A tapa book punishes a bad file in a way a sample puzzle never shows. Your buyer judges puzzle forty-one, deep into a larger grid, when the wall they have been painting cell by cell quietly breaks into two pieces, or a clue that promises two blocks of shaded neighbours can only ever produce one, or four shaded cells meet in a solid square the rules forbid. A tapa, also presented as Paint by Numbers, is a contract in four parts: the buyer shades cells to form one single connected wall; no block of two by two shaded cells appears anywhere; clue cells stay unshaded and are never part of the wall; and each clue cell's numbers describe, in order around that clue, the lengths of the consecutive blocks of shaded cells among its neighbouring cells, with at least one unshaded cell separating blocks. One broken part does not make the puzzle slightly harder. It makes it false as stated. A good tapa book is a verified set: the wall always one connected whole, no forbidden square anywhere, every clue sequence exactly true, every puzzle holding exactly one solution, every key re-checked against the final page, and every difficulty label graded on grid size, clue density, wall length and tightness, and neighbourhood ambiguity together.
The book is the puzzle set, not the sample grid
Buyers do not solve your listing image. They judge it by whether each grid can be finished as stated, with a legal wall, no forbidden square, and every clue around every clue cell reading exactly as printed.
Ship the book as a first-class solving experience, not one sample grid with the rest implied. Name exactly what the download contains. If the book holds eighty puzzles, say the count plainly, name the grid sizes it really holds, and say how many puzzles sit in each difficulty band claimed, because a buyer who bought gentle grids has not bought sprawling challenge grids. State the cell system honestly: the grid starts empty apart from its clue cells, clue cells are never shaded, every other cell is either shaded as wall or left unshaded by deduction, and the numbers inside a clue cell are read in order around that cell. Say whether full answer keys are included for every puzzle and where they live — at the back, on the following page, or in a separate file — before purchase. State the page size, whether it is meant for print, screen, or both, and flag any variant rules in plain language before a buyer meets them mid-book. Put the rules on an instruction page the buyer opens first — the wall rule, the forbidden-square rule, how a clue sequence is read, what the difficulty bands mean, and how the keys are marked — before puzzle one. Check the solving as the buyer experiences it: download it fresh, print it, and paint one full section with no help from your source files.
Every claimed puzzle has exactly one solution: the tapa contract
A tapa buyer cannot improvise around a defective grid: a wrong clue number, a second valid wall, or a layout whose only legal walls break into pieces or close into a forbidden square makes the book untrue in a way patience cannot fix.
Hold every puzzle to the contract, with no loose readings. First, the wall rule: in the finished puzzle, all shaded cells form one single connected wall. Describe that connectivity carefully in your own instruction page and hold your checks to the same wording you print. As tapa is commonly stated, wall connectivity is the shaded cells forming a single connected group rather than several separate shaded islands, and a solution that leaves two shaded groups with no connection between them has failed, however neatly each group obeys its local clues. Second, the forbidden-square rule: no two-by-two block of shaded cells appears anywhere — not in a corner, not at an edge, and not buried in the middle of a long wall run. Four shaded cells meeting in a solid square is a defect even when every clue nearby is satisfied. Third, the clue-cell rule: a cell that carries a clue is never shaded and never counted as part of the wall. It stays open for the whole solve, and the wall must flow around it, never through it. Fourth, the clue-sequence rule: look at the cells surrounding a clue cell — up to eight neighbouring cells, fewer at an edge or corner — and read the shaded cells among them in order around the clue. They fall into consecutive blocks, with at least one unshaded neighbour separating one block from the next. The numbers printed in the clue cell give the lengths of those blocks, in order. A clue printed as two then three promises a block of two shaded neighbours, then at least one unshaded neighbour, then a block of three shaded neighbours, in that order around the cell. A clue with a single number promises a single block of exactly that length and no other shaded neighbours. Some books also use a question mark in place of one value to mean a block of unknown length; if you use that variant, say so on the instruction page, keep it simple, and never let a buyer discover the convention by failing a puzzle. Fifth, exactly one solution. Not one intended wall — one wall, full stop. A grid that admits two different shadings, each obeying every rule, has broken the promise on the cover — the buyer who finds the second wall has discovered the puzzle was never verified.
