Someone makes a cardboard version of a shallow organizer for an art cooperative. It slides neatly into its spot on an empty desktop. The team takes a picture. Now load the desktop the way it really looks on a busy day. Put the proof stand beside the organizer, leave the pencil holder where it lives, and ask the person stocking cards to remove the organizer and carry it to the back shelf. Holding one cardboard wall at a time, it still fits. Holding two protruding hand grips as the traveler will have to, it catches the stand. The original photo wasn't wrong. It just proved something much narrower than the people looking at it assumed.
A real object has to arrive at its resting spot, get used there, and eventually leave. The static outline only covers one of these. A cardboard stand-in can reveal what happens at the bend before the door, when a lid opens, when a stack is drawn out, or when somebody needs a different grip. It can make the problem cheap to discover before a vendor cuts anything.
This is a guide to simple geometry and handling questions for light, unpowered, low-consequence things such as desktop displays and paper organizers. Cardboard is not a load test, an accessibility audit, a materials certificate, or a model of equipment clearances governed by safety rules. Keep prototypes clear of access routes and moving equipment. Don't simulate dangerous handling or use these tests to release something that bears weight, operates near heat, plugs in, or protects people. The benefit is that you can answer one question without pretending you've answered ten others.
Select the move you want to learn from
“Make a prototype” invites people to build a convincing-looking model. A better instruction is “find out whether opening the organizer's lid forces us to move the existing display stand.” Write the move, the normal starting condition, and what you would observe if it failed. The Stanford d.school No-Build Hack encourages small prototypes that test an idea and teach something. Its supplies list includes ordinary paper, tape and cardboard. Those are useful precisely because they make adjusting the first theory cheaper than making it look done.
Use a verb that can be acted out. Carry, turn, set down, slide, grip, open, withdraw, and restock are different actions. Each has a particular contact point and a particular position. A prototype answer “yes, it fits” can easily collapse several of them into a claim you never observed. “One person can slide it onto the shelf without touching the backing stand” is a useful statement only if the test actually included the backing stand and one person tried the specified movement.
Don't fixate on the most dramatic failure first. A beautiful organizer can make it all the way to its position and still be irritating because someone has to remove it before replenishing cards. Ask the person who normally works with the object which move they do ten times per day and which they do twice per year. Found a bad twist at the rare access point? Record it, but fixing the frequent move may matter more. If a move must meet a real accessibility or safety requirement, treat this exercise only as an early observation, followed by the correct evaluation.
Write what would change your decision. If the object can't clear the stand by the agreed amount, would you reduce the width, move it elsewhere, swap the grip design, or change the orientation? Different choices demand different levels of fidelity. A flat sheet suffices to compare footprints in several places. A sidewall and stop tab are more relevant when deciding if a hand can reach the back. A color sample may be irrelevant. Underline the consequential question and make a model whose highest fidelity lives right there.
Keep the outside honest and the inside cheap
For route work, the basic artifact is the outer envelope. That means the outermost thing that could collide, not just the convenient width and depth on a product sketch. Grip tabs, an overhanging lid, feet, a nameplate, and a strip of soft trim could change a measurement at the point you care about. Tape paper flags onto the corresponding model locations, mark their intended projection, and mark which flags aren't realistic enough to touch. If the object is expected to be a square corner rather than a flexible curve at the bottleneck, give that corner a simple stiff fold or record the limit of your result.
Label the reference faces. A protocol “38 millimeters from the front” will be read against whatever surface looks like the front to the next person. Physical arrows and a single “face A” marking can eliminate that ambiguity. Distinguish the thick wall from the clear opening. Cutting a paper hole the same size as an external dimension doesn't model an internal opening if the production wall is thicker, the label wraps over the edge, or hinges occupy part of the aperture. Write dimensions at the opening where they matter instead of trusting one universal outline.
