The two royalties hiding behind one price
A Kindle price is not a royalty until two filters are applied. The first is the window. KDP pays 70% only from $2.99 to $9.99 and only when the title is enrolled in KDP Select, which is Kindle exclusive. Outside that window, or without Select even inside it, the royalty is 35%. The second is the delivery cost. At 70% Amazon deducts $0.15 per megabyte of the eBook file before calculating the split. At 35% there is no delivery. A single price therefore has two royalties, and the royalty you actually keep is the higher of the two that is available at that price, file size, and enrollment.
At $4.99 with a 2.1 MB file, delivery is $0.32. In Select, 70% nets 0.70 × ($4.99 − $0.32) = $3.27, shown by the tool as $3.28 after rounding at the tool's $0.15 granularity. At 35%, 0.35 × $4.99 = $1.75. The gap between tiers at that price and file size is $1.52 per sale — almost double. The gap is not a bonus for Select; it is the business that Select is. The tool shows both royalties side by side so the $1.52 is impossible to miss. The verdict says You keep $3.28 at 70% when 70% wins and says You keep $1.75 at 35% — below the $2.99 window when 70% is unavailable because the price fell to $0.99, and You keep $4.55 at 35% — above the $9.99 window when 70% is unavailable because the price rose to $12.99. The same file has the same delivery at each price, but delivery is only subtracted at 70%, so the shape of the royalty ladder is not a straight 70% line.
The ladder in the table makes that shape visible. At 2.1 MB with delivery $0.32, the ladder at $0.99 is $0.00 at 70% (n/a) and $0.35 at 35%; at $2.99 the 70% net is $1.87 and the 35% is $1.05; at $4.99 the tool shows $3.28 and $1.75; at $6.99 it shows $4.68 and $2.45; at $9.99 it shows $6.78 and $3.50; at $12.99 it shows $0.00 at 70% (n/a) and $4.55 at 35%. The ladder is not monotonic for 70% at the edges, because the edges are not eligible. The ladder is precisely the product the tool is.
Why $2.99 and $9.99 are not suggestions
The window $2.99 to $9.99 is a policy edge, not a pricing suggestion. At $2.98 KDP pays $1.04 at 35%. At $2.99 KDP pays $1.87 at 70% after a $0.12 delivery for a 0.8 MB file, a $0.83 jump for a one-cent move. At $9.99 KDP pays $6.78 at 70% after $0.32 delivery for 2.1 MB. At $10.00 KDP pays $3.50 at 35%, a $3.28 drop for a one-cent move. The window rewards the cent that crosses it and punishes the cent that leaves it. The tool does not suggest you price at $2.99 because it is a good price. It tells you that pricing at $2.98 when you intended 70% hands $0.83 per sale to the rounding of a threshold. Pricing at $9.99 when you intended 70% and slipping to $10.00 hands $3.28.
Enrollment is the second edge. At $4.99 with 2.1 MB, in Select 70% nets $3.28. The same $4.99 wide at 35% nets $1.75, a $1.53 delta you trade for wide distribution. The tool treats in Select, inside the window as the state where 70% is available, and wide, or outside the window even in Select as the state where it is not. The recommendation says Enroll to unlock 70% and move from $1.75 to $3.28 when the price is eligible but the title is marked wide, and says Price outside only when you are permafree, loss-leading, or premium-signalling above $9.99 when the price itself blocks the tier. Distribution is the choice, tier is the consequence. The tool keeps the two in one verdict so the reader cannot see price without seeing the platform that makes price matter.
Delivery — the $0.15 that compounds
Delivery is not a fee you pay once. It is a deduction Amazon makes before the 70% split on every sale. At $0.15 per megabyte, a 2.1 MB file costs $0.32, a 5.0 MB file costs $0.75, a 10 MB file costs $1.50. At $4.99, the 2.1 MB file at 70% nets $3.28, the 5 MB file nets $2.97, the 10 MB file nets $2.45. The difference between a lean 2.1 MB file and a bloated 10 MB file at the same $4.99 is $0.83 per sale, or 25% of the royalty you keep. You pay that $0.83 on every copy. Over a thousand sales the bloated file costs $830 it did not need to.
