KEY TAKEAWAYS

✓
Flatness controls a single surface in isolation — it has no datum reference and measures how much a surface deviates from a perfect plane.

✓
Parallelism controls a surface or axis relative to a datum — it always requires a datum reference and constrains orientation, not just form.

✓
A surface can be perfectly parallel to its datum and still fail a flatness callout — these are independent controls.

✓
Tighter tolerances cost more to machine and inspect — only apply flatness or parallelism where function genuinely demands it.

✓
CMM inspection is the reliable method for verifying both controls; surface plate and indicator checks are acceptable for less critical applications.

In GD&T, every geometric control falls into one of five categories: form, orientation, location, runout, or profile. Flatness is a form control. It governs the shape of a single surface independent of anything else on the part. No datum is referenced, no other feature is involved. You are simply asking: does this surface stay within a tolerance zone defined by two parallel planes? Parallelism is an orientation control. It governs the relationship between a surface (or axis) and a specified datum plane or axis. It always requires a datum reference in the feature control frame. You are asking: is this surface oriented within a tolerance zone that is parallel to the datum? The distinction matters enormously in practice. Flatness can be inspected on a surface plate with a dial indicator without referencing any other feature. Parallelism cannot — you must establish the datum first, then measure the deviation of the controlled surface relative to it. Conflating the two leads to ambiguous drawings, incorrect inspection setups, and parts that pass one check while failing the other.

Rule of Thumb: If your feature control frame has no datum letter, it is a form control (flatness, circularity, cylindricity, straightness). If it references a datum, it is orientation, location, or runout.

The flatness symbol is a parallelogram — two horizontal lines — placed in the first compartment of the feature control frame. The tolerance value follows it. No datum is referenced in the remaining compartments. The tolerance zone for flatness is the space between two parallel planes separated by the tolerance value. Every point on the controlled surface must fall within this zone. A flatness callout of 0.05 mm means the entire surface must fit between two perfectly parallel planes that are 0.05 mm apart. The surface can be tilted, curved, or positioned anywhere in space — the only requirement is that all surface points fit within that 0.05 mm sandwich. Flatness is commonly applied to mating surfaces on flanges, gasket faces, valve bodies, and precision bases where surface contact quality drives function. It is also implied by Rule No. 1 (the Envelope Principle) in ASME Y14.5 — a feature of size at Maximum Material Condition (MMC) must have perfect form. When you invoke a flatness callout explicitly, you are overriding or supplementing that default behavior. Specifying flatness tighter than your surface finish or manufacturing process capability is a common and costly mistake — your machinist cannot hold 0.005 inch flatness on a part that is stress-relieved after machining without additional lapping or grinding.

Warning: Flatness tolerance must always be less than the size tolerance of the feature. If your thickness tolerance is 0.10 mm, your flatness callout must be tighter than 0.10 mm — otherwise the flatness control adds no constraint.

The parallelism symbol is two diagonal parallel lines in the feature control frame, followed by the tolerance value, followed by the datum reference. Unlike flatness, parallelism cannot exist without a datum. The tolerance zone is two planes — or a cylindrical zone for an axis — separated by the tolerance value, both oriented perfectly parallel to the datum. Every point on the controlled surface (or every point along the controlled axis) must fall within this zone. A parallelism callout of 0.05 mm to datum A means the controlled surface must fall within a 0.05 mm wide zone that is perfectly parallel to datum A. The zone itself can be anywhere in space relative to datum A — its position is not constrained by this callout alone. If you need to control both orientation and position, you would add a perpendicularity or position callout, or use a profile of a surface control. Parallelism is critical wherever two mating surfaces must maintain consistent contact or clearance over their full area — think sliding dovetail ways, bearing housing bores, cylinder head mating surfaces, and precision fixture plates. When applied to an axis (such as a bored hole), parallelism controls whether the hole axis runs parallel to a datum axis or plane, which directly affects whether pins, fasteners, and shafts assemble without binding.

Key Insight: Parallelism controls orientation only — it does not control the distance between the controlled surface and the datum. For that, you need a size tolerance or a profile callout. Many engineers incorrectly assume a parallelism callout also locks the gap.

Consider a precision aluminum spacer — a simple rectangular block with a top and bottom face. You specify parallelism of 0.05 mm on the top face relative to datum A (the bottom face). Does that also control flatness? No — and this is where engineers get into trouble. The top face could be warped like a potato chip and still satisfy the parallelism callout, as long as every point on that warped surface falls within the 0.05 mm zone parallel to datum A. If your application requires a flat mating surface — say, a gasket seat — you need both callouts. Add a flatness callout on the top face tighter than the parallelism tolerance. Now you are controlling both the form of the surface and its orientation to the datum. A useful rule: parallelism tolerance should always be equal to or greater than the flatness tolerance on the same surface. If flatness is 0.02 mm and parallelism is 0.05 mm, the surface must first satisfy the 0.02 mm form requirement (which is inherently the tighter constraint) and then that form must fit within the 0.05 mm orientation zone. Reversing this — specifying flatness looser than parallelism — is geometrically inconsistent. The flatness zone would permit deviations larger than the parallelism zone allows, creating an unverifiable condition. Most inspection software will flag this, but it still wastes review cycles.

