An experienced estimator can pick up a 7 page technical drawing and, within a few minutes, form an impression: this is a moderately complex machined part, probably aluminum, tight tolerances on the bore, standard finish elsewhere, we have done something similar before.
That impression is the result of pattern recognition built over thousands of drawings. It is extraordinarily valuable. And it is almost entirely invisible, even to the estimator themselves.
Ask that same estimator to write down every single piece of information contained in the drawing, and most would struggle. Not because they cannot read it, but because they have never needed to think about it taxonomically. They read drawings the way you read a sentence: absorbing meaning without consciously parsing grammar.
This chapter is about the grammar. I want to map every type of information that exists in a standard 2D engineering drawing, explain what each one means for quoting and manufacturing, and show you how much of it gets lost when the extraction is done by hand.
By the end, you will see your own drawings differently.
How Much Information Is Actually in a Drawing?
Before we get into categories, let me give you a number that surprises most people.
A moderately complex 2D engineering drawing (5 to 7 pages, multiple views, some cross sections, standard GD&T) typically contains 150 to 400 discrete, extractable data points.
That includes individual dimensions, tolerance values, material specifications, surface finish callouts, GD&T frames, notes, title block fields, revision information, and referenced standards.
A complex drawing package (15+ pages, detailed GD&T, multiple assemblies) can contain 800 to 2,000+ data points.
When your estimator reads this drawing manually, they are not extracting all of these data points. They are extracting the ones they consider relevant: the critical dimensions, the key tolerances, the material grade, the obvious cost drivers. Depending on the complexity and the estimator's experience, they capture maybe 40% to 70% of the total information.
The rest gets filtered out by their judgment. Most of the time that filtering is correct. But sometimes the data point that was skipped (a note on page 6, a surface finish callout on a secondary view, a tolerance modifier buried in a GD&T frame) turns out to be a significant cost driver. And nobody finds out until production hits a problem.
The Complete Taxonomy
I have organized the information in a 2D technical drawing into eight categories. For each one, I will explain what it contains, why it matters for quoting, and where the extraction difficulty lies.
Dimensions are the most obvious data in a drawing and the ones estimators extract first. They define the geometry of the part: how long, how wide, how deep, what radius, what angle.
But "dimensions" is not a single category. There are at least six sub types, and each one has different implications for manufacturing cost:
Linear dimensions (length, width, height). The most common. Straightforward to read, straightforward to manufacture. Cost impact is primarily material and machine time.
Angular dimensions. Angles between surfaces or features. These matter for fixturing: a 90 degree bend is standard, but a 47.5 degree bend might require custom tooling or multi axis positioning.
Radial dimensions (radii and diameters). Interior radii on sheet metal determine minimum bend radius, which constrains material thickness and tooling. A 1mm interior radius on 3mm steel might be impossible without cracking.
Depth dimensions. How deep a pocket, hole, or slot goes. Depth to diameter ratio on holes determines whether standard drilling works or requires special tooling. A hole deeper than 10x its diameter is expensive.
Reference dimensions (marked with parentheses or "REF"). These are not manufacturing dimensions. They exist for information only. An estimator who treats a reference dimension as a tolerance dimension will over price the part.
Theoretically exact dimensions (in rectangular frames). These are used with GD&T to define the ideal geometry that tolerance zones are applied to. They are not toleranced themselves, but they are critical for understanding the GD&T scheme.
Tolerances define how much a dimension is allowed to deviate from its nominal value. They are the single most important cost driver in a technical drawing, and the one most commonly misread or overlooked in manual extraction.
General tolerances are declared once, usually in the title block or a note, and apply to all dimensions that do not have specific tolerances called out. The most common standard is ISO 2768, which defines tolerance classes from "very coarse" (v) to "fine" (f) for linear dimensions, and from "H" to "K" for geometric tolerances.
The difference between ISO 2768 mH and ISO 2768 fK is enormous in manufacturing cost. Both are standard tolerance classes, but fK is roughly 2 to 4 times more expensive to achieve because it requires tighter process control, more precise fixturing, and additional inspection.