Prove all of it mechanically. Re-solve every puzzle from the final laid-out pages, checking that the shaded cells form one connected wall, that no two-by-two block of shaded cells appears anywhere, that every clue sequence matches the blocks actually surrounding its clue cell, and that no second shading satisfies the same printed clues. Do not trust the generator or your memory. Numbers mutate in editing — a digit changed in resizing, a clue shifted by one cell in re-layout, two values swapped inside a crowded clue cell — and a puzzle valid in the spreadsheet can be broken in the PDF. Run the mechanical checks that catch what eyes miss: a wall-connectivity check that walks the shaded cells of the finished grid and requires the walk to reach every shaded cell, so no shaded island survives unnoticed; a no-two-by-two check that inspects every possible square of four cells in the grid and requires at least one unshaded cell in each; and a per-clue block-sequence verification that, for every clue cell in every puzzle, lists the shaded blocks among its neighbours in order and compares that list, length for length and in order, against the printed numbers. Every clue, every puzzle, no sampling. Test the truth the way a buyer will: solve each puzzle cold, without the key, then run a second, independent pass — a solver program solving from the printed clues alone, a fresh manual solve on a different day, or a different solver altogether — and confirm both passes land on the same single wall. Re-running the same generator is not a second pass. Note every puzzle where it found a second solution, a wall that cannot stay connected, a forbidden square forced by the clues, or a clue sequence no legal wall can satisfy — that list is your defect list, fixed before release.
Complete and consistent: puzzle, clues, and answer keys must agree
A tapa book is a three-way system — the clued puzzle the buyer paints in, the shaded-and-open pattern that solves it, and the answer key that proves the pattern can be reached in exactly one way — and any inconsistency among the three reads as carelessness in a product whose whole value is trust.
Deliver the set in a stated, checkable order. Number every puzzle, keep the numbering identical on the puzzle pages, in any contents or difficulty index, and in the answer section, and never renumber silently between editions: a buyer halfway through version one who downloads your corrected version two must find puzzle forty-one in the same place, or your fix has cost them their progress. Keep the clues identical everywhere: the numbers on the puzzle page are the clues the key must honour, a clue cell printed with a two and a three must be the cell the key treats as a two and a three, in that order, and a key that paints a wall legal for different numbers is not a rounding difference — it is the book arguing with itself.
Run the puzzle, clue, and key consistency checks as their own pass, separate from solving. For the shaded pattern: every wall cell is marked unmistakably in the key, no cell is half-shaded or ambiguous, and every clue cell on the puzzle page is open in the key — a clue cell shaded in its own solution is a defect on its face. For the wall-connectivity check: walk the shaded cells of the key, require the walk to reach every shaded cell, and require the total reached to equal the total number of shaded cells, so a disconnected wall in the key can never pass as a quirk of style. For the no-two-by-two check: inspect every square of four cells in the key and confirm at least one cell in each square is unshaded, including squares that straddle clue cells, edges, and corners. For the per-clue block-sequence verification: take every clue cell in turn, read its neighbouring cells in order around the clue, record the consecutive shaded blocks and the unshaded gaps between them, and require the block lengths, in order, to match the printed clue exactly, with at least one unshaded neighbour between successive blocks. For the grid itself: row and column counts match the stated size, every clue cell carries its full set of numbers with none missing and none duplicated by a layout slip, no number is clipped at a layout seam or sitting half in the neighbouring cell, and the outer border closes cleanly.
Check every key against the final printed pages, because a re-laid-out page keeps its clean looks while silently invalidating the wall promised. Print each solution as the completed grid in the same size and orientation as the puzzle, wall cells solidly dark and clue cells clearly open with their numbers still legible, so the buyer can compare cell for cell — this cell shaded, that clue reading true — and arrange the keys so checking puzzle twelve does not force a glimpse of puzzles thirteen and fourteen. Never let the key be checked by trusting it. Check it by auditing against it: re-run the three checks on the key itself — connectivity of the wall, absence of any forbidden square, and the block sequence at every clue cell — and verify the key shading is a shading of the puzzle actually printed above it, clue for clue and cell for cell. If you include worked help, every claim in it must be true of the printed puzzle it describes. Test the way a buyer will: pick puzzles from the back half, solve them cold, then compare against the printed key cell for cell.
Difficulty is more than grid size
Grade difficulty honestly, and by more than how many rows a grid has. Four levers work together in tapa, and a book that labels its bands by size alone will mislead the buyers who trust its labels most.
The first lever is grid size. A small grid gives the buyer fewer cells to reconcile and fewer clue neighbourhoods to keep consistent at once, and it is where a new solver learns what a forced shaded cell feels like. A larger grid multiplies clues, hides consequences further from the cell that causes them, and asks the buyer to hold several half-painted wall sections in mind at once, remembering which sections must still join and which must never touch in a square. Size matters — it just does not decide difficulty on its own.
