The 5-year paper trail behind every lithium battery pack entry

A lithium battery pack entry file is a five-year record, and the clock starts on the date of entry, not the date of clearance. The importer of record — not the broker, not the forwarder — is the party that has to produce the complete package for a given entry number years after everyone has stopped thinking about that shipment. The two ways to fix something inside that file also run on separate clocks: a post summary correction while the entry is still open, and a protest inside the 180 days that follow liquidation. Clearance is the day the file starts working, not the day it retires.

The request I remember came as a two-line email from our broker with a scanned letter attached. I printed it, because I wanted to hold the thing that had just made my week complicated: a CBP request for information, form 28, one entry number from almost two years earlier, three questions, and a response date roughly four weeks out.

Stand at my shoulder while I open the drive. The folder is named for the entry number. Inside are four files — the 7501, the commercial invoice, the packing list, and a bill of lading. That is the entire record of a shipment of several thousand lithium-ion packs.

For this commodity, a four-file folder is thin. The three questions were about cell chemistry, the revision of the UN 38.3 test summary that applied at shipment, and how the entered value was built. I had none of it in front of me. The test summary had gone to the forwarder as an attachment two buyers ago. The dangerous goods declaration had gone with the shipment and lived in a shared inbox that had since been migrated. The value build was inside a purchase order in an ERP that had been re-implemented since.

It took the better part of two weeks and a chain of emails across three time zones to answer three questions. The answers themselves were boring and correct. The cost was the hunt, and what still bothers me is that nothing about the hunt was unusual.

Build the folder around the entry number

CBP asks by entry number. Not by supplier, not by purchase order, not by part number. That sounds obvious until you look at how most teams file: by supplier, because supplier folders are how purchasing already works, or by program, because that's how engineering thinks. Then the letter arrives and someone has to remember which supplier shipped in March of the prior year.

So the top level of the drive is the entry number with the year leading — something like 2023-04-118, then the supplier name, then the entry date. Year first means the folder list sorts chronologically without anyone maintaining a sort order.

Every folder also gets a plain text index at the top, written the week the entry is filed, listing what's present and, more importantly, what isn't and where it lives. The index is the part that makes the folder survive the person who built it. When the buyer who owned that supplier leaves, the index is still there saying the DG declaration lives in the forwarder's portal under the booking number, and the UN test summary is attached to the supplier's conformity file.

The documents that go missing are the shipping documents

The entry-side paperwork arrives on its own. Invoices, packing lists, the 7501, the bill of lading — those land in your inbox because they're transactional and someone is chasing them. The battery documents travel on the logistics side and never touch the customs entry: the shipper's declaration for dangerous goods is signed by the shipper and filed with the carrier, the UN 38.3 test summary is requested by the forwarder, the safety data sheet is maintained by the cell maker. Three different parties, none of them your broker.

That is precisely why they're missing when a reviewer asks, and precisely why a battery entry is one of the likelier ones to draw the question.

What I keep per entry, alongside the customs documents:

• The signed dangerous goods declaration for that specific shipment, not a template.
• The UN 38.3 test summary revision that was in force at the time of shipment, and I keep the superseded revision rather than overwriting it. The current summary in the supplier's portal does not prove what was tested two years ago, and overwriting it destroys the only document that does.
• The safety data sheet revision current at shipment, with the revision date visible on the page.
• The supplier's declaration of conformity or certificate of conformance, if one was issued.
• The purchase order and the pricing build behind the entered value.

A reviewer working a battery entry is trying to establish one thing: that the article entered is the article that was tested and the article that was shipped. Those documents answer it in about ten minutes. Without them, the same review turns into a supplier chase across time zones, and the response date doesn't move.

Two clocks, and only one of them is a courtesy

Liquidation generally lands somewhere around 314 days from the date of entry, and it gets extended often enough that you shouldn't treat the date on an ACE report as final. A post summary correction lives between entry summary acceptance and liquidation, with a short tail after the liquidation date — fifteen days is the number I work to, though this is one of those figures people misremember, so check the current guidance rather than my habit.

The practical consequence is that a PSC filed after liquidation isn't a PSC. So for entries where the classification, the origin claim, or the value was a judgment call, I keep a short list with the liquidation status on it. A list, not a memory. If a question comes up eight months later, the first thing I want to know is whether the window is still open, and I want to know it in a minute.

A protest is a different instrument with a different window — 180 days after liquidation — and which one applies to a given set of facts is a question for counsel. My job in the file is narrower: know which window is open, and have the record assembled before anyone starts arguing about it.

Answering a 28 without handing them a new question

The letter is usually narrower than it feels. Two or three questions, tied to specific line items. The failure mode isn't refusing to answer — it's over-answering. The supplier pack that gets forwarded "just to be helpful" frequently contains a superseded test summary, a sibling model with a different capacity, or a spec sheet naming a chemistry that doesn't match the invoice. Any of those converts one question into two.

Three habits I hold to:

Answer in the order asked, one question at a time, each tied to the document that answers it and a page cite. Attach the specific document, not the bundle. And don't reclassify inside the response — if the honest answer is that the entry was wrong, say so plainly, put a number on the exposure, and route the fix through the correction path with the broker rather than improvising a new position in a letter.

Two smaller ones: get the response in before the date, and if you can't, request more time in writing before the date passes. And run supplier technical questions through a single channel so you get one version of the chemistry instead of three.

What the broker keeps, and what you actually own

Brokers retain entry records too, generally for the same five years. That is not a substitute for holding your own copy. Broker relationships end, firms merge, historical files sit in a system you don't have credentials for, and the file does not walk out the door with your account manager. I have been quoted a research fee and a couple of weeks for a retrieval from a broker we no longer used.

The split I work to: the broker holds the transactional copy, the importer of record holds the authoritative copy. The obligation belongs to the importer, and treating a vendor's archive as your archive is a way of renting your own compliance from someone who can change the terms on renewal.

Five years from the entry date, not from the fiscal year

The retention period is generally five years from the date of entry. That single detail breaks most purge routines, because entries within one fiscal year expire on dates spread across twelve months. Purge by "files older than five years" and you will destroy some records early and hold others well past their date.

Two carve-outs from any sweep: anything with an open question on it — a pending request, a correction in flight, a liquidation that hasn't happened — doesn't retire on schedule. And if you're relying on a supplier or a third-party warehouse to hold a piece of the record, get the commitment in writing and re-verify it annually. A supplier portal is a convenience, not a retention program.

Then test the archive instead of admiring it

The habit that has bought me more time than any folder convention: once a year, pull five entry numbers from three years back and try to assemble the full package in one working day without emailing anyone outside the company. Retrieval is the only measure that matters, and it fails quietly. A folder that looks complete in a screenshot can be missing the one document a reviewer will ask for, and you won't learn that until the letter arrives.

Which is why the last request I got looked nothing like the first. It landed on an entry from a few years back, two questions, a date about four weeks out. I opened the folder, read the index at the top, pulled the invoice, the value build, and the test summary revision that shipped with that lot. The response went out the same afternoon with time to spare.

Nothing about that day was diligence. It was a folder built on the day of entry, by someone who had stopped believing that clearance was the end of anything.