1. When an off-the-shelf pack stops being the cheap option
An off-the-shelf pack looks cheaper until you count what it costs to make it fit.
The enclosure does not fit. You add a bracket, a spacer and a second wiring harness. The BOM goes up, the assembly time goes up, and the pack now has two extra failure points that nobody owns.
The discharge curve does not match. A standard 48 V pack rated at 20 A continuous will not run a 60 A peak tool, a kettle or a motor start. Buyers discover this after the first field return, not on the datasheet.
The certification does not transfer. A pack is certified as a system: cells, BMS, construction and connectors together. If your product needs UL 2054, IEC 621332 or UN 38.3 documentation, an offtheshelf pack with the wrong cell or a different BMS may need to be retested — and the certificate may not be in your name.Custom is not automatically more expensive. It is more expensive per unit and cheaper per delivered product, which is a different calculation.
Custom does not mean "we will build anything". It means six variables are decided together, and each one changes the others.
| Dimension | What you supply | What it changes |
|---|---|---|
| Voltage and capacity | Nominal voltage, Ah, and the usable Wh at your real duty cycle | Cell count, series/parallel layout, enclosure volume |
| Cell selection | Chemistry, format (18650 / 21700 / 32700 / prismatic), grade and brand tier | Cycle life, C-rate ceiling, low-temperature behaviour, cost |
| Discharge | Continuous amps, peak amps and peak duration | Busbar, nickel strip, connector, MOSFET rating, thermal design |
| BMS | Protection thresholds, balancing, communication, SoC method | Safety behaviour, host integration, whether the whole product re-certifies |
| Connectors and wiring | Connector type, wire gauge, length, routing | Drop-in assembly, voltage drop, serviceability |
| Size and housing | Envelope, mounting points, IP rating, thermal path | Tooling cost, MOQ, weight, shipping class |
A supplier who quotes you before asking about all six is quoting a catalogue item with your logo on it.
Vague specifications produce vague quotes, and vague quotes produce disputes at incoming inspection. Lock these eight before the first email.
| # | Specification | How precise it needs to be | Common mistake |
|---|---|---|---|
| 1 | Voltage window | Nominal V, upper and lower cut-off per cell | Quoting "48 V" without stating 13S or 16S LFP — the charger and the inverter both care |
| 2 | Continuous and peak current | e.g. 30 A continuous / 90 A for 10 s | Quoting one number; peak duration is what overheats the FETs |
| 3 | Cycle life target | "3,000 cycles to 80% capacity at 0.5C, 25 °C" | Quoting cycles without stating depth of discharge and temperature |
| 4 | Charge profile | Max charge current, CV voltage, charge time target | Assuming the pack charges as fast as the cell datasheet allows |
| 5 | Temperature range | Operating and storage range, plus the derating strategy | Ignoring 45 °C+ ambient — the most common field failure in the Middle East, Australia and the southern US |
| 6 | Mechanical envelope | 3D model or a drawing with tolerances, weight budget | Sending a photo and a rough dimension |
| 7 | Communication | CAN, RS485, UART or I2C, plus the exact data fields the host needs | Discovering at integration that the BMS speaks a proprietary protocol |
| 8 | Market and volume | Target countries, required certificates, annual and first-order quantity | Asking for certification after tooling is paid |
If you can answer these eight, you can compare quotes on equal terms. If you cannot, you are comparing sales skills.
Chemistry is the decision with the longest consequences, because it is the hardest to change later.
| Chemistry | Energy density | Typical cycle life | C-rate | Low temperature | Relative cost | Best fit |
|---|---|---|---|---|---|---|
| LFP (LiFePO4) | Medium | 3,000–6,000 | Medium | Weak below −10 °C | Low | Portable power stations, ESS modules, long-life industrial packs |
| NMC | High | 800–2,000 | High | Moderate | Medium–high | Power tools, drones, light-weight outdoor products |
| LTO | Low | 15,000+ | Very high | Strong | High | Fast-charge, high-power, extreme-cold applications |
| Sodium-ion | Medium–low | 2,000–4,000 | Medium–high | Good | Medium (falling) | Cost-sensitive, wide-temperature products |
Those are typical ranges, not guarantees. The pack is not the cell: a 5,000-cycle cell in a pack with a poor thermal path and an unbalanced parallel string will not deliver 5,000 cycles. Always ask for two documents — the cell datasheet and a pack-level cycle report from the supplier's own test bench.
