How to Accept or Reject Battery and Assembly Changes on an Electric Tricycle

The expensive mistake on a tricycle program is rarely the battery or the bolt you picked at quote stage. It is the one that replaced it after you signed off: a pack swapped to protect a lead time, a fastener set changed because the original grade ran out, while the frame, the controller part number and the paint code stay exactly as approved. That divergence shows up as rework on arrival, a container held while you explain yourself, and a liability file you end up building after a failure instead of before it.

Two baselines carry most of that risk, and both are documents rather than agreements: the approved battery pack and the approved fastener set. Because a factory can alter either one without touching the main drawing, they are the two subsystems that need a gate, and the gate is a recorded accept or reject before anything ships. What follows is the order to run it in, what each step has to produce, and where the paper stops proving anything at all.

What a Wrong Battery or Assembly Change Really Costs

The cost does not show up as a price gap. It shows up as divergence: a gap between what you believe you bought and what sits in the crate, and the unit still passes incoming inspection because it looks right, rolls right and charges right. It surfaces later, under conditions nobody designed for: a pack that was never matched to the charger, a fastener that gives out under a load case the frame was tested for with different hardware.

This lands on procurement rather than the factory because ownership runs one way. The factory is answerable for what it was told to build; only you hold the baseline that says what that was. Without a revision number attached to the pack and to the hardware, there is no fact of the matter about whether the delivered unit is the approved unit, and every argument after that becomes a memory contest between two teams in different time zones.

The Battery Pack Is the Part You Cannot Substitute

The regulatory language is unusually blunt for a safety letter. CPSC's February 2023 letter to e-mobility manufacturers and testing stakeholders tells the supply chain to "ONLY use battery packs recommended by the e-mobility device manufacturer" and to "NEVER use an e-mobility device with a battery pack that has been modified/reworked by unqualified personnel or with re-purposed or used cells."

The reason a pack is not a commodity part is that pack, BMS and charger are qualified as one matched system, and the BMS manages a known cell behaviour rather than a nominal voltage. Same voltage, same shape, same connector is a fitment claim, not an equivalence claim, which means a cell-source change is a pack change even when the enclosure and the connector are untouched.

That letter is a 2023 document carrying a 2022 compliance call alongside safety recommendations, and it certifies no specific tricycle. Its scope has to be split between UL 2272 and UL 2849 for the vehicle type in question and rechecked against later rulemaking before anyone treats it as the requirement set for a given model.

Run the Change Request in This Order

Freeze the baseline before any request arrives, while there is nothing to argue about: pack specification, BOM revision and hardware specification, each with a number and a date. Classify the request in the requester's own words, with pack, cell, BMS or charger on one side and assembly hardware and fasteners on the other; cosmetic goes last, since it still needs a record but not a sample comparison. Then attach each item to the document that governs it. A cell change lands on the pack specification, a bolt change lands on the hardware specification, and a request that fits neither is a new BOM item rather than a footnote to an old one.

Ask for the supplier's written documentation of the proposed version, naming the part rather than the product category, and compare it against your frozen baseline instead of against the supplier's summary of the improvement. Record accept or reject with a revision number. If it is accepted, treat the sample as unapproved until a unit built to the new revision has passed the same checklist the old one passed. That last step is the one teams skip, and it decides whether the change exists on paper or in the crate.

The Checkpoint That Decides Each Step

Every step has one narrow completion test. Freezing the baseline is done when a revision number exists that someone can quote back to you; before that you have a conversation. Classifying the request is done when the change is written down rather than named, because paraphrase is where "we may need to look at the cell source" quietly becomes "no change." Requesting documentation is done when the document names the part, the manufacturer and the version rather than describing a family of products.

The reason checkpoints beat approvals comes down to what survives a staff change. An approval is a memory held by whoever said yes, and two memories of one phone call differ across a language and a time zone. A revision number sitting in a file does not. Which means the deliverable of each step is an artifact, and a step without an artifact did not happen, however clearly everyone remembers agreeing.

