The most expensive dimension on a drawing is the one that is not there. The sheet metal worker already has the blank on the press brake, looks for the position of the second bend and discovers that no dimension on the sheet gives it. Machine stopped, call engineering, wait for someone who can open the model. One missing number pulls three people out of their rhythm.
Hiccups like these are rarely a matter of skill. The same engineer who forgot this dimension puts flawless dimensions on nine other sheets. Dimensioning errors on drawings are born from haste, from copy work and from models that change after the dimensions went on. In this article we line up the four errors that hold up the shop floor most often, show how they sneak onto the drawing and which checking habits catch them before production does. Anyone producing drawings from Inventor will recognise the examples; a toolbox like Thundercad plays a modest supporting role further on.
To be clear about scope: this is the deep dive on one topic, dimensioning. The full final check of a drawing, from title block to BOM references, is covered in Checking drawings before release: the complete checklist.
The four errors that stop the shop floor most often
1. The missing dimension
The classic, and it appears precisely on views that look finished: there are twenty dimensions on them, so it seems complete. Yet exactly the position of that one hole is missing, or the dimension is hidden in a detail view that never got made. The machinist cannot continue, or worse: he measures the value from another view himself and guesses the reference. Then the part comes back from assembly instead of from the saw.
2. Duplicate dimensions that contradict each other
An overall dimension of 1480 while the intermediate dimensions add up to 1485. Almost always the result of a change: one dimension chain updated, the other forgotten. Putting the same dimension on the sheet twice feels like a service to the shop floor, but it is a time bomb: as long as the values match, nobody notices, and after the next change they contradict each other and the machinist has to pick which one to believe.
3. Dimensions from the wrong reference
Everything adds up, and still the part comes out wrong. The hole pattern is dimensioned from a sawn edge that may vary by a millimetre, while the holes have to line up with a mating flange to a tenth. The tolerances stack up exactly where it gets tight. The cause is almost always that dimensions were placed from whatever snaps conveniently in the view, not from the face the shop floor measures from and clamps against.
4. Tolerances that make no sense
A fit tolerance on a hole that a bolt passes through with generous clearance, or five hundredths on a sawn tube. Usually copied along from another part or left over from a template. The shop floor then does one of two things: take the tolerance seriously and pour expensive machining time into it, or ignore it. The second option works fine for a long time, until the one occasion the tight tolerance was actually intended.
Where they come from: haste, copy work and changes
If you want to prevent the errors, you need to know where they are born. Three sources together supply nearly the whole catalogue:
- Haste: dimensioning is the tail end of the work, and that is where the deadline sits. Whoever finishes three more sheets at the end of the day does not forget difficult dimensions but easy ones: the position of the first hole, the thickness that "everyone knows anyway".
- Copy work: reusing the drawing of a similar product is fast, but old dimensions, tolerances and notes move along silently. Overridden dimension texts in particular survive every copying round, and nobody sees them sitting there.
- Changed models: the model changes after the drawing was dimensioned, dimensions jump or come loose, and the drawing gets repaired instead of checked. Whoever reattaches a loose dimension without walking the chain is already setting up error number two.
The overridden dimension value deserves a special mention: typing in 12 for now because the model will be updated tomorrow anyway. From that moment the drawing lies about the model, and lies of that kind always come out at the worst possible moment. A hard ban on overridden values is the cheapest quality measure there is: if the dimension is wrong, the model is wrong.
Checking takes time, and that time is the first thing lost to the routine work around it: exporting, publishing, printing. Thundercad takes those chores in Inventor off your hands, so the last look at the dimensions is not the first thing to be cut.
Try 30 days freeChecking habits that catch the errors
Opposite each of the four errors stands a habit that catches it reliably, and none of them costs more than a few minutes per drawing:
| Error | Habit that catches it |
|---|---|
| Missing dimension | walk the drawing in production order: can I saw, drill, bend and weld with this? |
| Contradicting dimensions | every dimension on the sheet only once; extras explicitly marked as reference |
| Wrong reference | dimension from the faces the shop floor measures from and clamps against |
| Nonsense tolerance | for every tight tolerance, ask: what has to fit here, and what does achieving this cost? |
The most powerful habit is the first one. Read your own drawing the way the machinist reads it: in the order of production, not the order of drawing. Sawing first, so which length do I need? Then the holes, so what are they measured from? Whoever asks at every step which dimension he needs right now finds missing dimensions within minutes. How to tune the layout and views of a sheet to the people at the machine is covered in Drawings your shop floor can actually work with.
After every model change: five minutes of fresh eyes
Most dimensioning errors on existing drawings are not born during creation but during changes. So turn the re-check into a fixed reflex: after every model change, walk the affected views, look for loose or jumped dimensions, and follow every dimension chain that touches the changed part. For critical or outsourced parts a second pair of eyes helps: a colleague reading along for ten minutes looks without knowing what is supposed to be there, exactly like the shop floor will.
This check will not become fully automatic any time soon, although software can find part of the signals just fine; a Drawing Checker that walks drawings for exactly these points is on the Thundercad roadmap. Until then, the combination of production order, marker and a second pair of eyes is the most reliable check there is.
Frequently asked questions
How many dimensions belong on a drawing?
Exactly enough to make and to measure the part, and every dimension only once. Too few dimensions force the shop floor to call or guess; too many contradict each other after the next change. If you want to add extra convenience dimensions, mark them explicitly as reference, so everyone knows which dimension leads.
What do I do when the shop floor calls about a dimension?
Answer the question, then do two things: fix the dimension in both model and drawing, not just verbally, and write the question down. Three calls about the same kind of dimension are not coincidence but a pattern, and that pattern tells you exactly which checking habit is still missing in your team.
Does a second pair of eyes really pay off, or is it disguised delay?
For critical and outsourced parts it almost always pays off: ten minutes of reading along is cheaper than one return shipment from the coater or one misdrilled hole in an expensive plate. Work it out for yourself in hours and the trade-off is quickly made. And if you want to free up those ten minutes from the daily export work, you can try Thundercad free for 30 days.