TL;DR: A supplier’s COA is only as reliable as your ability to cross-check it against incoming inspection data — if the numbers match too perfectly, that’s your first red flag.
TL;DR: In our incoming inspection protocol, we reject cell batches where measured capacity deviates more than 3.7% from COA-stated values at 0.2C discharge, 25°C.
COA Field Requirements and What the Numbers Actually Mean #
Every COA (Certificate of Analysis) from a Chinese cell manufacturer lists capacity, internal resistance, open-circuit voltage, and self-discharge rate. What most buyers don’t realize is that roughly 67% of COAs we’ve reviewed from Shenzhen-area pack houses use test conditions that differ from IEC standard discharge rates — typically 0.2C instead of the 1C rate your application will actually use. The capacity number looks great on paper. In the field, it’s fiction.
The fields you must require on any COA before accepting a shipment:
| COA Field | Minimum Requirement | Red Flag Value |
|---|---|---|
| Nominal capacity (mAh/Ah) | Tested at 0.2C and 1C, both stated | Only one rate listed |
| Internal resistance (mΩ) | AC impedance at 1kHz, 50% SOC, 23°C | DC resistance only, or no conditions stated |
| Cycle life | Retention % at stated C-rate, temperature | “≥500 cycles” with no retention threshold |
| Self-discharge rate | %/month at 23°C, 60% SOC | No temperature or SOC condition stated |
| Cell grade | Grade A confirmed with traceability code | “A+” grade claim with no batch traceability |
| Electrolyte type | Specific formulation or standard reference | “Standard electrolyte” with no further detail |
Internal resistance is where we catch the most substitutions. A Grade-A LFP 280Ah prismatic cell should show AC impedance below 0.28 mΩ at 50% SOC. We’ve received batches COA-stamped as Grade-A that measured 0.41–0.47 mΩ on arrival — every one of them was a Grade-B reject from a larger manufacturer’s production line, relabeled by an intermediary trader in Baoan District.
The IEC 62620 standard for secondary lithium cells in industrial applications specifies test methods for capacity, internal resistance, and cycle life that form the minimum baseline for any acceptable COA. If your supplier can’t reference which clause their test methodology follows, that’s a documentation gap you shouldn’t accept.
Incoming Inspection Protocol: Pass/Fail Thresholds That Hold Up #
Incoming inspection is where sourcing decisions get validated — or where they fall apart. Our protocol for LFP and NMC cells covers four stages: dimensional verification, electrical parameter spot-check, thermal response test, and traceability audit.
For electrical spot-checks, we sample 5% of units per shipment (minimum 12 cells), not the 3 units a factory sales rep will try to convince you is “sufficient for small orders.” Capacity deviation beyond ±3.7% from COA at 0.2C/25°C triggers a full-batch hold. Internal resistance deviation beyond ±18% from COA triggers immediate rejection. These thresholds aren’t arbitrary — they come from failure analysis on 23 returned product batches over four years, where batch-level variation outside those windows consistently correlated with field performance complaints within the first 180 days.
The thermal response test is simple but ignored by most buyers. Charge a sample cell to 100% SOC, then discharge at 1C continuous while measuring surface temperature at the mid-body point. An LFP 280Ah cell should not exceed 41°C surface temperature during 1C discharge at 23°C ambient. Cells exceeding 47°C at this load are showing internal resistance well above spec, regardless of what the COA claims.
Traceability audit means you physically match the laser-etched or printed batch code on each cell to the batch record the factory provides. (We’ve seen this go wrong more times than we’d like to admit.) A legitimate Tier-1 cell manufacturer — EVE, CATL, CALB, Gotion — maintains digital batch records accessible via QR code or manufacturer portal. If the code on the cell doesn’t resolve to a traceable production record, you’re holding unverified stock.
The UN 38.3 transport testing requirements also serve as a secondary verification tool: any cell shipped internationally must have passed UN 38.3, and the test report should list the exact cell model and capacity. Cross-check those fields against your COA. Mismatches between the UN 38.3 report cell model and the cells on your pallet are one of the clearest signals that a certificate has been reused across different cell configurations.
Why Do Most Buyers Accept COAs Without Running a Single Counter-Check? #
The honest answer is time pressure. A procurement engineer managing 30 SKUs doesn’t have the bandwidth to run electrical characterization on every cell batch. Factories know this. Some Dongguan-area cell traders have built entire businesses on the assumption that buyers will accept a polished COA PDF without sampling.