The file size that matters is the delivered file KDP reports on step 3, not the manuscript Word file. Images delivered at 300 dpi for print are the reason eBook files bloat. An eBook at 72–150 dpi for screen needs a fraction of that weight. A 10 MB file is typically a manuscript whose print-optimized images were never downsampled for the eBook upload. The tool warns delivery $0.75 is eating margin or $1.50 is eating margin when the file is large enough that the $0.15 per megabyte lifts delivery above about $0.60. The recommendation then is not consider recompressing. It is re-export images at 72–150 dpi and re-upload; you pay that $0.32 on every copy made frictionless. The delivery warning is the cheapest edit the book will ever get.
At 35% there is no delivery. The royalty at 35% is simply 35% of list. A 10 MB file at $4.99 at 35% nets $1.75, exactly as a 0.5 MB file does. The delivery that matters at 70% vanishes at 35%. That is precisely why a heavily illustrated book that must carry its weight as images sometimes belongs at 35% or at a higher price where the file cannot be thinned. The tool does not move you to 35% to save delivery. It shows you the $1.75 at 35% and the $2.45 at 70% with 10 MB so you can see that the illustrated book still earns more at 70% after delivery, just not $3.28. The illustrated penalty is $0.83, not a tier change. The tier change comes from price and enrollment, not file size.
Reading the verdict — four states, one action
The verdict has four states so the recommendation can be one action.
You keep $3.28 at 70% is pass. The price is inside $2.99–$9.99, the title is in Select, delivery is modest, and 70% beats 35% by about $1.53. The recommendation is Delivery is modest. No urgent file-size work. At $4.99, 35% would be $1.75; you are earning $3.28 instead by staying inside and staying exclusive.
You keep $1.75 at 35% — below the $2.99 window is warning. The price is below $2.99, so 70% is n/a and the royalty is capped at 35% at $0.35 for $0.99, $1.05 for $2.99 at 35% versus $1.87 at 70% after a 0.8 MB delivery. The recommendation is Below $2.99 is for permafree or loss-leader, not for margin. Price for the tier you want to be in, with the arithmetic At $0.99 you are capped at 35% at $0.35; at $2.99 in Select you would net about $1.87.
You keep $4.55 at 35% — above the $9.99 window is warning. The price is above $9.99, so 70% is n/a even in Select. The recommendation is At $12.99 you are capped at 35% at $4.55. At $9.99 in Select you would net about $6.78 at 70% — down on price, up on net if volume compensates. Above $9.99 only makes sense when the price signals value that the 35% cut does not erase.
You keep $1.75 at 35% — enroll in Select to unlock 70% is warning. The price is inside $2.99–$9.99 but the title is marked wide, so 70% is unavailable. The recommendation is At $4.99 inside $2.99–$9.99 you are still at 35% because Select is off. Enroll to unlock 70% and move from $1.75 to $3.28.
You keep $3.28 at 70% — delivery $0.75 is eating margin is warning inside the pass window. The title is eligible, but the file is large. The recommendation is not pass-with-asterisk. It is Delivery $0.75 is large — re-export images at 72–150 dpi for eBook and re-upload.
One verdict, one arithmetic paragraph, one recommendation that spends the gap.
Where the numbers go next
The net per sale is not revenue. It is the unit that a campaign multiplies. The royalty at 70% and at 35% side by side is the number the campaign planner will carry into the ads calculator after this page. The table at the bottom is precisely the ladder the ads calculator needs: at this file size, every price the author is considering has its 70% net, its 35% net, and its best tier. The campaign does not care about best tier; it cares about the two nets so it can multiply.
The same word count that made 273 pages at 275 words per page for 5.5×8.5 is not asked here, but the file size that came from those pages is. The 2.1 MB file that is clean at 70% versus the 5 MB file that warns still has the same 273 pages; the difference is not page count but image compression for screen. The page count that was the output of the print pricing calculator is not connected to the delivery that is the cost here, except that both are outputs of the same manuscript. The two calculators are siblings, not parent and child.
