What a supplier's UN 38.3 test summary must show you

A UN 38.3 test summary is evidence only when every identifier on it traces to one exact cell and one exact pack configuration: the same cell manufacturer and model, the same series and parallel count, the same protection circuitry, the same mechanical construction. It also has to be a summary of tests that were actually run on samples built to that configuration. If you cannot match the document line by line to the bill of materials and the cell data sheet you are buying against, it describes some battery. It does not describe yours.

I read these with the BOM open on the left of my screen and the PDF on the right, and I read the identifiers before I read a single result. That order is most of the job. A page full of passes tells you nothing until you know what passed.

The file usually shows up as three or four pages riding along with a quote, sometimes with a lab stamp scanned slightly crooked in the corner. There is a description block at the top, then a table of tests with pass or fail in the right-hand column, then a line about samples. Most buyers scroll straight to the pass column. I start at the top, in the description block, and I stay there long enough to get uncomfortable.

The description block is the argument, not the header

A summary is supposed to identify the cell and the battery separately, because they are separate test articles. So there should be a cell description and a battery description, each with a manufacturer, a model designation, and enough electrical detail to pin it down: capacity and energy for the cell, rated voltage and watt-hours for the assembly, and a note on how many cells are inside and how they are arranged.

I read the cell model number character by character, including the suffix. Suffixes are where reformulations live. A cell that gets a new cathode blend or a new separator often keeps its family name and picks up a letter at the end, and the old summary keeps circulating with the new quote because nobody updated the file.

Then I do the arithmetic. If the summary says the pack is 48 volts and 20 Ah and the configuration line says 13 series by 6 parallel, I multiply it out and see whether the numbers hold together. They frequently do not, because the summary was written for a 13S4P variant and the quote is for a 13S6P. Same cell. Same housing family. Different battery, and the test evidence does not transfer.

Here is the sheet I actually keep in the procurement file, filled in by the supplier, not by me:

Field	What goes here
Cell manufacturer and model	Exact string, with suffix
Cell capacity and chemistry	Ah, Wh, chemistry
Pack configuration	Series x parallel
Pack rated voltage and energy	V, Wh
BMS or protection circuit model	Part number and revision
Housing and internal construction	Material, potting, cell holder type
Summary document number and date	As printed on the summary
Standard revision cited	Revision and any amendment

When that table is filled and the three or four identifiers overlap with the summary, I have something I can defend. When a supplier will not fill it, that is information too.

The cell was tested, so what

This is the sentence I hear most: the cell is already tested, the cell maker has the summary. Yes, and that summary covers the cell. A completed battery assembly is a different test article, and the pack-level test set is broader. Cells go through altitude simulation, thermal cycling, vibration, shock, external short circuit, impact or crush, and forced discharge. Assemblies go through those plus an overcharge test on the finished pack.

So I look for pack-level results, on the assembly, with the overcharge line present. If the document I have been handed only shows the cell test set, then what I am holding is a cell summary stapled to a pack quotation. It is not wrong. It is just not about the thing on the purchase order.

I have also seen the reverse, where the summary is a pack summary but the samples line only counts cells. Read that line. It should tell you how many of each article was tested, and a summary that is vague about sample count is vague for a reason.

The revision line nobody reads

Near the document number there is a reference to which revision of the standard the testing was done against, and sometimes an amendment. I check it, and I check the date against the date the design was frozen. A summary that is several years old for a cell that has been through two formulation updates is a historical document, not a current one. The standard itself gets revised, and a summary citing an old edition is worth a question to the supplier about whether anything has been retested since.

I am not the person who decides what a carrier will accept, and nothing about holding a summary forces a carrier, a forwarder, or anyone else to take the shipment. That is between the shipper and the carrier and it varies by mode and by operator. What the summary does is give you a defensible answer when someone upstream asks what your evidence is. Whether they accept it is their call.

The word certified is a small tell. UN 38.3 is a test regime, not a certification scheme. A spec sheet that says UN38.3 certified, with no document number behind it, is a marketing claim about a document. Ask for the document.

Summary, report, and the safety listings are three different things

A test summary is a defined, short format: description, tests performed, results, samples. A full test report is the long version, with the lab's raw data and the conditions. Suppliers almost always send only the summary, and that is normal. If I need more, I ask for specific pages of the report under NDA, and I ask for the report number that appears on the summary so the request is easy to fulfill.

Separately, UN 38.3 addresses transport. It is not a product safety listing. UL 2054 and UL 2271 are different programs with different scopes, and a pack can have one without the other. I have watched a program stall because someone assumed a UN summary satisfied a customer's product safety requirement. Build two separate evidence columns in your file and do not let them substitute for each other.

The paper only stays true if the build stays frozen

This is the part that gets lost after the purchase order closes. The summary is tied to a tested configuration. Change the cell, change the protection circuitry, change the internal construction or the housing in a way that affects the cell environment, and the tested configuration no longer exists. That can mean retesting, and at minimum it means new paperwork.

So I write the configuration into the agreement rather than into a nice email. A change notification clause with a defined notice period, a requirement that any change to cell, BMS, or construction triggers a documented review, and a line that says the supplier will not ship a changed build against the original summary. I also name the configuration in the PO itself, by cell model and series-parallel count, so the summary, the PO, and the first article inspection all point at the same build.

When the pilot units arrive, I pull one apart far enough to read the cell wrapper and compare it to the summary. It takes ten minutes and it has caught a substitution for me. The wrapper is the ground truth; everything else is a document about the wrapper.

What I actually do now, in order

I open the description block and find the cell model with its suffix. I find the pack configuration and check the arithmetic against the rated energy. I confirm there are pack-level results, not just cell results, and that overcharge is among them. I note the revision cited and the test date. I copy the document number into the procurement file so the underlying report is one email away. Then, and only then, do I look at whether anything failed.

If the identifiers line up, the results column is the least interesting part of the page. The identifiers are where the risk sits, and they are the part almost nobody reads.