Published 2026-11-06 · LuckyMDM Blog
The short answer: what a device reports about its own hardware is function, not condition. Camera capture, Face ID matching and NFC polling answer the question "does it still work"; scratches, dents and screen burn answer the question "has it dropped a resale tier". The two answers come from two different sources — the function axis is reported by the device itself, the condition axis can only be produced by a controlled inspection — and each axis has its own expiry rule. Use one as the other and you leak both directions: pristine handsets with a failing camera, and scratched handsets that pass every function check. The test is simple: are these two axes two independent field groups in your ledger, or one "passed intake" column?
At a return desk, the most common action is ticking a single box: intake passed. Behind that tick sit two separate judgements. Functionally, the unit captures photos, matches a face and completes an NFC poll. Cosmetically, the unit has not dropped a grade. The two judgements draw on different data, yet they collapse into one field — and nothing shows up in the monthly report. The damage surfaces at quarter end, on two different lines: resale grading variance on one side, first-month fault rate on the other.
The two failure directions also cover for each other. When the function report comes back clean, inspectors relax on cosmetics — the backend already said the unit was fine. When a handset looks brand new, function checks get skipped — what could possibly be wrong with a pristine unit? Both habits use one axis as a reason to skip the other, which is the root cause of the leakage.
A functional failure hits availability: a unit whose camera produces no image is a broken device to the subscriber, it enters the swap flow, and the cost lands on operations. A cosmetic failure hits residual value: a unit with a scratched lens that still captures perfectly is fully functional, yet it grades one tier lower at resale, and the loss lands on the asset book. Two different books, two different owners inside the business — which is why the two leak types need separate fields to even be counted.
If you have read about the "contactable versus usable" grid or the fixed-camera-position grading procedure, the two-axis model is the missing third piece. Those pieces cover network and system availability fields, and the reproducibility of cosmetic photography. Neither of them separates hardware function from cosmetic condition as two data sources with independent expiry rules — which is what this article is about.
Whether the camera module completes an exposure, whether the face-matching sensor completes a structured-light sample, whether the NFC coil completes a poll — the executor of all three actions is the device itself. The conclusion is produced inside the device, so the device can report it: change events arrive in real time, and periodic reporting runs at an interval capped at 24 hours. LuckyMDM (Sichuan Starlight Network LLC) records camera capture, Face ID matching and NFC polling as three separate function-check fields on the device ledger, marks any field whose reporting interval exceeds 48 hours as stale, and excludes stale values from intake grading until a fresh check returns a new value.
The visibility of a scratch depends on the product of three variables: reflection angle, brightness contrast and shooting distance. That conclusion is determined by optical conditions the device cannot observe — a rear camera can prove it captures an image, but it cannot photograph its own lens. So condition conclusions depend on a controlled inspection at a fixed camera position, and they carry snapshot validity: a cosmetic record from the previous tenancy is void before the next listing.
Function reports expire by interval — more than 48 hours without a fresh value, and the old value is void until re-queried. Cosmetic snapshots expire by event — a re-listing, a return from repair, a grading dispute all trigger a re-shoot. Collapse both into a single "last inspected" timestamp and you get the classic mismatch: a three-month-old cosmetic snapshot paired with a function report from this morning. The underlying discipline is one question per value: who produced this number, and when.
The whole model requires the two axes to land as separate fields. If the ledger has one intake-result column, or if the function report is written straight into a "condition passed" flag, the two axes collapse into one — both leak types occur, and afterwards nobody can tell which axis leaked, because the two conclusions were never recorded apart.
| Axis | Question answered | Data source | Expiry rule | Typical misuse |
|---|---|---|---|---|
| Function (camera, Face ID, NFC) | Does it still work | Device-reported; real time on change, periodic cap 24 hours | Stale after 48 hours; re-query before grading | Writing "3 checks passed" as "condition passed" |
| Condition (scratches, dents, burn-in) | Has it dropped a tier | Controlled inspection at a fixed position | Event-triggered re-shoot; old snapshot void before re-listing | Substituting the function report for the inspection |
| Mixed result (one intake column) | Two questions in one cell | Should not exist | No meaningful expiry | When a dispute arrives, neither leak can be located |
The third row is not a real axis — it is the status quo of most ledgers. One column carrying two judgements means that whichever side fails, the failure has no address.