Decide how fine the model needs to be. Ordinary tape and corrugated sheet can give a convincing answer about which of two positions saves 100 millimeters. Around a two-millimeter interference it may have little power to distinguish success from failure. Your mockup itself has build variation, a cardboard edge may squish, and the fixed shelf isn't always perfectly square. Nothing in a caliper reading turns the test into a precision fit. Give close cases a threshold for escalation to a better model, a direct measurement of the unusually tight spot, or a material-appropriate pilot. Don't trick yourself into thinking that iterating indefinitely in cardboard will eventually make cardboard behave like the product.
Thin masking strips and taped-on labels can be changed while retaining the overall model. If you do change them, revise the ID and take a new note. Some groups redraw a whole mockup from scratch for every idea, losing the benefit that rough materials were supposed to bring. Others silently bend a corner and trade away the baseline. The compromise is to keep one identified body and a named set of pieces you can add or remove. A photo with an ordinary ruler gives sense of scale; the actual dimensions and version under test still belong in the record.
Use three kinds of fit evidence
The endpoint check asks what the thing occupies at rest. Set the model down without forcing it, compare against the planned surface, and identify where hands, other objects, and an occasionally opened drawer go. Look for overlap that is hidden at one level. A narrow desktop foot can clear everything at the base while the organizer top projects over an adjacent paper stack at grab height. A paper outline is an effective first screen for the clear surface. Add only enough height or overhang pieces to see what that outline missed.
The route check starts somewhere other than the endpoint. Starting with an empty desk and raising the organizer straight into place from above may be possible only in the mockup session. If stock is replenished from the side, start there. A synthetic bench test might follow a sequence: pick up from normal storage, shift left to clear a stand, turn the long edge forward, then lower onto two marks. Try the reverse route too. Lack of a sensible place to put down the model halfway can change the answer even when it could be passed between two imaginary people in a drawing. If the movement requires tools, installation, real load or cooperation beyond safe tabletop objects, stop and evaluate elsewhere rather than treating the sketch as an operational demonstration.
The use check covers what changes after arrival. Model the sweep of a lid, the extension of a removable paper liner, and the space one hand needs to pick up a card without tipping the stack. A strip of stiff paper attached at the intended pivot can show a simple lid obstruction without representing a real hinge. A larger strip can mark a padded hand zone but cannot validate movement or dexterity across all users. Pay attention to how people actually use it rather than explaining how they are supposed to. If you have to teach every person “you need to turn it upside down and hold this tab back with a pencil,” that may be a finding.
A clean photograph isn't the entire result. A phone clip of a specific snag can be helpful if everyone agrees, but the most useful output is the measurement or contact site the clip points to. Record “upper left hand tab touched base of card stand while turning inward toward the desktop.” Do not write “failed to fit” and expect the next maker to identify which quarter of the motion mattered. IDEO.org's Design Kit guidance on running prototypes emphasizes fast feedback and documenting what you learned. Adapting that idea here means carrying a record of the event, not holding onto every demonstration.
A fictional example: the last thirty millimeters
The cooperative's tabletop paper organizer has two candidate designs. Model R1 is 420 mm long and 240 mm front to back, with two projecting side grips. The desktop slot is 465 mm wide where the organizer sits. Both models can use the slot at rest. To get there, you approach from the front and move around the existing proof stand. The narrowest place isn't along a straight portion. It is where the right grip arcs toward the stand during a turn.
The team uses a small piece of bent card to represent the widest proposed right grip and leaves the other controls out of the model. They mark one regular motion start, then repeat the move with the mockup from two natural starting positions. On a third attempt a person succeeds by letting the cardboard flex around the stand. That isn't “pass on third attempt.” The flexible corner is a limitation of the trial and the equivalent rigid part could not be assumed to follow. They label the contact corner, record the alternate path as a possible change to the setup, and preserve an image of the model position during the snag.
