A doll code is usually an identifier, not a secret language with one universal grammar.
That distinction matters because buyers, sellers, forums, and sometimes manufacturers themselves reuse heights, names, suffixes, and option labels in ways that look more systematic than they really are.
The safest rule is simple: preserve the exact token first and interpret it second.
A code identifies something before it explains something
If a manufacturer calls a body ZX166K, 161M, 158T, or 171D, the first useful fact is that exact string.
The tempting next step is to decode every character into a universal formula. That is where research starts turning into fan fiction.
A suffix may describe a cup class, revision, construction, series, option family, or something else entirely. Unless the manufacturer defines the meaning clearly enough to generalize, DJP Dolls keeps the token literal.
ZELEX: structured codes, but still not permission to merge things
ZELEX provides some of the cleaner examples in the current corpus.
SLE 3.0 bodies use identities such as ZX160J, ZX164G, ZX165D, ZX166K, ZX170A, ZX171C, and ZX172E.
K-Series bodies use a different family of tokens, including ZK170F, ZK168B, ZK168B-1, ZK168B-2, and ZK159D. K-Series heads use codes such as KE01_1 through KE05_1.
Fusion introduces another family. The current canonical evidence includes ZFE01_1 and ZFE01_2, both associated with the display name Gwen in preserved first-party rows.
Those two head codes remain separate identities because the current evidence does not establish that they are the same sculpt, a simple revision, or interchangeable aliases. A shared display name is not enough to collapse them.
That is the pattern to remember: code-level differences beat convenient name matching until stronger evidence says otherwise.
Irontech: suffixes are especially easy to over-interpret
Irontech body identities include examples such as 158T, 160 Minus, 161M, 162 Plus, 164T, 166 2.0, 158BA, and 164LN.
The current research does not support one authoritative global definition for T, BA, LN, M, Plus, or Minus across the entire catalog.
That means DJP Dolls does not turn Plus into a universal bust rule, Minus into a universal lightweight rule, or M into one permanent meaning just because one product context suggests a relationship.
Revision wording matters too. 166 and 166 2.0 are not silently treated as one timeless body identity when option and robotics evidence differs across dates and sources.
FIREDOLL: numbered heads are part of the identity system
The Zhongshan FIREDOLL catalog uses a different naming pattern again.
Preserved first-party examples combine body information with numbered head identities, such as a height/cup-style body token paired with a head like #87.
Those numbered FIREDOLL heads should not be confused with Top Fire Doll’s Txx head codes. The two manufacturers have similar English names and dealers sometimes blur them together, but they remain separate manufacturer identities in the DJP corpus.
A code pattern can therefore help identify not only the product but sometimes which manufacturer system you are actually looking at.
HimeKingDoll: sometimes there is no neat alphanumeric sculpt code
HimeKingDoll is useful because it breaks the assumption that every factory must publish a tidy code system.
The current first-party layer centers on named identities such as Flower Hime / 花姬 and Snow Hime / 雪姬, together with portal keys and product forms rather than a ZELEX-style public sculpt code system.
The same Snow Hime sculpt is used on a detachable 172 cm full doll and an integrated 45 cm halfbody. Those are different product assemblies sharing a head/sculpt identity, not two unrelated heads just because the finished products differ.
If a factory does not publish an alphanumeric sculpt code, DJP Dolls does not invent one to make the database prettier.
STARPERY: names and revisions can be identities too
STARPERY relies heavily on named heads rather than a single obvious public head-number sequence.
The canonical layer keeps Amy and Amy Large separate because the evidence supports them as distinct named identities.
It also keeps the 171 cm D-cup old and new bodies as separate revisions rather than flattening them into one generic 171D entry.
This is another reason a human-friendly display name is not enough by itself. Revision state can be part of the identity.
Seller character names are not automatically factory identities
Retailers often sell a complete look under a character name. That can be useful for shopping, but it creates a second naming layer on top of the manufacturer system.
A seller may rename the same body/head pairing, reuse a name across different configurations, or attach a photoshoot name that the factory never uses.
DJP Dolls therefore separates:
- manufacturer body identity
- manufacturer head or sculpt identity
- exact evidence-backed body/head pairing
- seller listing identity
- seller or photoshoot display name
Keeping those layers separate makes it possible to compare listings without pretending every store speaks the same language.
Do not build combinations from a selector just because you can
A configurator may let a buyer click dozens of heads onto dozens of bodies. That does not mean every theoretical combination deserves its own canonical model page.
DJP Dolls creates exact combination records when there is evidence for a presented or documented pairing. Compatibility evidence can be stored separately when the model supports it.
This prevents the database from exploding into thousands of fictional “models” produced by multiplying every dropdown against every other dropdown. Computers love doing that. Reality is less cooperative.
A quick identity checklist
Before comparing two listings, check these in order:
- Are they the same manufacturer?
- Is the body token exactly the same?
- Is the head or sculpt token exactly the same?
- Is the revision the same?
- Is the material or construction the same?
- Is the seller using its own character name?
- Is the claimed pairing actually documented, or merely selectable?
- Are any suffix meanings being assumed rather than sourced?
If those answers are stable, the comparison is probably meaningful. If they are not, a price difference may be comparing two different products that merely happen to look similar in a thumbnail.
Ask the seller to map its listing back to the factory identity
When a retailer uses a character name or its own model title, ask for the exact manufacturer identities instead of trying to reverse-engineer them from height, cup size, or photos.
The useful mental model
Treat manufacturer codes like part numbers on a print: preserve them exactly, do not casually normalize them, and do not infer a missing revision because the neighboring numbers look convenient.
That approach is less glamorous than decoding every suffix into a grand taxonomy, but it produces a much better buying guide and a much less ridiculous database.