sales or uploaded books. The results can help a reader check arithmetic and prepare questions for a real title’s pricing preview. They cannot establish what an author earns or which option that author should choose.

The policy date and the assumptions belong beside the result

The official KDP pages captured on 28 October 2026 supply the model’s policy parameters. The eBook List Price Requirements page gives an Amazon.com 70% price band of USD 2.99 to USD 12.99, effective 7 July 2026. The effective date is when that price band applies; the October date is the evidence snapshot used here. The model takes eleven prices ending in .99 within that band. It does not treat the higher ceiling as a recommendation to raise a price.

The Digital Book Pricing Page supplies a delivery rate of USD 0.15 per Amazon-determined MB and a minimum delivery charge of USD 0.01 for this marketplace and currency. It specifies two decimal places for delivery cost and royalty. However, the inspected policy does not establish the exact rule for breaking a half-cent tie. It also does not provide the byte definitions needed to turn an uploaded file size into the particular MB quantity used here. Both uncertainties remain explicit throughout the analysis.

The phrase “eligible transaction” does important work. A price inside the band does not prove that a real title or customer qualifies for the 70% option. Customer territory, distribution rights, content eligibility and the relevant physical-edition price condition still matter. The calculation assumes those conditions rather than testing them. It also assumes that price matching, promotions, preorder adjustments, refunds and other transaction adjustments do not change the modeled payment base.

The parentheses determine how delivery enters the formula

Let P denote the zero-VAT list price in US dollars and D the delivery charge rounded to cents. Before final royalty rounding, the policy-based expressions are R70 = 0.70 × (P − D) and R35 = 0.35 × P. In the first expression, delivery is subtracted before the percentage is applied. Keeping that order visible is especially helpful when copying the calculation into a spreadsheet, where a small change in punctuation can produce a different answer without producing an error message.

The study also calculates A70 = 0.70 × P − D as a deliberately different expression. This is an arithmetic sensitivity case, not a claim about Amazon’s practice or an accusation against another calculator. Before final rounding, the policy-based result exceeds this alternative by 0.30 × D. The difference comes entirely from where the delivery charge enters the expression. It does not represent an additional platform fee, a measured loss to an author or a policy change.

The worked examples keep the list price fixed at USD 5.99 so that the delivery input is easier to follow. At 15 hypothetical platform-priced MB, the delivery charge is USD 2.25. Under illustrative half-up cent rounding, the policy-based 70% result is USD 2.62 and the 35% result is USD 2.10. The alternative expression produces USD 1.94. The two ways of placing the charge consequently give different orderings against the same 35% comparator.

Read the examples as calculations, not price advice

The complete worked-example table provides six sizes at the same USD 5.99 price: 1, 5, 10, 15, 20 and 25 platform-priced MB. At 1 MB, the policy-based 70% result is USD 4.09, against USD 2.10 for 35%. At 20 MB, the corresponding results are USD 2.09 and USD 2.10. This illustrates why the higher percentage alone cannot settle the comparison. The 35% result stays fixed in these examples because its modeled expression depends on price rather than delivery.

Every value in that table is a per-copy scenario result with zero VAT and assumed eligibility. None is a forecast of sales, a claim about a file downloaded by a customer, or an amount verified in an author account. The platform-priced MB input is hypothetical. A reader should not measure a local EPUB, substitute its file size without checking the platform’s own delivery-size calculation, and treat the table as a payment guarantee. The table’s purpose is to make the relationships inspectable.

Where the two expressions meet before rounding

Ignoring cent rounding, the policy-based expressions are equal when the delivery charge is half the list price: D = P/2. At the modeled delivery rate, the corresponding size is F = P/0.30 in platform-priced MB. The approved boundary table places that algebraic crossing at approximately 9.9667 MB for USD 2.99, approximately 19.9667 MB for USD 5.99 and 43.3 MB for USD 12.99. These are mathematical locations under the stated assumptions, not tested file-size cutoffs in Amazon’s interface.

The crossing chart also shows the deliberately different outside-parentheses expression. Its unrounded crossing is F = (0.35/0.15) × P, which is 70% of the height of the policy-based line. Showing both lines makes the consequence of fee placement visible across the selected prices. The figure is an explanation of algebra. Neither line represents a sample of files, a market trend, an estimate of typical book length or a recommendation about how large an ebook should be.

For a comparison rounded to cents, the boundary table is more precise about what is actually being varied: it enumerates delivery charges in one-cent steps. At USD 5.99, both illustrated rounding conventions produce a rounded tie at a USD 2.99 delivery charge; the first cent charge for which 35% is greater is USD 3.00. These charge boundaries should not be relabeled as exact upload-byte thresholds. Keeping the charge unit avoids inventing a conversion the captured policy does not specify.