R2 uses a pair of top grips within the same outer plan. A new mockup gets comfortable turning clearance. Unfortunately its simulated front flap sweeps through the space in front of the adjacent proof stand. The team asks why the flap opens full width, then sketches a shallow fixed front with two small openings for picking cards. That resolves the observed flap obstruction but introduces a question the low-sided mockup didn't answer: can somebody withdraw paper of all the intended sizes, including the thick sample pack? They put differently sized scrap bundles at the intended opening and test removal. It is better to identify and run that test than to congratulate the new grip design for solving a different problem.
| Trial | Question | Observation | What changes |
|---|---|---|---|
| R1, bare desk | Sits within assigned outline? | Yes, static endpoint only | Keep desk outline |
| R1, normal desk | Can turn in without contact or flex? | Right grip touches stand | Change grip or path |
| R2, normal desk | Can turn in and open flap? | Turns clear; flap meets stand | Revise front |
| R3, fixed front | Is largest intended paper pack removable? | Next targeted test | Not ready to order |
None of the numbers are a general rule. A custom space or model has its own dimensions. The shift in the trial structure is the point: add a piece only when it can settle a specific question and leave the untested feature in plain view.
Estimate room without promising a fit
For a simple straight pass between two fixed boundaries, arithmetic can help describe what you still need to test. If a measured gap is about 450 mm and a grip-to-grip envelope is about 430 mm, there is about 20 mm total nominal clearance across both sides. Ten per side is the centered ideal, not the clearance you get automatically. A real sample may be slightly wider, the doorway may be narrower at a different height, and the hand carrying it may extend outside both edges. Even modest uncertainties can eat this number easily.
Write a small “worst plausible” note using evidence rather than a magic safety factor. For a fictional rectangular gap with an ordinary shop-measuring estimate of 450 ± 4 mm and possible part-and-grip width of 430 ± 5 mm, the lower gap estimate is 446 and the upper part estimate is 435. The simple difference suggests 11 mm total under those stated assumptions, before any room for a hand, tilt, or rotation. These are example bounds, not certified tolerances. If the route depends on turning through that space, the changing swept envelope still has to be evaluated directly. No one needs to report “11 mm guaranteed” on an object that hasn't been made.
Don't try to simulate a heavier object by loading a cardboard test shape. Weight, stiffness and durability are different questions, and flimsy prototypes failing unexpectedly don't teach what a better design would do. For an ordinary paper artifact, test available hand space with the stand-in itself, and record that you're not testing loaded handling if you are not. The next evaluation might be another paper model with a different shape; a stronger specimen with a qualified handling procedure is a separate decision. The prototype record should make that difference impossible to miss.
Write a useful closeout instead of a big report
A closeout can fit in a few lines: “R2, envelope dimension check only, normal desk arrangement, two operators handled empty mockup. Right grip clearance improved. Front panel sweep interferes with stand. R3 test largest pack with the revised front. No production release.” Attach a view of the two interference points if useful. If a model tells you nothing because the station is still being designed, stop adding details. Look for the next input that matters, such as the dimensions of the reference object or a representative desktop layout.
Use the Smithsonian Designing for Change toolkit as a broader model-building resource. Its examples include simple materials to communicate a proposed design. The Stanford Experiment Expedition separates deciding what to test, capturing observation, and choosing what to change. You don't need to adopt an entire workshop language to put those distinctions on a piece of masking tape. Make the trial live at the actual scale and place where your mistake would show up.
Discard the cardboard when it becomes a distraction, not when it gets ugly. An unpainted edge with a useful mark is often better evidence than a beautiful finished render. Save the dimensions and the result, identify what's next, and let the mockup earn its keep on the short trip between two places rather than merely looking good in one.
Build a path story before building a shell
Suppose the fictional workshop has an organizer for index cards. It is light, unpowered, and lives in a shallow open cubby when not in use. During an exhibit, someone carries it to an adjacent desk, opens the organizer's flap, then slides it back into the cubby. A cardboard model can answer whether certain fixed boundaries intersect that intended motion. It cannot show that the actual finished material will resist damage, fit every similar cubby, or be comfortable after repeated use. Write down the intended operation and what the disposable model does and doesn't represent before cutting any pattern.
Draw the normal starting point, the direction the person faces, which parts are open, and the space already occupied by the ordinary fixture. In the imagined workspace, a small sign stand stays on the desk. It is not reasonable to remove the stand to make a demonstration pass if a person would normally leave it there. A movable object can still be part of the stated starting condition. The effect of changing that condition can be tested separately: one trial with the stand in its approved location, and another with a proposed rearrangement. These answer different questions even when the outline of the model has not changed.