An estimator who reads "ISO 2768" but misses whether it says mH or fK will price the job incorrectly by a meaningful margin.
Specific tolerances are called out individually on critical dimensions. They override the general tolerance for that dimension. Typical formats include bilateral tolerances (150 ±0.05), unilateral tolerances (150 +0.00 / −0.10), and limit dimensions (149.90 to 150.10).
Why tolerances are the #1 cost driver: A part dimensioned at 150mm with a general tolerance of ±0.5mm can be made on virtually any equipment. The same part at ±0.02mm requires precision grinding or CNC with compensation cycles, climate controlled inspection, and possibly 100% measurement. The cost difference can be 5x to 10x for the same geometry.
GD&T is the most complex and most frequently misunderstood layer of information in a technical drawing. It uses symbolic language to control the form, orientation, location, and runout of features relative to datum references.
There are 14 GD&T characteristics defined in ASME Y14.5 (and equivalent ISO GPS standards). The most common ones in fabrication work:
Flatness (▱). How flat a surface must be. A flatness of 0.05mm on a large weldment surface might require stress relieving and post weld machining. Expensive.
Parallelism (▷). How parallel two surfaces must be relative to each other. Drives fixturing complexity and inspection requirements.
Perpendicularity (▒). Squareness between features. A perpendicularity callout of 0.02mm on a machined surface relative to a datum hole requires very different machining than one at 0.2mm.
Position (⌖). The most common GD&T callout. Controls where a feature (usually a hole) is located relative to datums. Positional tolerance directly affects fixturing strategy and inspection method.
Runout and total runout (↻). Controls how much a surface or feature can wobble relative to an axis. Critical for rotating components. Tight runout drives machining cost significantly.
Why GD&T matters for quoting: A drawing without GD&T relies entirely on linear tolerances to control geometry. A drawing with GD&T adds an entirely separate layer of requirements that can dramatically affect manufacturing method, fixturing, inspection, and cost. An estimator who reads the linear dimensions but glosses over the GD&T frames may underprice the job by 20% to 50%.
The challenge: GD&T is genuinely hard to read quickly. Each frame contains a characteristic symbol, a tolerance value, possible modifiers (MMC, LMC, projected tolerance zone), and datum references. Parsing a complex GD&T frame takes a trained estimator 1 to 3 minutes per callout. On a drawing with 30 GD&T frames, that is an hour of reading just the geometric tolerances.
Surface finish specifications define the roughness, waviness, and lay of manufactured surfaces. They are expressed in Ra (arithmetic average roughness), Rz (average maximum height), or surface class designations.
Common Ra values and their manufacturing implications:
Ra 12.5 μm (500 μin). Standard machined finish. Achievable with basic turning or milling. No special requirements.
Ra 3.2 μm (125 μin). Fine machined finish. Requires sharp tooling, appropriate feeds and speeds. Standard for bearing surfaces and sealing areas.
Ra 1.6 μm (63 μin). Very fine finish. May require grinding or polishing operations. Significantly more expensive than Ra 3.2.
Ra 0.8 μm (32 μin). Ground or honed finish. Requires secondary operations. Cost multiplier of 3x to 5x versus standard machining.
Ra 0.4 μm (16 μin) and below. Lapped or superfinished. Very expensive. Specialized equipment required.
The cost impact is non linear. Going from Ra 3.2 to Ra 1.6 roughly doubles the surface finishing cost. Going from Ra 1.6 to Ra 0.8 roughly triples it. An estimator who reads "surface finish: 1.6" instead of "3.2" on a large surface area has just mispriced a major cost element.
Surface finish callouts can appear as symbols on the drawing (the check mark or fork symbol with Ra value), in the title block as a general specification, or in notes. They sometimes appear on specific surfaces, sometimes as a general callout with exceptions.
Material specification defines what the part is made of. It appears in the title block, in notes, and sometimes directly on the drawing as callouts. For fabrication shops, this is one of the first things an estimator looks for because it determines material cost, machinability, weldability, and forming characteristics.
The challenge is that material specifications come in multiple naming systems that overlap and sometimes conflict:
EN/ISO designations: 1.4301, 1.0038, EN AW 7075 T6. Used primarily in European drawings.
AISI/SAE designations: 304, 316L, 4140, 1018. Used primarily in American drawings.
Trade names and UNS numbers: Inconel 625, Hastelloy C276, UNS N06625.
Customer specific callouts: Some OEMs use their own material specification systems (GM, Boeing, Airbus all have proprietary material standards).
A single drawing might reference material in the title block as "1.4301" and in a note as "AISI 304 equivalent." These are the same material, but an automated system that does not understand equivalencies will treat them as conflicting specifications.
Material callouts also specify condition and temper: T6 vs T4 for aluminum (different heat treatment, different machinability, different cost), HR vs CR for steel (hot rolled vs cold rolled, different surface quality), annealed vs hardened (dramatically different machining parameters).
Knowing just the material grade without the condition is like knowing the ingredient but not the cooking method. The same alloy in different conditions can have 50% different material costs and very different manufacturing characteristics.
The title block is the metadata layer of the drawing. It typically occupies the bottom right corner and contains structured fields. Most estimators scan it quickly for part number and material, then move on. But the title block often contains information that directly affects quoting:
Part number and revision. Essential for tracking. Quoting on the wrong revision is a surprisingly common and expensive error.
Material specification. Usually duplicated here from the main drawing.
Weight. If calculated, this tells you the blank size and material volume immediately.
Scale. A drawing printed at 2:1 scale can mislead estimators who do not notice the scale factor.
General tolerance reference. "ISO 2768 mK" or "ASME Y14.5" declared here applies to the entire drawing.
Surface finish default. The general surface finish that applies everywhere unless specifically overridden.
Projection method. First angle (European) or third angle (American). Reading a third angle projection drawing as if it were first angle results in completely wrong geometry interpretation.
Drawing standard. Which standard the drawing conforms to. This determines how to interpret every symbol and notation.
Approval signatures and dates. Indicates the drawing's maturity and whether it is a preliminary or released document.
This is the most dangerous section of any drawing. I use the word "dangerous" deliberately.
Notes contain requirements that do not fit neatly into the graphical language of the drawing. They are written in plain text, usually in a notes block, but sometimes scattered across pages. And they frequently contain the single most expensive requirements on the entire drawing.
Examples of notes that dramatically affect cost:
"All sharp edges to be broken 0.3 to 0.5mm." On a complex machined part with 40 edges, this is hours of deburring labor.
"Heat treat to HRC 58 to 62 after machining." This adds an entire process step (and possibly distortion correction machining afterward).
"100% dimensional inspection per first article report." The inspection could take longer than the machining.
"No weld repair permitted." Any casting or weld defect means scrap, not rework.
"Surface to be free of tool marks visible to naked eye." A cosmetic requirement that adds polishing labor and increases reject rate.
"Material certificate EN 10204 3.1 required." Certified material costs 10% to 30% more than standard stock.
"Cadmium plating per AMS 2400." Outside processing with hazardous material handling, lead times, and restricted vendor lists.
Notes are where experienced estimators earn their salary. A junior estimator might read the dimensions and tolerances perfectly but miss a note on page 4 that adds $2,000 to the job. This is one of the reasons quoting experience takes years to develop: you learn what notes to look for by getting burned.
Drawings reference external standards that define requirements not explicitly drawn. These are usually listed in a notes block or specification block and include:
Manufacturing standards: AWS D1.1 (welding), AMS 2750 (heat treatment), ASTM specifications.
Quality standards: ISO 9001, AS9100, IATF 16949. These determine documentation requirements, traceability, and inspection procedures.
Testing standards: Magnetic particle inspection (MPI), liquid penetrant inspection (LPI), X ray, ultrasonic testing. Each adds cost and lead time.
Coating and treatment standards: MIL specifications for plating, painting, anodizing.
A drawing that says "all welding per AWS D1.1" has just imposed a comprehensive quality system on the welding operations. A drawing that says "all welding per customer specification XYZ" might impose even more stringent requirements that need to be obtained and reviewed before quoting.
Referenced standards are often the most time consuming element to research during quoting, especially when they reference customer specific or military specifications that the estimator does not have immediately available.
The Information Decay Problem
Now that we have mapped what lives inside a drawing, let me show you what happens when this information is extracted manually.
In a typical manual extraction, the estimator reads the drawing and captures the information they consider relevant. They type it into a spreadsheet, an ERP field, or their own notes. This process is selective by necessity: nobody writes down all 300 data points from a complex drawing. They extract what they need to produce a quote.
The problem is what gets left behind.
Based on observations at fabrication shops where I compared full structured extraction to manual extraction by experienced estimators, here is what typically gets captured versus what gets missed:
| Information Category | Manually Captured | Commonly Missed |
|---|---|---|
| Critical dimensions (overall, key features) | 90% to 95% | Non obvious dimensions, secondary views |
| Secondary dimensions (radii, chamfers, fillets) | 50% to 70% | Detail views, small features |
| General tolerance class | 80% to 90% | Specific class (mH vs mK) sometimes missed |
| Specific tolerances | 70% to 85% | Tolerance modifiers, unilateral specs |
| GD&T callouts | 40% to 65% | Complex frames, datum references, modifiers |
| Surface finish | 60% to 80% | Per surface specs (vs general), Rz vs Ra |
| Material grade | 95%+ | Condition/temper occasionally missed |
| Title block data | 50% to 70% | Scale, projection, default tolerances |
| Notes (critical) | 70% to 85% | Notes on later pages, cross referenced notes |
| Notes (secondary) | 30% to 50% | Packaging, marking, certification requirements |
| Referenced standards | 60% to 75% | Customer specific specs, sub tier requirements |
On average, a good estimator manually captures about 60% to 70% of the total extractable information from a moderately complex drawing. The remaining 30% to 40% is either filtered out by judgment (often correctly) or simply missed.
Here is the critical point: even small misses can be expensive.
Missing one GD&T callout with a tight positional tolerance can mean underpricing a job by 15%. Missing a note about material certification requirements can mean eating $500 in cert costs that were not included in the quote. Missing a surface finish specification on an interior surface can mean absorbing hours of hand polishing.
When estimators miss information, the errors are asymmetric. Missing a requirement almost always means under pricing, because the missed requirement adds cost that was not estimated. Under pricing erodes margins. It turns winning jobs into losing propositions.
Over extraction (capturing something that does not matter) is inefficient but harmless. Under extraction is expensive. This asymmetry means the cost of "good enough" manual extraction is systematically higher than it appears, because the errors it produces are almost always in the wrong direction.
The Structural Problem With Manual Extraction
There is a deeper issue beyond accuracy. Manual extraction is ephemeral.
When your estimator reads a drawing and extracts the information they need, where does that extraction go? Into their notes. Into a spreadsheet. Into their head. Maybe some of it gets typed into an ERP system, but rarely the full extraction.
Six months later, a similar RFQ comes in. The estimator opens the new drawing. Does the full extraction again from scratch. They might remember the previous job, but they have no structured record of what they extracted, what they priced, or what they learned during production.
Every extraction is a one time event. The work is done, the quote is sent, and the information dissolves.
Over 10 years, an estimator processes perhaps 15,000 to 20,000 drawings. The knowledge they accumulate is enormous, but it exists only in their pattern recognition, their instincts, their memory. None of it is searchable. None of it is transferable. None of it survives their retirement.
Structured extraction, where every data point from every drawing is captured and indexed in a searchable system, transforms drawings from one time documents into permanent, searchable institutional knowledge. When the next similar RFQ arrives, the system does not just read the new drawing. It connects it to every drawing the shop has ever processed, surfacing similarities, pricing history, and lessons learned.
That is the difference between reading a drawing and understanding a drawing in context. The first is a task. The second is intelligence.
We explore how this intelligence compounds in Chapter 04.
Stop reading.
Start seeing.
Drop in your hardest drawing - aerospace, sheet metal, precision machining. Watch every dimension, GD&T frame, and tolerance get extracted in minutes.
Upload a Drawing