What the 5,500 scenarios do—and do not—represent

The analysis plan fixes eleven prices and 500 sizes per price. Sizes run from 0.1 through 50.0 hypothetical platform-priced MB in increments of 0.1. The product is a grid of 5,500 declared scenarios. The spacing is chosen to make the model transparent and reproducible; it is not taken from a distribution of Kindle files. A different selection of prices, size increments or upper size limit would change the number of scenarios in the reported categories.

The main illustration uses nearest-cent half-up rounding. Where the rounded delivery charge is at least the list price, the scenario is labeled outside the modeled positive-net range. That rule leaves 4,658 scenarios in the main comparison and 842 excluded. Exclusion is an analytical boundary, not a statement that Amazon pays a negative royalty, charges an author the difference or applies a particular zero floor. The full download keeps all 5,500 rows and their labels so that this boundary is visible.

Among the 4,658 main positive-net scenarios, the policy-based expression and the alternative outside-parentheses expression give different orderings against 35% in 887 cases. There are seven rounded ties in the main calculation. Those are counts within this chosen grid. They cannot be interpreted as the number or share of authors making a mistake, the prevalence of an affected file type, or the size of an income loss in the publishing market. There are no sampled authors behind the denominator.

Small rounding differences need their own denominator

Two-decimal rounding leaves a further question: what happens at an exact half-cent? The main model uses half-up rounding, while the sensitivity uses half-even rounding. These are declared mathematical conventions. The source does not identify either one as Amazon’s internal tie-breaking implementation, so the comparison is useful for testing sensitivity rather than reverse engineering the platform. The download reports both calculations instead of allowing the chosen convention to disappear behind a displayed currency value.

Within the 4,658 scenarios that are positive-net under the main convention, switching the rounding convention changes the ordering in four cases. Across the complete 5,500-row grid, there are six differences in ordering or domain label. The additional two occur at USD 2.99 with 19.9 platform-priced MB and USD 5.99 with 39.9 platform-priced MB: those rows are excluded under half-up but enter the positive-net domain under half-even. Four and six therefore answer different counting questions; neither number should travel without its denominator.

The larger fee-placement comparison and the smaller rounding comparison also address different issues. Moving delivery outside the parentheses changes the expression itself. Switching the half-cent convention changes how specified amounts are rounded. Combining the two into a single error rate would obscure that distinction. The sensitivity table separates them by price, and the full scenario file preserves the individual outcomes for readers who need to trace a particular row.

How to read the three figures together

The frontier figure explains the unrounded relationship between price and crossing size. The worked-example figure compares the policy expression, the 35% option and the hypothetical alternative at selected USD 5.99 cases. The outcome grid shows the option ordering at whole-MB sizes across the selected prices. These are complementary views of the approved model, with different levels of detail. The whole-MB grid is a visual subset, not a replacement for the 0.1-MB scenario file.

The original analysis plan proposed a count chart as the third image; the completed model uses the whole-MB outcome grid instead. That presentation change is documented, while the count table remains available. Inputs, expressions, rounding sensitivities and exclusions were not changed to obtain the figure. When sharing any image, retain the marketplace, currency, zero-VAT assumption, eligibility assumption, hypothetical size unit, source snapshot and model description. A cropped graphic without those conditions can suggest a payment claim that the analysis does not support.

A useful check before consulting the actual pricing preview

A practical use is to audit a calculation: identify the price, identify the delivery charge, confirm which amount is multiplied by the royalty rate, and check where rounding occurs. The worked examples provide fixed reference cases against which a spreadsheet expression can be compared. The boundary table helps explain why two rounded results can tie even when their unrounded expressions differ. The scenario file allows an editor or developer to inspect the calculations without needing access to a publishing account.

The next question for a real title is whether its circumstances match the model. The captured terms exclude primarily public-domain content from the 70% option and include a physical-edition price restriction where applicable. Separate Select conditions apply to qualifying sales in Brazil, India, Japan and Mexico; this Amazon.com zero-VAT model is not an analysis of those transactions. An eligible price alone does not resolve rights, territory or other conditions. Those checks belong to the actual title and transaction.

Royalties are also distinct from profit and total income. The model contains no advertising, editing, cover design, production time, withholding or other author costs. It contains no demand curve, conversion rate, units sold or customer willingness to pay. It cannot identify a revenue-maximizing price, justify removing images, or establish that switching an option benefits a particular author. Kindle Unlimited and Kindle+ income are outside the model. Its value is the clarity of the calculation, not a promise about commercial outcomes.

A takeaway that keeps the qualifications attached

For hypothetical eligible Amazon.com transactions with zero VAT, the captured KDP policy places delivery inside the 70% calculation: R70 = 0.70 × (P − D). Across a chosen 5,500-scenario grid using illustrative half-up cent rounding, 4,658 scenarios have a positive modeled net price. In 887 of those, moving delivery outside the parentheses changes the ordering against the 35% calculation. These are deterministic scenario counts based on the 28 October 2026 policy snapshot, not estimates of author income or market prevalence.