Break the move into three or four useful frames. At the cubby, the model is presented with its long face visible. During the move its left corner enters first while the tab stays accessible. At the end, the lid lies flat and the intended gripping face is still available for removal. Make those positions observable rather than prescribing a single exact angle if normal users would choose several. Ask whether the product needs one documented path or reasonable tolerance for common variations. A puzzle demanding one secret orientation might technically fit and still be a bad choice for the shop.
Don't confuse photographs of the finished position with evidence about motion. A real-sized model can often be placed straight into an otherwise empty tabletop footprint because the material overhang is flexible or the person knows which part to remove first. That is neither the normal starting condition nor the normal move you meant to test. Capture what happens along the actual path as a separate ordinary observation. A useful failure report names where interaction stopped; it does not use an elaborate visual rendering to imply a move that nobody can repeat.
For a first pass, specify one route and one objection. “The left-hand tab hits the stand as the box twists before it clears the cubby” is enough to choose a change. Do not start by testing every possible orientation of every environmental variable. Later a second person may uncover another practical route; distinguish it from a deliberate attempt to cheat the fixture. Good modeling reveals which design assumptions were convenient fictions rather than simply punishing someone for moving a prototype differently. The Stanford d.school's No-Build Hack is a primary teaching resource for reducing prototypes to questions rather than polish.
Capture what was represented and what was not
Use a model that is realistic about the external dimensions that govern the test. A drawing taped flat on a shelf can answer whether two resting footprints overlap. It cannot represent a raised corner that swings into a sign, a protruding lip, or the amount of space a hand needs to keep hold. A simple shell can carry the outside points without pretending to represent the interior mechanism. Set a few clear references and mark which dimensions were copied from known specifications, which are proposed, and which remain rough. If the actual model's folded layers add thickness, record the difference instead of making it disappear in annotations.
Cardboard's useful weakness can also conceal failures. It bends past narrow points and springs back, whereas a finished shell may be more rigid. In a toy trial suppose the wall flexes noticeably when the model clears the cubby, leaving no visible crease. That still counts as evidence of interference unless the actual product intentionally allows the same flex with suitable evidence. Don't treat flex clearance as design clearance. Similarly, a crooked fold can create a false snag not present in the intended geometry. Check the model itself before assigning cause. An interference observation is strong evidence about that trial and not automatically a verdict on the entire proposal.
Label geometry features that are not yet known. If the choice between two commercial holders is unresolved, fabricate a measured neutral envelope for each proposal using only benign materials. The test can compare base shapes while leaving hinge behavior and grip unresolved if those have not been modeled. A note that says “tried the larger outer profile, lid motion not represented” helps a reviewer select the next high-value question. It is far more useful than translating a flat mockup into a confident statement that the drawer works. One model can answer more than one question only if the relevant geometry actually appears in it.
Use a known, permitted environment in the workshop or recreate the landmarks on a spare desk. The latter is useful when the normal bench can't be blocked for a trial. If a key fixture has been modeled rather than directly used, label the simulation and retain the changed constraints. Nobody should supply access to a closed, raised, or unsafe area just because the model is made of harmless cardboard. The use case deliberately avoids heavy loads, powered mechanisms, hazardous clearance, fire or building approval questions, and operational machinery. A disposable fit model is useful for modest tabletop geometry precisely when making a poorly chosen motion safe is not the challenge.
Test the turn, the return, and the ordinary mistake separately
Prepare three passes for the invented index-card holder. In pass one, the maker tries the normal insertion from the assigned desk position. It reaches its final spot, but the inspection shows an upper corner briefly meets the edge of the sign stand. Check whether the maker changed the stand's location mid-move or bent the model by squeezing it. Document the contact and any visible deformation in its own line. A pass/fail result tied to the final spot alone would report success, but the observed problem occurred during movement and will likely matter on the actual product.