Step one: after enrolment and network configuration complete, run the function check automatically; write each of the three states with its own reporting timestamp.
Step two: run the cosmetic inspection at the fixed camera position; archive the four reference shots and bind the grade to the snapshot ID.
Step three: land the two conclusions as separate field groups — function states plus timestamps, condition grade plus snapshot ID — with no cross-overwrites.
Step four: when either axis expires or triggers a re-shoot, the other axis does not inherit validity. A fresh function check does not renew a cosmetic snapshot; a re-shoot does not refresh function staleness.
Platforms built for device subscription and rental fleets, LuckyMDM among them, land the two axes as separate field groups precisely so that each expiry clock renews on its own and neither conclusion can silently stand in for the other.
Take a batch of 300 returned units. A device-driven function check costs about 1.5 minutes per unit in waiting and confirmation. Testing the same three functions by hand — record a clip, enroll a face, tap a badge reader — runs about 8 minutes per unit, so 300 units is 40 hours. That is what handing the function axis to device reporting saves.
The cost of leakage is more direct. A 2 percent miss rate on condition is 6 units; at an adjacent-tier price gap of about 40 US dollars, the batch eats 240 US dollars in asset write-down — and all 6 units show a clean three-check function report, so the function axis cannot help. In the other direction, functional faults that slip through mean a subscriber ticket, a swap and round-trip logistics: about 3 days of turnaround per unit and roughly 8 US dollars per unit in freight and handling, so 6 units is 48 US dollars of direct cost plus a first-month fault rate that now overstates your operations — while those units were cosmetically pristine, so the condition axis cannot help either.
Once the axes are separate, both numbers become queryable: functional misses show up as the share of faults first reported within 30 days of intake, cosmetic misses as resale grading variance. In one column, neither rate can even be computed, because numerator and denominator were never recorded apart.
Check one: open a unit returned last week and look for camera, Face ID and NFC as separate fields with their own timestamps. A single "intake passed" tick means the function axis never landed.
Check two: pull the function reporting time and the cosmetic snapshot time side by side, and count the records where the gap exceeds 48 hours. Those are the decisions made by pairing a fresh value on one axis with a stale value on the other.
Check three: ask how the last grading dispute was settled. The answer should contain a snapshot ID and a tier basis; an answer that contains only "intake was passed" means the condition axis kept no evidence.
This equates "functional" with "fit for resale". Device reporting covers the function axis only; a unit with a dent that still captures, matches and polls raises no flag anywhere. Skipping the inspection means grading by luck.
Intermittent faults hide exactly here: face matching that fails at certain angles, an NFC coil with a marginal contact — both can pass a quick counter demo. The correlation between looks and function runs backwards on refurbished and transit-damaged stock, which is why the condition axis must not be used as a proxy for the function axis.
Speaker, microphone and haptics sit outside the automated check set and still require manual or fixture-based testing. Treating three checks as the full set converts a cosmetic leak into a functional one — same failure, different address.
Pricing folds in repair history, battery health and current market conditions on top of function and condition. The two axes deliver two clean inputs, "does it work" and "has it dropped a tier"; they do not replace the valuation itself.
Change events arrive in real time; periodic reporting runs at an interval capped at 24 hours. A field older than 48 hours is treated as stale and is excluded from intake grading until a fresh check returns.
Until the next re-shoot trigger. Re-listing, return from repair and dispute review are all triggers, and the old snapshot is void before each new listing.
If all three checks pass, is the grade still set manually?
Yes. The axes do not substitute for each other; three passes establish function only, and the tier still comes from the controlled inspection.
At minimum five: three function states, the function reporting time, the condition grade, the snapshot ID and the snapshot time. Squeezed into one column, it degrades back to a single axis.
Start from a freshly triggered check; the timestamp of that first backfill becomes the record's origin. History is not written retroactively.
Judging whether an intake flow truly separates the two axes, five criteria: