When did someone else last take a serious look at one of your drawings before it went to the shop floor? In many small engineering teams the honest answer is: a very long time ago, or never. Reviewing sounds like something for large organisations, with stamps, workflows and meetings nobody has time for.
That is a shame, because a second reader is the cheapest quality improvement there is. A drawing review on the four-eyes principle does not have to mean bureaucracy: fifteen minutes of reading by a colleague catches exactly the errors you can no longer see yourself. This article lays out a lightweight review routine that holds up even when there are only two of you, plus an emergency plan for those who draw completely alone. Tooling such as Thundercad plays a supporting role at most: it frees up time, the looking itself remains human work.
Why you can no longer see your own mistakes
Whoever makes a drawing reads it differently from whoever has to use it. You know what should be there, so you see what you meant: your head fills in the missing dimension, you read straight past the wrong sheet thickness because it was still correct yesterday. That is not sloppiness but perception: the longer you work on something, the worse you become at reading it as if for the first time.
Self-checking against a fixed list already removes a lot; how to approach that is covered in the complete checklist for checking drawings before release. But self-checking has a ceiling. The errors that remain are precisely the ones you are blind to: assumptions that appear nowhere on paper, references that only make sense inside your head. That is what a second reader is for, and it does not take them hours.
When four eyes are needed, and when not
Reviewing everything is the fastest recipe for quitting. Make it risk-driven: a short list of criteria decides whether a drawing gets a second reader.
- New type of construction or first order for a new customer: always.
- Outsourced work, such as laser-cut or turned parts made externally: always, because there is no walking distance to fix things quickly.
- Unusual or expensive material, long delivery time, one-offs: always.
- Safety-related parts, lifting points, parts under pressure: always, no exceptions.
- Repeat orders of proven drawings: not in full again, but do have someone read the change itself.
At an average machine builder, criteria like these put perhaps a quarter of the drawing work in the review category; the rest goes ahead after self-checking. That is manageable: a few quarter-hours per week, not a full-time job.
The reviewer's fifteen minutes: reading like the shop floor reads
The reviewer does not redraw anything and does not recalculate everything. They read the drawing the way the shop floor will: from the outside in, with only the paper as information. In fifteen minutes they answer four questions. Can I make this as drawn, think bend sequence in sheet metal, weld accessibility, assembly order? Is everything I need on here: dimensions without arithmetic, sections where it gets tricky, material and quantity in the title block? Does the drawing agree with the rest of the package: BOM rows, references to adjoining drawings, fits with the mating part? And finally: where would I pick up the phone?
That last question is the most important one. Every spot where the reviewer hesitates or has to ask is a phone call from downstairs later on, and every phone call is waiting time. Doubt is not an opinion but a finding.
The easiest way to find fifteen minutes for a review is to cut routine work. Thundercad automates updating, exporting and publishing drawing packages, so the time you free up goes into the look that actually matters.
Try 30 days freeRecording feedback without building a system
A review that only happens verbally evaporates. At the same time, keep the recording as light as possible: a marked-up PDF or a print with a red pen is enough. What matters is that every finding has a place and gets ticked off, not that a form exists.
A workable format: the reviewer numbers their remarks, the author walks through them, fixes each one or explains why not, and every finding gets a tick. The author stays the owner: the reviewer flags, the author decides, and in a genuine disagreement the person carrying responsibility for the release has the final say. And feedback is about the work, never about the person; get that wrong once and you will never receive an honest review again.
Keep the findings per category, even if only as tally marks: dimensioning, material, references, readability. After ten reviews you can see where the team structurally slips up, and that is exactly where failure costs are born. How to follow those costs from the shop floor back to the source is something we covered earlier; the review tallies are the best starting point for it there is.
When there really is nobody
Sometimes there is no second pair of eyes: your colleague is away, or you are simply the only person drawing. Then you simulate distance. The cheapest substitute is time: let the drawing rest overnight and read it first thing the next morning, before opening your mail and your model. Also change the medium: print on paper or view the PDF on a different screen at a different scale; errors that were invisible on your own screen suddenly become readable. A third technique is explaining out loud: describe the drawing as if you were talking the shop floor through it by phone, and notice where you stumble.
And do not underestimate the shop floor itself. Five minutes of reading by the person who has to build it is not a vote of no confidence but the best review there is, especially for bending work and tight assembly. Organise that structurally and your second reader simply comes from downstairs.
Frequently asked questions
How much time does a review take per drawing?
Count on fifteen minutes for an assembly drawing with a parts list and a few minutes for a simple part. Assuming a quarter of the drawing work falls into the review category, that stays well under an hour per person per week at team level. One caught bending error already pays that time back.
Does the reviewer have to be senior?
No, a fresh pair of eyes often works better. The reviewer reads like the shop floor reads: what is not on the paper, they cannot fill in, and that is exactly what you want surfaced. Trade knowledge helps with manufacturability, but completeness and readability can be checked by anyone who can read a drawing.
How do I keep the review from being the first thing dropped?
Make it small and visible: only high-risk drawings, a fixed moment per release, findings on a list everyone can see. And make room: automating routine work, for instance with the free trial month of Thundercad, leaves you the fifteen minutes the review needs.
