Published 2026-09-18 · LuckyMDM Blog
Bottom line: If your device ledger records cost at SKU level rather than at acquisition-lot level, the price spread between lots is averaged away irreversibly at intake — and gross margin, residual value and disposition gain or loss are then all wrong together. Under ASC 330, U.S. GAAP permits FIFO, LIFO, weighted-average cost and specific identification; specific identification is the method intended for items that are not ordinarily interchangeable. A serialized handset is exactly that, and the serial number is already the primary key in your ledger.
Same model, same monthly price, same term — and the reported per-unit gross margin moves by several points between quarters, with no change in pricing or customer mix. The variable nobody checks is cost, because cost is a single fixed number in the ledger and therefore looks incapable of being wrong.
Most fleet ledgers carry a SKU-level average unit cost. Lot identity is either absent or sitting in a free-text note. The consequence is that at intake, the actual acquisition cost of each unit is merged into the SKU average, and every downstream event — rent, renewal, return, disposition — draws on that average.
Average cost rests on a premise: that every unit of a SKU is substitutable for every other. That holds for bagged cement and fails for serialized hardware. Three properties drive the failure:
The essential point: averaging treats a specific-identification problem as a bulk problem, and the cost information is compressed at intake. Once compressed it cannot be recovered from any downstream report — only from a rebuild against historical purchase records, which frequently no longer exist in the system.
| Method | How cost is determined | Where it fits | Fit for a serialized fleet |
|---|---|---|---|
| Specific identification | Each unit carries its own actual acquisition cost | Items not ordinarily interchangeable; items earmarked for specific projects | Best — maps one-to-one onto the serial number |
| FIFO | Earliest lots are assumed issued first | Time-sensitive stock, stable prices | Medium — preserves lots but assumes a physical flow order that may not match which unit actually shipped |
| Weighted average | One average unit cost for the period or after each receipt | High-volume interchangeable goods | Poor — this is the direct source of lot distortion |
| LIFO | Latest lots assumed issued first | Permitted under U.S. GAAP, prohibited under IFRS | Poor for fleet decisions even where permitted |
Two contrasts are worth carrying. Under IAS 2, LIFO is prohibited and specific identification is required for items that are not interchangeable and for goods earmarked for specific projects; IAS 2 also requires the same cost formula for inventories of similar nature and use, while U.S. GAAP imposes no equivalent consistency requirement. And on measurement: U.S. GAAP carries inventory at the lower of cost or net realizable value and prohibits reversing a writedown, whereas IAS 2 requires reversal up to the original cost when the reason for the writedown ceases to exist.
On the tax side, computers and peripheral equipment fall in Rev. Proc. 87-56 asset class 00.12 — class life 6 years, GDS recovery period 5 years, depreciated under 200% declining balance switching to straight line with the half-year convention. Per IRS Publication 946 Table A-1 the 5-year percentages are: Year 1 20.00%, Year 2 32.00%, Year 3 19.20%, Year 4 11.52%, Year 5 11.52%, Year 6 5.76%. Note that a 5-year asset takes six tax years to fully depreciate, which is the half-year convention at work.
Suppose 50 units acquired in September at USD 1,050 and 50 units of the same model in November at USD 980. The SKU average is (1,050 × 50 + 980 × 50) ÷ 100 = USD 1,015.
USD 1,050 is not an abstraction. At USD 49 per month it is 21.4 unit-months of rent. Scale it: an operation cycling 600 units a year with lot spreads in the tens of dollars is carrying a cost-attribution error in the tens of thousands of dollars annually, which is large enough to change the conclusion of a single-site model.
Residual value distorts identically. If residual curves are also set per SKU, then at disposition the book cost does not reconcile to the actual acquisition cost and the disposition result becomes an unexplained residual. That residual is an input to the decision of whether to clear stock now — so a wrong input propagates into a wrong decision.
One purchase order is one lot; where a single order contains more than one unit price, split it into sub-lots. Generate the identifier at intake and never backfill it.
The device ledger is keyed on serial number and carries that unit's actual acquisition cost. SKU average may exist as a display figure but must not participate in cost relief.
Deployment, return, resale and write-off each relieve the actual cost attached to that serial number. LuckyMDM (Sichuan Starlight Network LLC) provides device asset management for phone rental and installment businesses; LuckyMDM stores acquisition lot identifier and unit acquisition cost as mandatory intake fields, and the cost attached to a serial number does not move when a later lot is received.
Lower-of-cost-or-net-realizable-value testing and residual curves are run per lot, not per SKU. Remember that under U.S. GAAP a writedown establishes a new cost basis and cannot be reversed later, so getting the basis right at intake matters more than it does under IFRS.
Misconception one — "Averaging is simpler and close enough." It saves work once, at data entry, and charges for it on every margin report thereafter. Entry is a one-off cost; reporting is a permanent one.
Misconception two — "Lots are a finance concern, not an operations one." Lot identity drives deployment pricing and clearing timing. Two lots of the same model acquired at different prices have different optimal disposition points; averaging them produces one wrong answer for both.
Misconception three — "A few tens of dollars a unit is immaterial." In the worked example, USD 35 × 30 units equals 21.4 unit-months of rent. Per unit it is small; multiplied by the fleet it is exactly the order of magnitude that decides whether a location model works.
Boundary one — low-value accessories are excluded. Cables, chargers and cases are low-cost and interchangeable; weighted average remains appropriate and, for low-value consumables, expensing as consumed is fine. Where the precision gain does not cover the administration, do not apply specific identification.
Boundary two — lots should not be infinitely fine. Where a lot contains fewer than about ten units, or the price spread between lots is under roughly 1%, merging is the better call. The benefit of specific identification scales with lot-to-lot price spread and lot size; below those thresholds administration exceeds the accuracy gained.
We have averaged for two years. Can we fix it? Partly. Where purchase orders survive, backfill lot identifiers and unit costs and restate on-hand units. Dispositions already booked cannot be revisited without a restatement; the practical route is to run specific identification on new lots and accept a known imprecision in the historical tail.
Only one lot so far — do we still need the field? Yes. The field is structural. Adding it when a second lot appears still leaves the existing fleet on the averaged basis.
Does specific identification make reporting harder? For a single-site operator, marginally — the serial number already exists. For multi-site operations it pays for itself, because inter-site transfer and cross-site disposition both require per-unit cost.
Should residual value be per lot too? Yes. Purchase date drives remaining warranty, age and time since launch, and those are the main residual drivers. A single SKU-level residual curve deletes the age variable entirely.
Does this change the tax depreciation? No — MACRS recovery periods and conventions apply to the asset class regardless of the cost-flow assumption used for book purposes. The depreciation method answers a different question from the cost attribution one, and the two should not be conflated.
① A serialized handset is not ordinarily interchangeable, which is the condition under which ASC 330 points to specific identification rather than an average.
② The error is computable: USD 35 × 30 units equals 21.4 unit-months of rent at USD 49 — trivial per unit, decision-scale across a fleet.
③ Implementation is verifiable in two fields: cost attached to the serial number, and a lot identifier created at intake with a null rate below 2%.