Flatness and parallelism tolerances are not free. Every tighter tier of tolerance translates directly into process selection, cycle time, and inspection rigor. Standard CNC milling can typically hold flatness in the range of 0.025 mm to 0.075 mm depending on part size, material, fixturing, and thermal stability during machining. If your design requires 0.005 mm flatness, you are almost certainly looking at surface grinding, lapping, or hand scraping — operations that add significant time and cost. Parallelism is similarly process-driven. For a machined prismatic part, the parallelism you achieve is a direct function of your fixturing accuracy, spindle runout, and thermal compensation. On large parts, thermal growth during machining is often the dominant error source. For tight parallelism on large aluminum or steel parts, single-setup machining (where both datum and controlled surface are machined in one clamping) dramatically reduces the opportunity for error. Material matters too. Castings and weldments with residual stress will move after machining. If you need 0.01 mm flatness on a cast iron surface, you need stress relief before finish machining — and possibly again after rough machining. Nimble’s certified partner network performs free DFM reviews on every order, and this is exactly the kind of issue that gets flagged before a single chip is cut. Catching a process mismatch at the quote stage is far cheaper than a first-article rejection.

Cost Tier Guide: Flatness 0.1 mm+ — standard milling, no premium. 0.025–0.1 mm — controlled milling with quality fixturing. 0.005–0.025 mm — grinding or precision milling required. Under 0.005 mm — lapping, scraping, or specialty processes only.

Verifying flatness and parallelism correctly requires understanding the measurement setup for each. Flatness inspection involves placing the part on a surface plate (or using a CMM) and measuring a grid of points across the surface. The software or analyst calculates the minimum separation of two parallel planes that contain all measured points. The result is compared to the tolerance value. With a dial indicator on a surface plate, you scan the surface and record the total indicator runout (TIR) — the full range of the indicator reading. For flatness, that TIR value is a valid approximation of the flatness deviation, provided the datum surface is itself flat. Parallelism inspection requires establishing datum A first — either by resting the datum feature on a surface plate or registering it in a CMM fixture — and then measuring the controlled surface. The CMM computes the deviation of the surface points from a perfect plane parallel to the established datum. The TIR value of the indicator sweep, with the datum locked, gives you the parallelism deviation. CMM inspection is the authoritative method for both controls, particularly for tight tolerances or complex geometries. Surface plate methods are acceptable for tolerances above approximately 0.025 mm on parts where manual setup does not introduce significant error. All parts manufactured through Nimble’s network include CMM inspection reporting, which documents actual measured values against the callout — not just pass/fail.

Inspection Warning: A common error is measuring flatness by rocking the part on a surface plate and noting the gap. This measures convexity at best — it does not capture concavity or saddle-shaped deviations. Always use a grid of points across the full surface.

Several recurring mistakes show up in drawings submitted for manufacturing. Mistake 1: Applying parallelism without a datum. This is a drawing error — parallelism has no meaning without a reference. The drawing will be returned or the shop will guess at intent. Always include the datum reference. Mistake 2: Over-tolerancing for the application. Specifying 0.005 inch flatness on a non-critical cover plate wastes money and can cause unnecessary rejections. Ask yourself: what is the functional consequence of exceeding this tolerance? If the answer is ‘nothing significant,’ loosen it. Mistake 3: Applying flatness to a non-flat nominal surface. Flatness applies only to nominally flat surfaces. If your surface has a designed crown or contour, use Profile of a Surface instead. Mistake 4: Forgetting that parallelism includes flatness implicitly — but only when tighter. As discussed, a parallelism callout does not control form unless the tolerance is tighter than what the size tolerance already permits. If you need form controlled independently, call it out separately. Mistake 5: Not accounting for free-state variation on thin or flexible parts. Sheet metal and thin-walled parts will deflect under their own weight. GD&T provides a free-state modifier for these situations. If you inspect a thin flange clamped to a surface plate, you may pass it in a constrained state that it cannot hold in service. Define your inspection conditions on the drawing.

The choice between flatness, parallelism, or both comes down to the functional requirements of the feature. Use flatness alone when you need a surface to mate consistently, seal reliably, or support a precision component — and that surface’s orientation relative to other features is already controlled by your assembly or by a profile callout. Sealing surfaces, lapping targets, and precision reference surfaces often fall here. Use parallelism alone when a surface’s orientation to a datum matters for assembly — such as a lid that must close flush, or a bearing bore that must align with a shaft — but the local form of the surface is adequately controlled by the size tolerance. Use both controls together when you need a surface that is both locally flat and correctly oriented to a datum. A cylinder head fire deck is a classic example: it must be flat to seal the head gasket, and it must be parallel to the crankshaft bore axis to maintain consistent compression. When in doubt, run a tolerance stack analysis before locking in callouts. Over-constrained drawings with multiple redundant tight tolerances are harder and more expensive to make and inspect — and they can make first-article approval unnecessarily difficult. Working with a sourcing partner that performs DFM review before manufacturing begins helps catch these issues early. Nimble’s 24-hour quote process includes that review at no charge, so tolerance decisions can be revisited before any manufacturing cost is committed.

Quick Decision Rule: No datum needed? Use flatness. Relationship to another feature matters? Use parallelism. Both the surface shape and its relationship to a datum matter? Use both — with flatness tolerance tighter than or equal to parallelism tolerance.

READY TO SOURCE?

Table of Contents

Related Services

Get a Quote in 24h

Upload your drawing and get a detailed quote with free DFM review.