In pass two, start with the item properly stored and try removing it under the normal hand placement rule. Maybe it's easy to slide into the cubby using its broad face and hard to get a hand onto a recessed holder once seated. A perfectly symmetrical diagram might falsely predict that withdrawal is just reverse insertion. In use, doors and hands are differently positioned on each side of the operation, and guiding the model from above can be possible when placing it but unavailable later. Check whether the planned handhold is visible and ordinary to use; don't force fragile grips or reach into an unsuitable work area for the sake of a clean demo.
In pass three, let a second adult tester try without coaching beyond the tasks someone would actually receive. The designer may have memorized the one path that avoids a snag. A reasonable second orientation may encounter a different conflict. Record the attempted path rather than classifying the person as wrong for doing something plausible. The test can result in a marked orientation, a changed hand tab, a way of leaving more margin, or accepting the current design only under clearly stated conditions. The correctness of that design choice depends on who actually uses it and the room available; a paper mockup makes the difference visible.
Try what happens after placement. Can the flap be opened far enough for the ordinary card change without knocking into the sign, and can the model be put away with the flap safely in the modeled position? Some projects need a separate deployment footprint and a storage envelope. A physically settling model can meet one and fail the other. Keep a small checklist with observation frames such as begins from permitted base orientation, turns clear in ordinary path, grip remains usable from stored position, and reaches intended open presentation state. None of those rows asserts load-bearing or durability performance.
A useful test purposely keeps an apparently good setup alongside a known bad one. For example, move the synthetic stand just enough that the conflict appears at the same part of a repeatable turn, then restore it to the defined position. This is a controlled comparison, not a recommendation to move whatever blocks someone's workplace. If the detection method can't distinguish a clear interference case from the intended fixture, refine the note-taking or prototype geometry. The IDEO Design Kit prototype guidance emphasizes learning from concrete trials; here that means saving which expected observation changed when one fixture detail was deliberately changed.
Read modest dimensions without claiming a guaranteed tolerance
In a toy example, the straight opening is roughly 250 millimeters, and the model's relevant outside measure in the same plane is roughly 240. Ten millimeters as a simple difference is not a proof that the item can turn, that the full opening is that size along the entire route, or that someone has ten millimeters where their hand actually needs it. It is a useful preliminary clue. At a particular point the corner projection may sweep across an occupied region, or the space above the model may have another constraint. Describe the dimensions at the exact positions they represent. When a measurement is approximate, record it as approximate rather than collecting more digits than the setup justifies.
Compare a revision using an independent blank model where useful. One trial changes the hand tab from projecting to recessed, and the new model still contains the old dimensions everywhere else that matter for the comparison. Ask whether the impression that it is easier actually comes from the changed feature or from a new path learned on the first try. Counterbalance the sequence between two observers if it's meaningful and reasonably simple. No elaborate statistics are needed. The purpose is to notice when the facilitator changed three things and then attributes the whole improvement to one.
Keep the margin discussion qualified. A rough model's seam, environmental fixture, and hand grip each contain uncertainty. It is reasonable to use a readily observable threshold as a screen for this low-risk project, but not to generalize it into a tolerance for real fabrication. A built sample may introduce process variations that the model doesn't represent. If the real decision depends on a narrow difference or a predictable result across many locations, choose a more appropriate review process. The article's examples are a practical inquiry about a small benign object, not a method for critical clearance or formal ergonomic certification.
Write down a useful rejection threshold too. “Corner and sign touch under ordinary setup during two observed trials” is sufficient cause to reconsider the model in the fictional pilot, even if both determined makers can nurse it through with modification. “Looks cramped” is worth noting but gives another team less to act on. Capture where and whether the contact persists, then change something small enough to learn from. If you are unsure the clipping came from a warped prototype, rebuild only the cheap outside contour and compare rather than announcing a final failure. Better evidence beats cleverer descriptions of the same poorly observed demo.