The second lever is clue density: how many clue cells the grid carries, and how they are spread. A dense grid pins the wall down in place after place: nearly every stretch of wall is measured by a nearby clue, deductions arrive early, and the solve feels guided even when the grid is large. A sparse grid is the opposite — long stretches of wall with no clue nearby, where the wall route must be inferred from connectivity alone and a wrong turn may not surface for many cells. Say which pattern your harder bands use, because a sparse medium grid can honestly outrank a dense large one.
The third lever is wall length and tightness: how long the finished wall runs, how often it turns, and how narrowly it threads between clue cells and edges. A short, open wall leaves room to correct course. A long, tight wall that doubles back, squeezes through one-cell gaps, and repeatedly approaches itself without forming a forbidden square asks for planning ahead, because one shaded cell placed for a local clue can seal off the only route the wall still needs. Tightness, not just length, makes a late-book tapa feel relentless.
The fourth lever is the ambiguity of clue neighbourhoods: how many plausible block arrangements a clue allows before its neighbours are settled. A single-number clue with a short block is often settled locally. A clue carrying several numbers can fit its neighbourhood in more than one arrangement until distant cells decide it, and overlapping neighbourhoods can each look satisfiable alone while only one joint arrangement satisfies both. Grade on what actually made a puzzle hard — size, clue density, wall length and tightness, and neighbourhood ambiguity — and never promote a puzzle to a harder band simply to fill a quota.
Print, pencil, and eyes: what the page does that the preview cannot
The grid a buyer approves on a bright phone screen is not the grid they solve on paper, and tapa makes the gap physical: the buyer shades cells with a pencil, marks cells proven open, and erases shading that breaks the wall or closes a forbidden square — and the printed page is their only truthful reference once the screen is off. State the printed page size, and tell the buyer to print at full size, actual size, with fit-to-page off, because a shrunken page squeezes cells until a shaded cell, a half-erased shading, and a tiny clue number stop being distinguishable.
Hold shaded-cell solidity without mercy, in the key above all. In the answer key, a wall cell must print as solidly, evenly dark from edge to edge, with no pale centre that could be misread as open, because a misread key cell also falsifies every clue sequence whose neighbourhood includes it and can make a connected wall look broken. On the puzzle page the opposite discipline applies: a cell that starts open must stay visibly open, its surroundings crisp, and the buyer's own pencil shading must read as a deliberate mark, not a printing flaw. Check a printed key beside a printed puzzle page in poor light before any buyer sees it.
Print the clue numbers with honest legibility, and remember that tapa clue cells are the hardest numbers in puzzle publishing to keep readable. A tapa clue cell often carries several small numbers inside a single cell, arranged together in the space other puzzles give to one digit. At final print size those numbers must still read instantly and in order, because the puzzle turns on telling a two from a three at a glance, mid-solve, and on never losing which value comes first around the cell. If the numbers are small, give them weight; if a cell carries three or four values, test that exact cell size on paper, not on a zoomed screen. A clue that can be misread is not a harder puzzle — it is a different, usually impossible, puzzle wearing the first one's numbers.
Print grid lines with a weight that does the job and no more. Lines must stay visible so rows, columns, and wall continuity can be traced without drifting into the next cell, heavy enough that a number in a clue cell stays readable under pencil shading beside it, and never so heavy that the lines crowd the numbers themselves or make a lightly shaded cell look pre-shaded.
Leave honest pencil room. A buyer needs room inside each cell to shade it fully, mark a cell proven open, and erase cleanly without tearing away neighbouring clue numbers or leaving a ghost dark enough to be mistaken for wall later. If you sell a large-print edition, say what is actually larger — cells, numbers, grid, keys — and solve one full printed puzzle yourself before deciding the page is ready.
Licensing a tapa book is licensing the solving
Solving feels private, yet the same file could supply a family's evening, a teacher's logic lesson, a puzzle club, a care home's activity hour, or another seller's new puzzle book. If your terms do not answer those uses in advance, the buyer either asks and waits, or guesses and infringes.
Answer the questions buyers actually ask, in short sentences inside the download, and keep personal solving, group use, and republication separate. May the buyer print the book more than once for their own use — a second clean attempt after a solved copy covered in shading, open-cell marks, and erased trial walls? Say yes or name the limit, because a painted grid is a spent grid, and reprinting is the genre's normal life. May they print copies for family, a classroom, a club, or a care setting, free of charge? Each needs its own answer, and if group use needs an extended licence, name and price it before purchase. May a buyer share their own solved grids — a photograph of a finished wall, a video of a solve? Say so exactly, separate from the file itself: what is plainly out is reselling or giving away the digital file, and republishing the collection as a competing book where your terms prohibit it.
Handle provenance with particular care, because tapa puzzles arrive by two very different roads. If your grids were generated, say what generated them and what check confirmed each one — the solver pass, the uniqueness test, the wall-connectivity check, the no-two-by-two check, and the per-clue block-sequence verification — and keep those records, because the generator said so is not a verification trail. If your puzzles were hand-made, or licensed from a collection, state the terms in your records — whether they may appear in future volumes, whether the licence covers the answer keys as well, and whether the grids are cleared for republication — and do not imply a right to republish puzzles you have not cleared.
A broken wall cannot be un-painted: the errata rule and the version line
Every genre has an error that cannot be taken back, and a tapa book's is trust in the clues: a buyer who solved honestly for forty minutes and then discovered the only legal-looking wall breaks in two, or closes into a forbidden square, cannot get that solving back. A wrong clue voids the puzzle it sits in — not the page, the puzzle — and the correction is not a footnote. It is a replacement, sent as a corrected file to every buyer.
Run corrections as a stated system, not an apology when someone complains. Your errata rule is simple and it covers the failures only this genre makes: a second solution — a puzzle whose printed clues admit a second, different legal wall as well as the key's — a disconnected wall — a solution whose shaded cells fall into separate groups with no connection joining them into one wall — a forbidden square in the key — a two-by-two block of shaded cells appearing in the printed solution — or a clue sequence mismatch — a clue cell whose neighbouring shaded blocks, read in order, do not match the printed numbers in lengths or in order — each triggers the same response: a corrected file to every buyer. Keep a public errata list, dated by edition, naming the puzzle numbers affected and what changed — a puzzle replaced because it admitted a second valid wall, a puzzle replaced because its wall disconnected, a key re-drawn because it held a forbidden square, a clue corrected because its sequence did not match its solution, a difficulty label fixed — so a buyer with an early copy can check in minutes before they solve, the only moment a correction can still save the painting. Version the file: put the edition and date inside the download — the version line — and never silently change a number or move a clue by a cell. When one wrong clue is confirmed, re-solve the whole book, not only the reported puzzle, and say what the re-checking covered. Notify past buyers, keep the corrected edition as the only file new buyers receive, and be honest in the listing while a known defect is unfixed.
A short launch checklist for tapa book sellers
Before the listing goes live, work through the book as a buyer will — on paper, in order.
- Download your own file fresh and print it at full size, actual size. Solve from that printout only, pencil in hand, and shade, mark open, follow clues, and erase as you go.
- Re-solve every puzzle mechanically on the final laid-out pages: the shaded cells forming one connected wall, no two-by-two block of shaded cells anywhere, and every clue sequence matching the shaded blocks around its clue cell in lengths and in order. Fix every broken puzzle before release.
- Check uniqueness by a second, independent pass, not by trusting the generator: confirm no puzzle admits a second valid wall, and confirm every printed answer key is the one solution.
- Audit the three-way agreement on the final pages: puzzle, clues, and answer key must agree exactly — the wall-connectivity check reaching every shaded cell, the no-two-by-two check on every square of four cells, and the per-clue block-sequence verification at every clue cell. Solve a sample from the back half cold, then compare against the printed key.
- Confirm the listing states the puzzle count, the grid sizes, the difficulty bands and what they mean — grid size, clue density, wall length and tightness, and ambiguity of clue neighbourhoods — and where the answer keys live.
- Check the page with your own eyes on the largest grid: every clue number legible inside its cell at print size, including cells carrying several small numbers, key shading solid enough to survive a tired printer, no pale key cell that could be misread as open, grid lines clear without crowding the numbers, and real pencil room to shade a cell, mark a proven open cell, and erase cleanly. If a number can be misread, re-laying out beats a buyer's frustration.
- Put the licence in the download in short sentences — repeat printing for personal use, family, classroom, club, and care-setting use, sharing solved grids, puzzle provenance for generated and hand-made grids, and the file-sharing line, each answered separately — and start the errata page and the edition line inside the file on day one.
A closing note on delivery: FileFreight is one option for that delivery step: flat-rate download delivery plus license keys on your own Stripe, so payments settle into your own Stripe account — it is not a tax or merchant-of-record service, not a payment processor in its own right, it does not handle tax, it has no importer, and it is not a Shopify app. Check every clue, paint only what the rules force — and let delivery be the one part a solver never thinks about.
Top comments (0)