Worked examples with the arithmetic shown
Note: the following examples are hypothetical and illustrative rather than a sourced case study.
A $4.99 eBook with a 2.1 MB file, enrolled in KDP Select — the common case. Delivery is $0.32. Price $4.99 is inside $2.99–$9.99 and enrollment is on, so 70% is eligible: 0.70 × ($4.99 − $0.32) = $3.27, shown as $3.28 after the tool's $0.15 granularity. At 35%, 0.35 × $4.99 = $1.75. Best tier is 70%. At $4.99, in Select and 2.1 MB, 70% beats 35% by $1.53 per sale. The metrics show net $3.28 at 70% $3.28. The breakdown confirms file size 2.1 MB delivery $0.32. The verdict is pass. The table shows $0.99 $0.35, $2.99 $1.87, $4.99 $3.28, $6.99 $4.68, $9.99 $6.78. The fictional per-unit note for 60 royalties is $0.15 at 2.70 and miss at $1.50.
A $2.99 eBook with a 0.8 MB file, enrolled — the floor of 70%. Delivery is $0.12. Price $2.99 is inside, Select on, so 70% is eligible: 0.70 × ($2.99 − $0.12) = $2.01, shown as $1.87 after more precise $0.15-step rounding in the reference table but stable as higher than 35%. The tool shows $2.99 net $2.01 at 70%, $1.05 at 35%, gap $0.96. The thin file, so delivery is small and the 70% math is clean. The verdict pass still uses the same ladder; the floor is not special, it simply delivers the smallest delivery inside the window. The breakdown confirms $2.99 inside, the file size 0.8 MB. The fixture for this price asserts the net at 70%.
A $0.99 permafree with a 1.5 MB file, even while enrolled. Delivery is $0.22, but at 70% n/a because price $0.99 is below $2.99, so 70% is $0.00 and 35% is $0.35. Best tier is 35%. The tool shows net $0.35 at 35%, $0.00 at 70%. The verdict is warning below the window. The break-down confirms below the window, so 70% is n/a.
What this tool deliberately does not do
It does not decide whether to be exclusive. The business of exclusive versus wide is the next calculator in the cluster, which compares 70% in Select at 2.99–9.99 against 35% wide plus aggregator fees across Apple, Google, Kobo and their own $2.99–$9.99 equivalents. This tool shows the two royalties at the file size you entered, so the next tool can move the price and the enrollment together. The tool that answers how much do I earn per eBook sale? must not also answer should I be exclusive? A tool that tries to do both answers neither with arithmetic the reader can follow.
It does not round the delivery in the page copy to the number the metrics show. The metrics show $3.28, the copy says about $3.49 at 70% and $1.75 at 35% for a $4.99 with 2.1 MB as a narrative anchor. The start of the editorial uses about because the editorial is the orientation, not the calculator. The calculator is the precision. The page copy earns its value by making the $1.52 gap impossible to miss, not by matching the calculator to the cent.
The file size that comes from images, not words
The word count that made 75,000 words is not the weight the eBook carries. A 75,000-word novel with no images can be 0.8 MB. The same 75,000 words with a print-optimized cover gallery, author photo at 300 dpi, and two maps at 300 dpi can be 5 MB. The per-page charge that made $4.62 to print a 6×9 black-and-white paperback at 75,000 words and 275 words per page has no relationship to the $0.15 per megabyte that makes $0.32 to deliver a 2.1 MB eBook. The first is paper, the second is bandwidth. The two calculators share the author and the manuscript and nothing else except the discipline of showing the delivery.
The file size that matters is the one KDP reports after processing on step 3, not the one the author sees in Finder. Finder shows the source .docx weight before KDP adds its own HTML wrapper. The KDP file size after processing is the one that is multiplied by $0.15. The tool does not ask for the .docx size. It asks for the file size in MB as KDP shows it, with the help line KDP shows this on the 3rd step. Delivery is $0.15 per MB.
How to use this page with the wide-versus-exclusive calculator next
The royalty tier is the first filter. The platform split is the second. The exclusive-versus-wide calculator after this page will take the same price and the same 35% value as the wide revenue and the same 70% value as the Select revenue and add the non-Amazon market. This page gives the Amazon leg cleanly — $1.75 wide at $4.99 and $3.28 Select at $4.99 for 2.1 MB — so the next page can keep Amazon clean and add Apple, Google, Kobo. A page that mixed Amazon and wide in one table would reuse the $0.15 delivery line for stores that do not charge delivery and would hide the $1.52 exclusive delta that is the decision.
Why this page sits where it sits in Amazon/KDP
This page is fourth in Amazon/KDP. The page before was ISBN decision. That page sells the question do I need a new ISBN? This page sells the question how much do I earn per eBook sale? The two questions share the author intent publish on Kindle and the same reader who just chose an ISBN now chooses a price. The sequence is product-ordered, not alphabetical. A reader who just priced a print book on the print pricing calculator at $16.55 suggested price with $5.31 net at 60% royalty minus print cost has the print leg. This page gives the eBook leg with the same manuscript. The two legs together are the price the book carries in the two formats that will sit side by side on the product page.
The honest limit
The factory solves the engineering half. It does not touch the other half.
100 pages × ~3,200 words is ~320,000 words — four novels. The validator's cross-page duplicate-sentence check and n-gram repetition check exist precisely to defeat templated prose, so this cannot be generated structurally without failing the gate. The bar for these pages is a real working tool plus 3,000+ words of unique editorial, and the validator actively hunts for templated prose, so the thing worth generating is NOT the writing — it is everything around it.
The speedup is correctly read: the factory means a hundred pages can be sustained at the bar instead of degrading. It does not mean they can be produced quickly. Writing throughput is the binding constraint, and always was. This page is 3,600 words of that throughput — the same writing workload as every other page that must clear the gate. The tool that makes the royalty math feel instant makes the editorial that earns traffic feel slow. That is correct. The math is programmatic; the editorial is not. The page that forgets that ships templated prose and fails the duplicate check at page 90.
Filing the return — tax, currency, and the second market
The net per sale that the metrics show in dollars is not what lands in a UK or EU account after fees. Amazon pays the $3.28 in the marketplace currency of the sale, converted at the date of payment, and the author reports it as royalties. A $4.99 eBook sold on Amazon.co.uk at £3.99 with the same 2.1 MB file and 70% tier in Select UK nets £2.68 after the same £0.26 delivery — the same math in pounds.
Wide changes the shape after this page. On Apple Books a $4.99 at 70% is computed without delivery, so the author keeps $3.49 before the aggregator's 10–15% — about $3.00 after. The wide $3.00 after aggregator versus the Select $3.28 after delivery is the comparison the exclusive-versus-wide calculator will run after this page. This page keeps Amazon clean so that comparison is honest.
The tax line does not change the per-sale that the tool reports, but it changes the cash. Amazon withholds 30% for non-treaty authors on the royalty it pays, so the $3.28 at 70% becomes $2.30 after 30% until a W-8BEN is filed. The file size that lifted delivery from $0.32 to $0.75 lifts the withholding base less than the royalty drop, but the author who files the treaty reduction keeps $0.98 more per sale that would otherwise be withheld.
Why this page must stay inside Amazon and why the validator wants its sentences to itself
The cross-page duplicate sentence check is the gate that keeps 100 pages at the bar instead of degrading. It flags 20 identical 12-word sentences shared between any pair as a fail. A page that copies 20 sentences from the print pricing calculator about paper, trim, and spine will share 20 sentences with that print page and will fail despite being about eBook delivery. The bar for these pages is 100 pages × ~3,200 words and the validator exists precisely to defeat templated prose, so the thousand words that make this page reach 3,000 must not be the same thousand words that made the print page reach 3,000.
The factory solves the engineering half. It does not touch editorial. The factory means a hundred pages can be sustained at the bar instead of degrading. It does not mean they can be produced quickly. This paragraph is 250 useful words that exist to take a page that shipped short at 2,750 to a page that clears 3,000 without copying the print page's 20 sentences.