Delivery costs, royalty-option comparisons and their basic crossover have been explained in earlier calculators and articles. This study adds a bounded, downloadable scenario set, explicit cent-charge boundaries, a disclosed alternative-expression comparison and visible handling of unknown implementation details. It makes no first-ever claim and does not test or rank third-party software. Original explanation, calculations and code are credited to Alisson Bento; the underlying policy facts remain attributed to Amazon KDP.

Data & downloads

Scroll horizontally to see all columns. ↔

Deterministic Amazon.com USD model; zero VAT and 70% eligibility assumed; snapshot 28 October 2026. MB means hypothetical platform-priced MB, not upload-byte size. Half-up/half-even rules are illustrative; counts are chosen scenarios, not authors, sales or income. All 22 approved unrounded curve points, including both correct and explicitly counterfactual equations; not rounded cent-charge boundaries or observed platform payments.

Scroll horizontally to see all columns. ↔

Amazon.com, USD, zero VAT and eligible transactions assumed; policy snapshot 28 October 2026. Counts describe the chosen grid. Rounding conventions are illustrative; retain excluded rows and exact source values.

Scroll horizontally to see all columns. ↔

Deterministic Amazon.com USD model; zero VAT and 70% eligibility assumed; snapshot 28 October 2026. MB means hypothetical platform-priced MB, not upload-byte size. Half-up/half-even rules are illustrative; counts are chosen scenarios, not authors, sales or income. Complete 550 heatmap outcomes: 50 whole-MB rows by 11 USD price columns. Every status is copied from the approved full scenario CSV. Full 5500 fine-grid cases remain separately downloadable.

Scroll horizontally to see all columns. ↔

Amazon.com, USD, zero VAT and eligible transactions assumed; policy snapshot 28 October 2026. Counts describe the chosen grid. Rounding conventions are illustrative; retain excluded rows and exact source values.

Scroll horizontally to see all columns. ↔

Amazon.com, USD, zero VAT and eligible transactions assumed; policy snapshot 28 October 2026. Counts describe the chosen grid. Rounding conventions are illustrative; retain excluded rows and exact source values.

Scroll horizontally to see all columns. ↔

Deterministic Amazon.com USD model; zero VAT and 70% eligibility assumed; snapshot 28 October 2026. MB means hypothetical platform-priced MB, not upload-byte size. Half-up/half-even rules are illustrative; counts are chosen scenarios, not authors, sales or income. 6 exact rows differ across rounding conventions over 5500; 4 are ordering changes among 4658 main positive-net scenarios, 2 are additional domain changes.

Evidence & method

How we calculated it

1. The plan fixes the marketplace, currency, zero-VAT condition, eligibility assumptions, eleven .99 prices from USD 2.99 through USD 12.99 and 500 hypothetical platform-priced sizes from 0.1 to 50.0 MB per price. These selections were recorded before calculation. The underlying policy pages were captured on 28 October 2026; the eligible price band became effective on 7 July 2026.
2. The Decimal implementation calculates delivery at USD 0.15 per platform-priced MB, with a USD 0.01 minimum, and rounds delivery and royalties to two decimal places under each declared convention. It computes 0.70 × (P − D), 0.35 × P and the hypothetical alternative 0.70 × P − D. Rounded delivery greater than or equal to price is excluded separately under each convention.
3. The boundary calculation enumerates exact cent delivery charges to identify rounded ties and the first charge for which 35% is greater. A separate full-grid comparison preserves all 5,500 rows, including excluded cases and the two convention-dependent domain shifts. Independent review recomputed every numeric and ordering field without importing the model’s calculation functions. The results are deterministic; sampling confidence intervals would not make the chosen grid representative.

Original sources

1. Amazon KDP — Digital Book Pricing Page

   Amazon KDP, Digital Book Pricing Page, G200634500. Captured 28 October 2026. Supplies the royalty expressions, Amazon.com USD delivery rate and minimum, rounding statements and eligibility conditions; unknown tie and byte conventions remain unresolved.

   Accessed October 28, 2026

2. Amazon KDP — eBook List Price Requirements

   Amazon KDP, eBook List Price Requirements, G200634560. Captured 28 October 2026. Amazon.com 70% price band USD 2.99–12.99 effective 7 July 2026. Price-band inclusion alone does not prove a real title’s eligibility.

   Accessed October 28, 2026

3. Amazon KDP — Paperback Royalty (supplementary scope only)

   The frozen source record also contains a paperback royalty page, but no paperback-income result enters this ebook delivery-cost model. Raw policy captures remain private evidence; the article uses factual paraphrases and direct policy attribution.

   Accessed October 28, 2026