Buying Compatible Battery Packs from China: The Stuff I Check Before Anything Else

The first battery pack I ever sourced from China taught me the most valuable thing I know about this category, and I only learned it after a customer sent back a vacuum with a note that said "runs 9 minutes, not 35." The pack was labelled 4000mAh. It measured 2380mAh on a proper discharge rig. Nobody had lied on the invoice; there was simply a housing full of cells that had been previously used to build something else.

Compatible battery packs are the awkward middle of this market. You are not buying a house-brand pack with a factory's own name on it, and you are not buying a loose reel of cells either. You are buying a box of cells plus a protection board plus a plastic carrier plus a connector, and the quality of any three of them can be perfectly good while the fourth ruins the product. Here is what holds up in practice.

Start with the cell, not the pack

Ask who made the cell. There are now maybe a dozen names in 18650, 21700 and pouch that most factories can actually source reliably and that report honest specs: the large Korean and Japanese makers, and the big Chinese suppliers such as EVE, Sunwoda, Sunwoda-derived lines, Lishen and a few others. If the answer is "our own brand" with no cell maker named, the pack is probably being assembled from whatever the local cell market had that week.

Grade matters more than brand. In the trade you will hear A, A- and B. Grade A means capacity and internal resistance within the datasheet window, from one production lot. Grade B means the cell is out of spec somewhere and is sold cheap. The dangerous stuff is not B, which at least usually reports something close to reality. The dangerous stuff is welded-tab cells pulled out of returned packs, sometimes called "recycled" or discreetly "reclaimed." Those have already done a few hundred cycles before they are even built in. A pack built from them might pass an initial capacity test and then fall off a cliff after a month.

Simple way to catch it without a lab: ask for the cell production date codes and the factory's purchase invoice from the cell maker. Both are normal commercial documents. A supplier who cannot show a cell invoice for the quantity you are ordering is not buying that many cells.

Capacity is a measurement, not a label

Batteries are one of the few categories where you can measure the main claim yourself cheaply. Buy three sample packs, run them on an electronic load at the datasheet discharge rate, and write down the actual Wh. Then ask for the same test on the production batch you are about to accept, at your own load profile, not the factory's.

Two numbers turn the 4000mAh label from a boast into information. First, measured capacity at the discharge rate your product actually draws. Second, measured capacity at the temperature your product actually sees. A pack that delivers 3800mAh at 25 °C often delivers 2600mAh at 0 °C, and if your product is a cordless tool in a cold store, the second number is the only one your customer experiences.

The protection board does the quiet work

Almost all the safety-critical behaviour sits in the BMS board, which is usually a $1 part nobody inspects. What I want written into the spec:

• Over-discharge cutoff voltage, set per cell chemistry, and the recovery behaviour (does it latch or auto-recover).
• Over-current trip point and whether it is a one-time fuse or a resettable protection IC.
• Whether the board has a temperature cutoff and where the NTC actually sits, because an NTC buried in the middle of a pack reacts to the cells while one glued to the housing reacts to the room.
• Whether cell balancing exists in the multi-cell version, since an unbalanced 3S pack loses runtime long before it loses capacity.

Ask for the BMS schematic, or at least the IC part number. An uncovered marking on the chip is a red flag; it usually means a clone part made to look like a name brand, and the trip points on those drift.

UN38.3 is a transport document, not a safety guarantee

Every air shipment needs a UN38.3 test summary plus an MSDS. Buyers routinely treat those as proof of quality. They are not. UN38.3 says the design survived eight specific abuse tests, on the exact sample submitted. It does not tell you anything about cycle life, real capacity or whether next month's cell supplier is the same as this month's.

Check that the model number on the UN38.3 summary matches the pack you are ordering character for character, and that the summary is not for a different cell inside a similarly named pack. I have seen one report reused for four different products with different cell counts. Also look at age. A 2019 summary with a 2026 build is technically possible and practically meaningless unless the design is frozen.

Connectors, carriers and the boring stuff that kills returns

The most frequent field failure I have had in this category is not thermal; it is mechanical. Pack carriers are moulded with tolerances that drift, and a 0.4mm error on the latch means the pack rattles or does not seat. Ask for the mould to be run on the same machine for samples and production, and check latch force with your own gauge on the first article, not on the drawing.

Wiring also needs a look. Silicone insulated lead vs PVC matters in anything that flexes or gets warm. Nickel strip thickness on welded joints should be written into the spec, because 0.1mm strip with a partial weld heats up under sustained load. And the connector must be rated for the peak current, not the average: a JST-type connector rated 3A on a tool drawing 12A peak is a slow-burn failure.

How I structure the order

Three samples, measured, then a small pilot order of one or two hundred pieces with a written acceptance criterion: capacity at your discharge rate within 5% of the sample, at least two temperature points recorded, and cell date codes reported. Pay the extra 3–6% for the cell brand you specified. In my experience the cost gap between unknown cells and a named cell is smaller than one return shipment, and after two years of field data it is not even close.

One last thing. If the factory can quote a "same as original" pack without any question about your device's charging profile, they have not understood the job. Charging is where these packs fail most often, and a good factory will want to know what the charger actually does before they quote.