Two packs can use identical cells and behave completely differently, because the BMS decides what the product is allowed to do.
Protection thresholds. Over-voltage, under-voltage, over-current, short circuit, over-temperature and under-temperature. Ask for the actual numbers, not "full protection".
Balancing. Passive balancing bleeds a few milliamps; active balancing moves charge between cells. On a 16S string that drifts, the difference shows up as lost capacity in year two.
State of charge accuracy. Coulomb counting with open-circuit-voltage correction holds ±3–5%; a voltage-lookup table can be off by 20% under load, which is why some power stations jump from 40% to 15%.
Communication. CAN, RS485 or UART, plus the register map. If your host firmware is already written, the register map is a hard requirement, not a detail.
Sleep current. A pack that draws 2 mA in storage is flat in three months. Ask for the shutdown current and the wake method.
Firmware and logging. Can thresholds be updated in the field? Does the BMS record fault history you can read after a return?
One wrong threshold can force your entire product through re-certification, so this is the section where "we will check with the engineer" is an acceptable answer and a confident guess is not.
Most quality problems are created in three places, long before the finished pack is tested.
Cell sorting and matching. Capacity, internal resistance and self-discharge are measured, and cells are grouped within a tight window before assembly. Unmatched cells in one parallel string mean the weakest cell works hardest and fails first.
Assembly control. Welding energy is monitored rather than guessed, insulation and Hi-pot checks happen on the line, and terminal torque is specified. A cold weld looks fine and becomes a 200 mΩ hot spot six months later.
End-of-line testing. Capacity, internal resistance, self-discharge over 24–72 hours, Hi-pot, and a full charge/discharge cycle — plus vibration and drop testing for transport compliance.
Ask for the records, not the promise. A supplier who cannot show per-pack test data is selling you hope with a spec sheet attached.
Certification is the part of a custom pack project that most often slips the launch date, because it cannot be accelerated by paying more.
| Scope | Typical requirements |
|---|---|
| Transport (all markets) | UN 38.3 test summary, MSDS, 1.2 m drop test on packaging, Class 9 lithium battery labelling |
| European Union | CE (EMC and LVD), RoHS, REACH, WEEE, EU Battery Regulation documentation |
| United States | UL 2054 for the pack, UL 1642 for cells, UL 2271 for light electric vehicles, UL 1973 for ESS, FCC if the product is wireless |
| Japan | PSE (DENAN) where the product is a specified item |
| Korea | KC certification |
| Australia / New Zealand | RCM |
| International | IEC 62133-2 for portable packs, IEC 62619 for industrial packs |
Two practical rules. First, the certificate must be in the name of the party that will place the product on the market, or issued to the factory with your brand listed — check the certificate holder before you print packaging. Second, certification follows the cell, BMS and construction. Change any of the three and you may need a new test series, which is why the eight specifications in section 3 should be locked before tooling.
Two quotes that differ by 30% are almost never the same product. In a typical custom pack, cells account for roughly 40–55% of material cost, the BMS and electronics 10–20%, the enclosure and hardware 10–20%, assembly and testing 5–10%, with tooling, certification and freight on top.
What moves the number:
C-rate. Higher continuous and peak current means thicker busbars, more copper, better FETs and a bigger thermal path.
Temperature range. A pack specified for −20 °C to 60 °C needs different cells, heaters or derating logic, and more testing.
Communication and firmware. A custom register map and host integration is engineering time, not material cost.
IP rating. IP67 adds gaskets, potting and test time per unit.
