Golden record
A golden record is the single trusted version of a product's data, assembled attribute by attribute from every system that holds a claim about it. When a supplier feed, your ERP, and a manufacturer's spec sheet disagree on the thread pitch of a 3/8-16 hex bolt, survivorship rules decide which value wins. The result is not a copy of any one source. It is the best available value for each field, with its origin recorded.
What a golden record actually is
Most product data lives in more than one place at once. A 3/8-16 x 2" Grade 8 hex bolt might exist as:
- a line in the supplier's price file, with a short description and a pack quantity
- an item master row in your ERP, with a weight, a UOM, and a landed cost
- a PDF datasheet on the manufacturer's site, with the thread pitch, proof load, and finish
- a legacy row in your webstore, last edited by a merchandiser in 2019
None of these is wrong, exactly. Each was right for the job it was built for. The golden record is what you get when you decide, field by field, which of them to believe.
Two things make it golden:
- It is assembled per attribute, not per record. You do not pick a winning source and copy it wholesale. Thread pitch can come from the datasheet while pack quantity comes from the ERP.
- Every surviving value keeps its provenance. Which system it came from, when it arrived, and which rule let it win.
Survivorship rules: how the conflict gets settled
Survivorship rules are the written policy that resolves disagreements. They are boring on purpose: applied the same way every time, and reviewable when a customer service rep asks why the site says 5.2 lb and the ERP says 5.5 lb.
Common rule types:
- Source priority. Manufacturer datasheet beats supplier feed beats ERP beats web copy, for a given class of attribute.
- Completeness. A populated value beats a blank one. Always.
- Validation. A value that passes the attribute's rules beats one that doesn't. A GTIN with a valid check digit beats an 11-digit fragment.
- Recency. Newest value wins when two sources are equally trusted.
- Human override. A reviewed and locked value beats everything until someone unlocks it.
The priority order is not global. It is set per attribute. The manufacturer is authoritative on proof load and thread class. Nobody but your ERP knows your pack quantity, your case UPC, or your stocking UOM.
| Attribute | Supplier feed | ERP item master | Manufacturer datasheet | Rule applied | Golden value |
|---|---|---|---|---|---|
| Thread size | "3/8 hex bolt" | 3/8-16 | 3/8-16 UNC | Source priority (mfr) | 3/8-16 UNC |
| Grade | Grade 8 | (blank) | Grade 8 | Completeness | Grade 8 |
| Weight, each | 5.2 lb | 5.5 lb | (blank) | Recency + validation | 5.5 lb |
| Pack quantity | 50 | 25 | n/a | Source priority (ERP) | 25 |
| GTIN | 00012345678905 | 12345678905 | (blank) | Validation (length + check digit) | 00012345678905 |
| Finish | zinc | (blank) | Zinc plated | Source priority (mfr) | Zinc plated |
What a golden record is not
The term gets stretched to cover four different jobs. Keeping them separate is what makes survivorship rules writable.
| Often confused with | The actual difference |
|---|---|
| Product matching | Deciding which records describe the same physical part. Matching happens first; survivorship only runs after you know the supplier's SKU and your ERP item are the same bolt. |
| Data cleansing | Fixing a value inside one source: trimming whitespace, normalizing 3/8" to 0.375 in. Cleansing improves a candidate. Survivorship chooses between candidates. |
| Enrichment | Creating a value that none of your sources has. Survivorship can only pick among what already exists. If every source is blank on proof load, no rule will produce one. |
| MDM hub or PIM | The system that stores and serves the golden record. Storage is not the same as the decision logic that fills it. |
The enrichment row is the one that bites distributors. Survivorship is a tie-breaker, and most catalog gaps are not ties. They are silence.
Where golden records break in real catalogs
The failure modes are predictable:
- Bad matching upstream. Two similar MPNs get merged, and now one golden record blends the specs of a 600V connector and a 1000V one. Survivorship worked perfectly on the wrong pair.
- Per-source priority instead of per-attribute. "The ERP is authoritative" is easy to configure and wrong the moment you need a thread class or a certification.
- All sources blank. Roughly the entire long tail. The rules run, find nothing, and you ship a PDP with six populated attributes.
- No provenance. A value is in the golden record and nobody can say why. The next disagreement gets settled by whoever escalates loudest.
- Rules that live in someone's head. Or in a spreadsheet that one analyst maintains. This is a governance problem wearing a data problem's clothes.
Who builds it versus who stores it
A PIM is where the surviving values live — Akeneo, Salsify, Syndigo, inriver, Pimberly. It models the attributes and holds the reviewed value.
What they do not do is the labor underneath. Reading the manufacturer's datasheet to find the proof load that no feed carries. Deciding that this supplier is reliable on dimensions and unreliable on plating. Producing a defensible value for the attributes where every source is empty and survivorship has nothing to arbitrate.
Anglera does that work and writes the result back into your PIM with its source attached. Survivorship rules settle the arguments; enrichment ends the silences.
Frequently asked questions
What is a golden record in product data?
A golden record is the single trusted version of a product's attributes, built by choosing the best value for each field across every system that describes it. Supplier feeds, the ERP item master, manufacturer datasheets, and legacy web copy all get a vote. Survivorship rules decide which value survives per attribute, and the record keeps a note of where each surviving value came from.
What are survivorship rules?
They are the written policy for resolving conflicts between sources. Typical rules include source priority (the manufacturer wins on thread class, the ERP wins on pack quantity), completeness (a populated value beats a blank), validation (a GTIN with a valid check digit beats an 11-digit fragment), recency, and human override. The order is set per attribute, not once for the whole record.
Does a PIM create the golden record, or just store it?
A PIM stores it, models the attributes, and enforces locks on reviewed values. Configurable source-priority logic varies by system. What no PIM does is the underlying labor: reading the datasheet, judging which supplier is trustworthy on which field, and producing values for attributes where every source is empty. Storage and decision logic are different jobs.
How is a golden record different from product matching?
Matching decides which records describe the same physical part. Survivorship decides which value wins once you know they do. Matching comes first and matters more, because bad matching produces a confidently wrong golden record: blend a 600V connector with a 1000V one and every survivorship rule downstream runs cleanly on garbage.
What happens when all my sources are blank on an attribute?
Nothing. Survivorship is a tie-breaker among existing values, so it has no output when every candidate is empty. This is the normal case in the long tail, where a supplier line gives you a 60-character description and nothing else. Filling those gaps requires enrichment, which means sourcing the value from documentation rather than picking between systems.