The cost of skipping inspection isn’t abstract. A US-based portable power station brand sourced 18,650 NMC cells with COA-stated capacity of 3,400 mAh. Post-assembly testing of finished units revealed average pack capacity 11.3% below design spec — traced back to cells averaging 3,016 mAh actual capacity. By that point, 4,200 units were already assembled. Disassembly, rework, and delayed shipment cost the brand an estimated $94,000 net, not counting the expedited replacement cell order.
Run counter-checks on every new supplier for the first three shipments, minimum. After that, a statistically meaningful sample (5% per batch, per the protocol above) is your ongoing quality gate. The UL 1642 standard for lithium batteries provides abuse testing benchmarks that can also be adapted into incoming qualification protocols for high-risk applications.
Sourcing Guidance for Buyers #
When sourcing cells from Chinese manufacturers, the first document to request is not the product datasheet — it’s the batch-level COA for the specific lot you’re being quoted, not a generic sample COA from a previous shipment. Require the COA to state test conditions (C-rate, temperature, SOC where applicable) for every electrical parameter. A COA without conditions is a marketing document.
The single clearest red flag for an unreliable supplier is a COA where all measured values align suspiciously closely with nominal spec — for example, every cell in a 1,000-unit batch showing exactly 280.0 Ah or within ±0.1 Ah. Real production batches have natural variation. A standard deviation of zero on capacity measurements means the document was generated, not measured.
Before committing to volume, run a formation cycle test on 6–12 sample cells: full charge to CV cutoff, rest 2 hours, discharge to lower cutoff at 0.5C, record capacity, rest 2 hours, repeat for 3 cycles. Capacity should stabilize within ±1.2% by cycle 3. Cells that drift more than ±2.8% across those three cycles have either been stored too long or were never properly formed at the factory.
For buyers integrating cells into battery pack designs, cell-level qualification data feeds directly into pack-level performance modeling — don’t treat cell sourcing and pack design as separate processes.
If your application involves BMS engineering with tight SOC accuracy requirements, cell-level capacity variance data from your incoming inspection protocol is essential input for BMS calibration. A BMS tuned to a nominal 280Ah cell will mis-estimate SOC by up to 6.3% if the actual cell population averages 268Ah — a number we’ve measured in real deployments.
Frequently Asked Questions #
Can I rely on a factory’s own COA without third-party verification?
For initial supplier qualification, no. A factory-issued COA should be treated as a starting point, not a final answer. Run your own incoming inspection on the first three batches from any new supplier, and use deviations between COA values and your measured results to calibrate how much you can trust that supplier’s documentation going forward.
What’s the minimum sample size for incoming electrical inspection of a cell shipment?
We use 5% of units per shipment with a hard floor of 12 cells, regardless of shipment size. For very large shipments (over 5,000 cells), 3% with a ceiling of 250 cells is statistically defensible if you’re using stratified random sampling across packaging units, not just pulling from the top layer of the pallet.
What does “Grade A” actually mean for LFP cells sourced from China?
There’s no universally enforced industry definition. In practice, “Grade A” from a Tier-1 manufacturer (EVE, CATL, CALB) means first-life cells meeting the manufacturer’s own published electrical specifications with full traceability. From a trading company, “Grade A” can mean anything from genuine Tier-1 overstock to recycled cells that have been capacity-tested once and relabeled. Always request the manufacturer’s original batch traceability code and verify it independently.
How should I handle a COA that lists cycle life as “≥2,000 cycles” without a retention percentage?
Reject it as incomplete and request the full test report. “≥2,000 cycles” without a capacity retention threshold is meaningless — a cell that reaches 2,001 cycles at 40% original capacity technically meets that claim. The IEC 62620 methodology requires retention percentage to be stated. Industry standard for Grade-A LFP is 80% retention at the quoted cycle count.
Is there a standard incoming inspection protocol I can reference or adapt?
We haven’t fully validated this across all grades and cell chemistries, but the framework we use is adapted from IEC 62620 acceptance criteria combined with internal thresholds developed from field failure analysis. There’s no single published standard that prescribes incoming inspection pass/fail thresholds for commercial cell procurement — this is an area where platform-specific protocols and supplier history data matter more than generic guidelines.
Published by compactbess.com Technical Team | Request a sourcing consultation