Why a New Bolt Is Also an Assembly Change

Hardware gets filed as housekeeping and it should not be. The text captured in 16 CFR part 1512 requires that all screws, bolts, or nuts used to attach or secure components shall not fracture, loosen, or otherwise fail their intended function during the tests required in that part, and that all threaded hardware shall be of sufficient quality to allow adjustments and maintenance.

The reason a cheaper bolt is a design change rather than a purchasing saving is that the obligation attaches to function under test and to maintainability, and neither of those is a part number. A fastener swapped for a lower grade, a different coating, or a thread that stops tolerating repeated adjustment answers the same requirement differently, which puts it behind the same gate as a pack change.

Scope decides how far that reaches. The wording above is the version captured in the 2025 annual CFR edition, and whether part 1512 covers your particular vehicle is a per-model question rather than a category one. It is not a parts certification, and it says nothing about road legality.

A Mid-Run Cell Swap, Walked Through

Suppose the order is half built when the factory writes to say the cell source it quoted is short and proposes an alternative to keep the line moving. Nothing about the enclosure, the connector or the voltage changes. Walk the same steps and three different things happen at once.

Part of it is answerable immediately, with no test data at all: nothing gets approved on the strength of a sentence. An assurance that the new cells are "the same spec" describes a product category, not a part, and it does not satisfy the documentation step. Part of it has to be requested in writing and cannot be covered by your own judgment, namely a pack specification that names the recommended pack, a written statement from the supplier that no reworked or used cells are involved, and an updated BOM revision that supersedes the number you froze. Here is where it gets messy: the proposal almost always arrives as a helpful message from an account manager, and the first real answer is simply to ask for it in writing, because that one request turns a favour into a record.

The rest has to wait. Whether the proposed pack is equivalent to the approved one is a question for the sample comparison and for a written agreement on what gets tested and against which criteria, not something you settle in the same week the line is idle. That is the whole trap: the schedule pressure is real, and the only thing you can decide under pressure is what to request, never what to accept. No price, lead time or test result belongs in that decision until the supplier writes one down and someone checks it.

Prove the Change With Paper, Not Promises

The acceptance file is short enough to maintain and specific enough to inspect against. It holds the approved BOM revision, the pack specification naming the manufacturer-recommended pack, the supplier's written statement on cell sourcing, the hardware specification with the test reference it answers to, a deviation log recording what was asked for and what was refused, and an incoming checklist keyed to those same documents so the inspector compares the crate to the file rather than to recollection.

What the file cannot hold is anything you were never given. For suppliers in this category, custom capability, MOQ, pricing, lead time, available tooling and certificates stay open questions until someone answers them in writing, and no sample test or factory audit has been run here. Those live in the RFQ as explicit questions with a requested document behind each one, while test conditions, tolerances and acceptance thresholds get written into the agreement later against a specific model. Filling those blanks with an assumption is how an acceptance file turns into fiction.

What This Discipline Does Not Prove

A document check proves what was written and when. It does not prove the tricycle in front of you is compliant today. The two limits carried over from the sources above are dated ones: guidance issued in a given year, and a text captured in an annual edition rather than a live compilation. Neither tells you the rule in force on the day your container loads.

Which means the final step belongs before shipment, not in the file cabinet: confirm the rules in force, confirm the standard version that applies, and confirm the certificates for the exact configuration you are shipping, from the documents themselves rather than from a supplier page. Those three confirmations are dated facts. Your acceptance file is only as current as its last date.

The Tradeoff You Are Actually Accepting

Running this buys a specific set of things and costs a specific set of things. What you get is fewer silent substitutions, an acceptance file you can hand to a customer, a carrier or an insurer without reconstructing it, and a much shorter root-cause path when a unit fails in the field, because you can point at the revision that was shipped. What you give up is speed.

Every legitimate change now waits on three artifacts, a revision, a document and a re-approved sample, and that wait lands in the weeks when the line is already running. It also spends goodwill with factories that prefer a verbal yes and read a written request as distrust. Choose the gate where battery and assembly risk dominates your exposure, which on a US tricycle program is most of the time. Skip it and the saving is administrative convenience, paid for at the far end by whichever unit fails first.