Imprint is usually understood as the "branding" a publisher applies to a book – what's on the spine of a print book – but it's a marketing tool and may change as the publisher's needs change.
Publisher is the name of the company that owns or holds rights to a product. They receive the money from sales of the book and pay royalties to the copyright holder. Another way to understand it might be that it's the name of that appears on the Title Page of the book, and on the reverse of that page you would expect to find copyright and Cataloging In Publication information.
Supplier is the company who supplies the book to Retailers (not consumers) – so the name of the company who has been assigned by the publisher a market area (usually on a exclusive basis) and they supply retailers within that market. While that's makes the most sense in a print world supporting warehouses, EDI and distribution, it's no less true for digital books. The difference in the digital world is that it's common that the publisher = supplier and the market area is the "WORLD".
The very short form is: The Imprint is just a name but the money goes from the retailer through the Supplier to the Publisher who pays the author. Naming is important because it's branding of the product and the company, but it's also about accounting systems and identification of businesses. Consistency in the naming of these helps the extended supply chain of retailers, libraries and wholesalers understand and work with you.
Some on-line retailers, Amazon most notably, display "Imprint" as their "Publisher" value. Other on-line retailers may choose to display both Imprint and Publisher as distinct values and some may only display Publisher. All on-line retailers index both values for consumer search whether or not they are displayed.
Self-published authors may be providing the same information in each case. That's normal and fine, but you should be aware that within the metadata the functions are treated differently. You may be the business and you might even change your publisher name on every book, but within the publishing supply chain the publisher is the business name, the imprint is the brand and they need a name to contact for stock and to pay. They are distinct and changes made to them should be handled with notification and care.
Imprint codes are a means for end users to map and follow the continuity of a publisher's imprints which, as they are means of branding, may change over time. For example a publisher's 2010 imprint "Baker's Dozen" may in 2013 be presented in their list as "A Baker's 13". The continuity of the code would allow the end user to know that their records on both names should remain linked even though the formal branding has been updated. The code's consistency allows the end user to know the publisher has done this intentionally and not randomly set up a new Imprint and transferred old titles into it.
- Imprint codes are a "proprietorial" identifier and like any proprietary identifier relies on the metadata sender using a consistent self-chosen code. End users can only believe what you send.
- Amazon requests that publishers supply a proprietorial imprint code in their metadata. Other companies may use them as well.
What makes a good Proprietary Imprint Code?
It should be a code, meaning a continuous string without spaces or atypical characters in it. And it should be used consistently. After that it's up to you, but generally a string of 4 or 5 letters and numbers is probably as much as you need. If you know your Baker & Taylor imprint code you can use that. The above example might be BakDoz or BAKER, or BAK05.
It's a very poor practice to repeat the Imprint Name, spaces and all, as a code.
The need for accuracy
Publisher and Imprint names are often misspelled or have subtle differences introduced into them. Simply adding a period -- Ltd vs Ltd. -- will make a difference for computer selects. That in turn can cause your business partners to lose their associations to your records. Yes they will find them eventually, but why make it hard?
BookNet Canada strongly recommends that all publishers maintain a strictly controlled presentation of their names in metadata because they are key to how accounts are organized (and how you get paid). While simply doing it the same way each time is best, use of supplementary identifiers can help. In the case of Imprints which are held entirely within a business a well maintained proprietary code is sufficient.
The case for other identifiers
The same case made can be made for codes both Publisher and Supplier Names but the difference would be that an outside agency should supply the identifier because to be useful businesses need to to understand and trade it among trading partners. So Publishers or Suppliers would be best supported by the use of standard identifiers such as GLN (assigned by GS1), SAN (required for EDI in North America). Strictly speaking GLN and SAN are business "location" identifiers giving the business name plus the specific location of say head office vs a distribution warehouse. The most generic naming identifier is an International Standard Name Identifier (ISNI) which is mainly with associated with identifying individuals as content creators but can be used by any named organization associated with "content". Typically a publisher can be considered to be associated with content and could be assigned as ISNI, but a Supplier would not be as their association with specific content is tenuous. A Supplier's best identifier is their SAN or GLN.
Identifiers should NOT be seen as a replacement for spelling and presentation consistency, but as a means of providing "persistent" identifiers that enrich your ONIX feed as a discoverability and business tool.