The Search for a Universal Bar Number

In the bullion bars database we have the signatures for about 250,000 gold bars and 800,000 silver bars. I'm at the stage now where I need a consistent referencing number — currently the unique identifiers get created fresh each time I import a major set of data, which solves a bunch of other problems but makes it difficult to permanently catalog interesting bars. If you're not into 'Barspotting' please feel free to tune out - this is another of those posts where I'm just talking out loud, trying to solve some boring questions open for discussion.

The 2009 Project Mayhem article on Zero Hedge (original article, plus my backup copy) was to my knowledge the very first example of proper 'barspotting' - their document attempted to explain the presence of 'duplicate serial numbers' in the SLV data. A few times I have been tempted to adopt their classifications for my study but found them unworkable - the only real definition of significance is their 'Perfect Internal Duplicate' (match on serial number, weight, brand) but their study failed (IMO) due to bias but the effort was still a stepping stone in barology.

Finding a truly unique bar number is difficult because it is known that some refiners restart their numbering sequence (like Tanaka gold bars). Additionally, there is no standardization on Refiner names or codes and even the weights will sometimes vary - e.g. bars being delivered to GLD sometimes have weight inconsistencies and this may vary from vault to vault.

Close-up from the Bob Pisani sequence - correct bars weights have a tick ...
but some other bars show up as a different weight to six significant figures.
I've recently discovered a technique which highlights (easily) bar signature changes - the quantity is semi-significant - adding back over 5,000 bar signatures which were previously thought to be 'lost bullion'. This will change the Defrag Chart slightly and is just one of those things that I have to clean up before a good quality picture will emerge. Getting there is slightly more difficult and requires the construction of a universal bar number identifier. It will involve the following elements, in order of what is typically encountered on a standard bar list:
  • Refiner/Brand
  • Year
  • Serial Number
  • Purity/Assay/Fineness
  • Weight
The definition should be flexible enough to use for other metals (Silver, Platinum, Palladium). You'll may note the 2009 project mayhem study did not include 'year of manufacture' which is a shortcoming of their study but in fairness the information is not always recorded consistently in most bar lists except in cases where it is part of the serial number. It has only been the last couple of years where the LBMA has stipulated that recording the year is preferable and this is also a source of serial number changes because it is often used as a prefix.

Let's rework this with a bit more detail to get the rough parts. I'm also rearranging them in order of what I think works. Please bear in mind I am not a member of the international standards organization ; p
  • Refiner/Brand - as a four-letter alphanumeric code. Currently there is no standard universal code list but the best one is from CPM GROUP. Using this consistently would require the establishment of codes for the historical refineries and published somewhere permanent (which would be challenging). In the case of similar names, a number can be used in the fourth position.
  • Year (of manufacture) - where available. Four digits for simplicity, follows a formatting convention that most people are familiar with: YYYY. Where the full or exact year record is not available or known, "0000" should be used.
  • Purity/Assay/Fineness - expressed to four significant figures where possible. If the fourth digit is not known or is not available, an "X" should be used (rather than zero) to indicate the value was not recorded. This will force the fineness to be evaluated as text rather than a number.
  • Weight (Gross) - I have elected Gross weight as being the unit of measure because in the case of gold bars, the fine weight can be determined using the fineness/assay (and in the case of the lesser metals, a Fine weight is not recorded). Silver is typically only recorded to one decimal place, but gold has three decimal places so that will be the standard. Finally, silver has weights up to 1000 oz so we'll use seven significant figures. Measurement is in OZ, since that's the obvious industry standard. This does mean that some platinum bars measured in kg cannot be stored.
  • Serial Number - As appears on the bar list. Note: The fullest bar serial number seen should be used. i.e. leading zeroes included where they are recorded (even if removed later) and year included as the prefix where it has been included in the records, even if this duplicates the information in year of manufacture. Because the serial number is variable length, it will be placed last so that the positioning of all the other information is consistent.
Each section of data will be separated by a full stop, just like a web address. In practice, here is what it would look like. Note I have used some images from the old Bob Pisani Revisited post, to illustrate how we can sometimes use clues in bar pictures to know the full definition.

^^ This refiner is OJSC Krasnoyarsk bar as identified by their 'Россия' stemp, and while theirs is not in the CPM Group listing, the most obvious four-letter code is OJSC. The year of manufacture (6) we only know from pictures like these and fortunately we have records in the bar list which measure the weight in oz. The universal bar number is:


^^ Rand Refineries bar, which allows us visiblity to the year of manufacture (which may also apply for most of the 'ZA' series). The weight and fineness is known from the bar list data. Universal bar number is:


^^ A picture Bron sent me of a (beautiful) old silver bar, recently recovered from the seabed floor from a sunken WW2 ship. This bar is (obviously) not registered in any ETF records, and even the refiner (Indian Government Mint) is not seen in any modern records. Nonetheless with our numbering system we should still be able to catalog it, something like:


All you scientists and barologists out there, please feel free to provide suggestions to make this better (readability), etc. I'm just floating it as a proposal. The primary advantage is that it's reasonably easy to read and that the positions are fixed and consistent. Note that there is currently no distinction between metals - I'm not sure how to add that since 'G' for gold, 'S' for silver breaks down once you have P for platinum and palladium (in my database, gold = 1, silver = 2, etc). The type of metal could typically be derived from weight.

Update: 22nd April 2014.