MOQ and tooling. A custom enclosure amortised over 500 units costs far more per unit than the same enclosure over 20,000.
Cell grade. "A-grade" and "A-grade, matched" are different purchases. Ask for the sorting report.
The cheapest quote is usually cheapest for a reason that appears in the field, not on the invoice.
| Stage | Typical duration | What you need to provide |
|---|---|---|
| RFQ and quotation | 3–7 days | The eight specifications from section 3 |
| Design and samples | 10–20 days for a standard platform, 25–35 days with new tooling | 3D model, connector preference, target cost |
| Verification and certification | 4–10 weeks, longer for UL | Sample approval, importer details, test budget |
| Pilot run | 15–25 days | Frozen specification, packaging artwork |
| Mass production | 25–45 days depending on volume | Purchase order, deposit, forecast |
| Sea freight | 25–40 days to the US or EU | Correct UN 38.3 documentation and labelling |
Plan three to five months from specification to the first container, and six or more if UL testing is in scope. A supplier who promises eight weeks including certification is either not certifying or not shipping by sea.
Comparing price per Wh. It ignores cycle life, derating and the cost of a field return.
Accepting "we have all the certificates" without reading the holder name. A certificate issued to another factory does not protect your importer.
Skipping the pilot run. Fifty units will find the assembly problem that five samples never show.
Ignoring high-temperature derating. A pack that delivers 100% at 25 °C may deliver 70% at 45 °C — plan capacity for the hottest week, not the lab.
No spare parts or firmware clause. Ask what happens in year four when a customer needs a replacement BMS.
No test method in the specification. If the acceptance criteria are not written down, incoming inspection becomes a negotiation.
CHYU POWER builds custom lithium battery packs for B2B programmes, with a working range of 12 V to 80 V, discharge capability of up to 50C for high-current applications, and pack designs rated for 3,000+ cycles depending on chemistry and duty cycle.
Customisation is handled as the six decisions described in section 2 — voltage and capacity, cells, discharge, BMS, connectors and wiring, and mechanical housing — followed by three checkpoints: cell sorting and matching, controlled assembly, and finished-pack electrical and protection testing.
Typical programmes include battery packs for power tools, e-scooters and light electric vehicles, drones, portable power stations, outdoor power products, AGVs and medical devices, as well as modules for commercial and industrial storage.
Minimum order quantity: 〔MOQ:___ pcs〕
Sample lead time: 〔___ days〕
Monthly capacity: 〔___ packs / month〕
Certification support: 〔按实际填写,例如 UN 38.3 / IEC 62133-2 / UL 2054 报告是否已具备〕
Send a specification, a drawing or even a competitor sample, and you will get a pack proposal with the BOM assumptions and the test plan attached — not just a price.
What is the minimum order quantity for a custom pack? It depends on whether new tooling is required. Packs built on an existing platform start lower; a new enclosure or a new BMS layout carries tooling and engineering cost that has to be amortised. Send your target volume and we will tell you the honest break-even.
How long does a custom sample take? Around 10–20 days on an existing platform, and 25–35 days when a new enclosure or a new BMS configuration is needed. Certification, if required, runs in parallel and is the longer path.
Can I specify the cell brand or use cells I supply? Yes. Cells can be selected to your brand preference or supplied by you, provided the grading and matching data is available — mismatched cells are the most common cause of early capacity loss.
Who handles certification? The factory provides test samples, technical files and the documentation package; the certificate is normally issued to the brand owner or importer. We will tell you up front which tests are needed for your target market and what they cost.
What does the warranty cover, and what about spare parts? Warranty terms are agreed per programme, and replacement BMS boards and connectors can be stocked against your forecast so year-four service does not depend on a single production run.
Do you also build modules for commercial and industrial storage? Yes — the same cell sourcing, BMS and testing process is used for storage modules, which is why buyers often consolidate portable and stationary programmes with one supplier.
Still comparing suppliers? Send your specification and we will return a pack proposal, a test plan and a realistic timeline — including the parts of your current design that will cause problems in year three.