Choose the next model from the evidence you actually gained
Suppose the cardboard trial suggests shortening one outside corner. Three decisions remain: whether the material and design allow that shape, whether the same movement still suits a real user, and whether the actual finished object offers the promised fit. A revised paper shell can address the first spatial question cheaply and show whether it is worth a more representative low-risk model. It should not jump straight to such claims as holds its shape indefinitely or has an ideal grip. Think of the next model as buying another distinct answer. The Stanford Experiment Expedition is useful for naming what an iteration is supposed to settle rather than making prototypes ever more expensive because the process feels unfinished.
Document the quiet success case. The intended path works in two ordinary observed attempts under the documented layout, the mock-up has no recorded interference, and undoing the move is possible from its stored state. That supports continuing to a next approved step. It is not a public claim that every user can use it or that any other workstation has equivalent clearance. If the project only needs one ordinary desk location and the future sample has enough independent evidence, an unusually broad route survey isn't necessarily useful. Let intended use and consequence determine how much verification is needed. A small experiment should be proportionate to the decision it serves.
Document a scope expansion separately. An organizer first built for a particular exhibit desk might later be offered to librarians on other floors with different shelving and permissions. A new setting deserves a fresh record of the actual constraints and what kind of model would represent it. There is no guarantee that a model from the first fixture applies merely because both shelves share a name or were bought in a similar decade. Ask the ordinary question of which starting state, intended movement, and remaining access actually apply. Comparing a second environment is an experiment, not an extension of authorization inferred from a cardboard item's success elsewhere.
Share a brief handoff with the next maker. It can say what model revision was used, where its geometry came from, what ordinary move was attempted, what touched what, and which plausible change should be tried next. Add the limits: no weight, durability, safety clearances, or long-term fit verified; no intervention to the actual work setting during the trial. A few marked photos or measured drawings can be useful for resolving ambiguity, but don't upload faces or proprietary environments merely to show a toy example. The Smithsonian facilitator toolkit illustrates simple model construction to communicate an idea. In the shop that communication is strongest when the model's physical claim and its limitations travel together.
One small test can tell you the project doesn't need any new object at all. If a friendly rearrangement under proper ownership solves the thing you set out to learn, the mockup is still a valuable result. The point was to understand a use and its constraints, not justify constructing an organizer. A lasting practice is to let a modest paper trial teach a narrow practical fact and keep that fact visible as the design changes. An attractive shape at rest is easy to exhibit. A usable object has to make its ordinary trips, and the evidence for those trips is worth saving before you make the final piece.
Let a move fail for the right reason
Suppose two careful observers disagree about a contact. One sees a top corner touch the sign; the other thinks the model's temporary tape lifted during the turn. Their visual judgments need not decide the complete design. Check the model's shape before and after the attempt and repeat under the same approved setup with the tape repaired or deliberately excluded when it does not represent a final feature. If the touch disappears, report what changed without certifying the production version. If the contact remains at the same point, mark that location in a plan view and show which proposal you will test next.
In the fictional fixture someone also suggests removing a little pad from the base so the model passes cleanly. The pad is part of the proposed footprint, so omitting it during the ordinary run would remove the very constraint under investigation. A run stripped of the pad can still teach which geometry causes the interference, if labeled as a diagnostic alteration. The valid design test will either preserve the pad or describe a genuinely revised part. Keeping those results on different lines makes it possible for a future reader to learn from the controlled change without inheriting a false passing report.
Pair each outcome with a plain sentence about the next decision. The simple pass means a better model is worth checking; the observed top collision means review a constrained revision; uncertainty about the model shape means improve the fixture before judging the product. These sentences are easier for a small team to act on than a numerical fit score that combines all three categories. A carefully observed disagreement can improve the next experiment without anyone having to find a more dramatic failure to justify the work.
Sources and further reading
- Stanford d.school: The No-Build Hack, quick material prototypes to test one question.
- IDEO.org Design Kit: Build and Run Prototypes, capturing feedback and iterating.
- Stanford d.school: Experiment Expedition, test planning, capture, and decisions.
- Smithsonian: Designing for Change Facilitator Toolkit, low-cost model-building and communication.
AI disclosure: Drafted with AI assistance for Apiary. The organizer, model trials and dimensions are fictional; no physical prototype was built or tested.