Screen wear classification is where most grading rubrics break down in practice. The problem is not that operators lack categories. Grade A, Grade B, Grade C, and various sub-tiers like Grade A-minus or Grade B-plus exist in most systems. The problem is that the boundaries between those categories are defined in language that describes visible outcomes without specifying the viewing conditions under which visibility is assessed.
A hairline scratch that is invisible under overhead fluorescents reveals itself clearly under raking light at a 20-degree angle. Whether that scratch puts a device in Grade A or Grade B depends entirely on what kind of light your counter staff uses during inspection, and that decision was made implicitly the first day someone set up the counter, not in any written policy.
The Taxonomy of Screen Wear
For a grading system to produce consistent results, it needs to classify screen wear at a more granular level than "visible" or "not visible." The dimensions that matter in practice are depth of defect, location on the screen face, density (how many defects in a given zone), and how the defect pattern interacts with display content at normal brightness.
Hairline scratches in the oleophobic coating are the most common and the most contested category. These scratches are shallow enough that they do not affect display function, but they create a visible glare pattern under certain light angles. Most buyers can see them if they look. Many buyers will not look. The question for grading is not "can these scratches be detected" but "do they affect the buyer's perception of condition in the transaction context."
Deeper scratches that reach the glass surface are less ambiguous. They are visible under normal lighting conditions and will affect the buyer's experience of using the device. These reliably push a device into Grade B or lower.
Pressure marks and micro-abrasion patterns from heavy glass use are a third category that does not fit cleanly into the hairline-versus-deep binary. They appear as a haze in the screen center under indirect light, caused by fine scratching across the coating surface. This type of wear is common on heavily used devices, correlates with high cycle counts, and is often missed by inspection under direct overhead light because the haze only becomes apparent when viewing at an angle.
Bezel and edge wear is a fourth category. Chips, scuffs, and paint loss on the frame affect the overall visual impression of the device but do not change how the screen grades. Many rubrics lump screen and frame condition into a single cosmetic grade. This creates a problem when a device has a pristine screen and significant frame wear, or vice versa. The buyer's reaction to those two situations is different, and the pricing should reflect that.
Where Buyer Acceptance Actually Sits
We spent time analyzing handset return patterns in a set of refurbished device transactions to understand where buyers actually draw their acceptance line, as opposed to where sellers assume the line sits.
The clearest finding: buyers who inspect before purchase accept hairline scratches at a much higher rate than sellers expect. A device with light hairline scratching across the oleophobic coating, visible only at a specific angle, is accepted as Grade A by the majority of buyers who see it in person, particularly if the price reflects a slight adjustment. The Grade A-minus designation that many operators use for this category creates pricing friction without corresponding rejection reduction.
The place where buyers consistently refuse devices is at center-screen deep scratches longer than roughly 5 to 7mm, and at coating damage that affects viewing quality at normal use angles. These are the defects that Grade B classification is supposed to capture. When Grade B is applied consistently to this category and Grade A is used for the hairline-only category, buyer acceptance rates align with seller grade claims at a much higher rate than in operations where the boundary is blurry.
The failure mode is this: when sellers inflate Grade A to include what should be Grade A-minus or Grade B, buyers begin discounting the Grade A label entirely and building in their own condition adjustment at the time of inspection. The grade stops being informative, and every transaction becomes a negotiation from scratch rather than a confirmation of a pre-agreed condition standard.
How Image-Based Classification Handles Screen Wear
Image-based grading handles the classification problem differently than a text rubric applied by a human inspector. A well-shot screen photo under consistent lighting conditions captures the visible surface state at the time of capture. The classification system then scores the image against a labeled dataset.
What this means practically: the classification output reflects what is visible in the photo. A scratch that is visible in the photo at the capture angle will be classified. A scratch that is not visible in the photo will not be classified, even if it would have appeared under a different inspection angle.
This is not a weakness specific to image-based classification. It is a version of the same observation-condition dependency that affects human inspection, just made explicit. The difference is that the observation condition for image-based classification is defined by the photo capture protocol, which can be specified and enforced. The observation condition for human inspection is defined by the counter lighting setup, the staff member's inspection habit, and the ambient light that day, none of which are controlled.
For screen wear classification, the implication is that the photo capture protocol matters as much as the classification algorithm. A consistent intake lighting setup and a defined capture angle produce photos where the classification output has a known relationship to actual screen condition. A variable lighting setup produces photos where the classification output is inconsistent, not because the algorithm changes but because the input changes.
The Buyer Acceptance Alignment Problem
The deepest issue with screen wear classification is not the rubric or the measurement method. It is the alignment between what the seller classifies, what the buyer sees, and what the buyer's acceptance threshold actually is at the price point being offered.
A buyer who purchases a Grade A device at a Grade A price expects near-new screen condition. What "near-new" means to that buyer is shaped by their own experience with new devices and by whatever they saw in the listing. If the listing photo was taken under flat overhead light and the device actually has hairline scratching visible at any other angle, the buyer's experience of the device in hand will not match their expectation from the listing. That gap produces returns and renegotiations.
The practical fix is photo transparency. When the intake photo that produced the grade is the same photo the buyer evaluates before purchase, the buyer's expectation is calibrated to the same observation condition that produced the grade. The gap closes. Buyers who purchase after seeing the annotated intake photo have significantly lower return rates for condition-related complaints, because the photo pre-disclosed what the buyer was accepting.
We are not arguing that image-based grading solves all buyer-seller condition disputes. Some buyers have higher standards than any reasonable Grade A definition accommodates, and no grading system prevents all post-purchase complaints. The argument is narrower: consistent classification, combined with photo-based disclosure at the time of sale, moves the buyer acceptance threshold from a judgment call at pickup to a documented agreement at purchase. That shift changes the margin profile of the transaction at the point of sale and the return rate after it.