Added Bron's suggestion (Below) regarding currency code prefix, samples in this article become:


Now just gotta work on the delimiter. And also addition of other information like 'obsolete/superceded/dead' bar signature (probably as a suffix).


Marks said...

Off topic a bit. Is there any way to identify which 400 oz gold bars are being converted to 1 kilo gold bars? Is there any manufacturer of choice or fineness of choice. Can barspotting in anyway identify "dead bullion" (i.e. 400 oz bars recasted into kilo bars or jewelry). And most importantly, are serial numbers and/or refiner names listed on 1 kilo gold bars.

Marks said...

Meant to say, are serial numbers and refiner names REQUIRED on 1 kilo bars.

Warren James said...

Hi Marks, 'Dead Bullion' is a great label for what you describe! (can I use it with attribution?).

It's my understanding that once bars are destroyed (melted down) the only records kept belong to the refiners. I think I recall Bron mentioning that they don't bother recording the numbers they melt (for coins).

To my knowledge, smaller bars normally do have the refiner stamp (because it's branded) and some have serial numbers but as far as I know it's not consistent. There's a site I reference regularly which has a lot more info on those smaller bars.

I've been thinking about appending an extra value to the end of the unique bar number to indicate it's known status. Specifically it would be worthwhile knowing if the signature has been superseded (with extra detail), or 'dead' in your suggestion where it is known to have been melted down.

In theory, the LBMA chain of custody gets broken and thus dead bullion could be traced based on 'who had it last' but getting access to this information would probably be near impossible.


Bron Suchecki said...

I think you need a prefix for the metal for completeness, eg ISO currency codes for PMs:


I'm not so sure about the "." delimiter as it makes it hard to distinguish between the delimiter and decimal point in the weight. It may also be a problem if you want to use this format to transmit bar data or import into Excel (which we know the vaults use).

Go with the old comma delimiter?


I would note the use of commas as the decimal symbol in some countries so maybe some other symbol unlikely to be used in bar numbers?

Bron Suchecki said...

If you wanted to get industry acceptance, maybe alignment with SWIFT standards may help but these don't seem particularly advanced for PMs.

See MT604 for example

Scenario: ABC Investments, Toronto, instructs The Bank of Nova Scotia, Toronto, to deliver 500 Britannia coins to Samuel Montagu & Co. Ltd., London, for the account of Roberts Gemstones Ltd., London, on 12 November 1992.

Message Type 604
Receiver MGTCGB2L
Commodity and Commodity Account :26C:/LONDON/ALLOCBRIT1/1
Value Date :30:921112
Transaction Reference Number :20:TA00900
Related Reference :21:NONREF
Further Identification :23:DELIVERY
Quantity of the Commodity :32F:UNT500,
Receiver of the Commodity :87A:SMCOGB2L
Beneficiary of the Commodity :88D:ROBERTS GEMSTONES LTD. LONDON

Regarding bar numbers, the standard is limited

Field 26D: Certificate Number(s) and/or Bar Number(s)
FORMAT Option D 30*65x (Narrative)
EXAMPLE :26D:BAR NUMBERS D340055, D443357, G101020, G234576

Given the SWIFT use of ":" I would advise against that as a delimiter and either go with "," or preferrably "/" which is used in field 26C as a delimiter

Field 26C: Identification of the Commodity and the Commodity
FORMAT Option C [3!a]/15x/5!a4!a[4x][//8x]

Subfield Format Name
1 [3!a] (Delivery Details)
2 /15x (Availability)
3 /5!a (Allocation)
4 4!a (Type)
5 [4x] (Denomination)
6 [//8x] (Form)

Warren James said...

Great suggestions Bron - I've updated above with iso currency code prefix. The delimiter does pose a bit of a problem - main goal is human readability but I agree a full stop confuses where the decimal place would be used ... also I confirm there are bar serial numbers which use the full stop (~10,000), slash and sometimes a comma (352).

There are no usages of colon (:) or pipe (|).

Bron Suchecki said...

Note in the Bob Pisani clip that the weight corrections are only by 0.025oz, the rounding unit for LBMA standards. This is most likely due to different calibration of scales and natural variation between weighers and most likely occurs if the actual scale weight is on the borderline of the round up/down, so if one weigher reads a few hundreds of thousands more or less then they make a 0.025oz change.

There may be some value in recording the previous weight to see if when the bar shows up in another vault its weight gets changed back by the receiving vaults weighers.

FYI from an LBMA March 2007 paper to give an idea of how weights could change merely due to reading error:

"In the London Bullion Market, gold has always been weighed on beam balances. The smallest weights used to balance the bar are of 0.025 troy ounces and the registered weight of each bar will be expressed as a multiple of this weight. The balance also has a scale with divisions which represent 0.001 of a troy ounce. In order to ensure that the customer always gets “full measure” the practice in London is that the bar must weigh at least 0.002 of a troy ounce more than the weights on the balance pan. In other words, when a close balance has been achieved, the needle must settle on a point at least 2 divisions in favour of the bar. The recorded weight is then taken as the sum of the balance weights (i.e., a multiple of 0.025 a troy ounce). This means that (ignoring any errors in the balance weights or the set-up of the balance) the true weight of a bar can be anything from 0.002 to 0.027 of a troy ounce greater than the recorded weight."

Marks said...

Warren, no need for attribution... You can use "Dead Bullion" free and clear...

Marks said...

Thanks, that is everything I ever wanted to know about 1 kilogram gold bars...