What you are registering
Not just an identifier. A registration is the DOI plus a description of what it names: journal title and ISSNs, the article's title, its authors, the volume and issue, the publication date, the page or article number, the licence, and the address the DOI should resolve to.
You also submit the article's reference list, and this is the part most worth doing properly. Deposited references are matched against everything else registered, which is what allows a citation of another paper to be followed, and what allows citations of your own articles to be found and counted. A journal that registers DOIs without references gets the permanent name and none of the citation graph.
The prefix, and what the rest of the name is up to you
A registered publisher is given a prefix, which identifies you, and chooses the rest of each DOI, which identifies the article. That second half is a decision you make once and live with.
Two rules make it painless. Make it opaque - do not encode the year, the volume, the author or the title into it, because any of those can turn out to be wrong, and a DOI cannot be changed once it is published. And make it short enough to print on the first page of the article without wrapping.
The obligation you take on
A DOI is a promise that the link keeps working. If your journal moves to a different platform, changes its address scheme, or reorganises its archive, every registered DOI has to be updated to point at the new location. That is your responsibility and nobody else's, and a journal with thousands of broken DOIs is a journal nobody will cite with confidence.
In practice this means two things. Keep the article addresses on your own site stable, and treat a change to them as a deposit update rather than a redirect you will get round to.
Where registrations go wrong
The common failures are metadata ones, and the same few recur:
- The print and electronic ISSNs swapped, or the same one given as both.
- An author's name deposited as a single string, so it cannot be matched to the same author elsewhere.
- A publication date that disagrees with the date on the article.
- A resolution address that works today and is behind a login tomorrow.
- References deposited as plain strings, which register but match nothing.
- A licence omitted, which makes the article look closed when it is open.
Registering as part of publishing
Every field above exists in the article you just approved. If the registration is assembled from that same approved article, the metadata agrees with the published page by construction, and the DOI is live when the article is.
Assembled separately, by hand or by a supplier working from a PDF, it agrees by luck - and the mismatches are the ones in the list above.
Last reviewed
