How do journals improve their impact factor?

I am an Associate Editor with a journal that has improved its IF from a little over 1 to more than 4.5 over the past decade. Some of the things that we have done to achieve this have been:

• Better defining our unique niche within the field. Part of this has been placing an emphasis on publishing papers with finding that are generalisable across sub-fields, not specific only to a narrowly defined group of specialists.
• Publishing Position Papers that help set the standards within the field. These Position Papers are designed to help scientists in our field improve their procedures, practises and assessment of their own work. They quickly become standard references, so they receive a lot of citations themselves, but are also something we can point authors to for guidance if their work needs improving to meet the standards of our journal.
• When asking for revisions, working hard to make sure the feedback is as useful as possible and designed to help authors improve the impact of their papers. We try to get four reviews of each paper, usually including some comments from an editor about how to make the paper more impactful. The feedback that we have from authors is that, although our review process is longer than they'd like, they do find the extra feedback helps to make their papers better.
• Inviting contributions from scientists who we feel can help us to improve the profile of our journal.
• Changing the way we handle special issues. Like many journals, we used to let guest editors handle the review and acceptance of papers for special issues alone, with candidate papers taken from a group of invited authors after a workshop or major project. Now, guest editors must work with regular editors who know our standards and what we are looking for, and the opportunity to submit to the issue must be advertised generally, not just to workshop participants. We also make sure the special issue topics are broad enough to be of wide interest.

There are right ways and wrong ways to go about increasing a journal's IF. I think the above points are the right way. Some "wrong ways" that I have heard of include asking authors to cite more of the journal's previously published papers, and coming to agreements with related journals to encourage cross-citations between the two.

The general argument is to publish great articles. However, the challenge is that in many fields, authors will send their work to the most prestigious outlet that they think might accept the manuscript. So if you have a second tier journal, it will typically receive the work that has been rejected by higher ranking journals (or where the authors felt that the work would have been rejected by higher ranking journals, so they didn't bother submitting).

Here are a bunch of approaches for increasing impact factor from the ethical to the unethical.

Ethical approaches

• Publish quality over quantity. Of course, a journal needs a certain number of publications per year in order to receive an impact factor. But in general, higher quality publications will receive more citations. And impact factor is a ratio of citations to publications.
• Create belief in authors that impact factor will rise. Having a great editorial board or being from a prestigious publisher can create a belief in authors that the impact factor will be high in the future. I've noticed this particularly for new journals that are released by established publishers or by a group of famous academics. Authors know that the journal doesn't have an impact factor yet. But that given the profile of the editorial board or publisher, it soon will be a respected journal, and therefore, it is worth submitting work to the journal.
• Make the submission process more pleasant for authors. A pleasant review experience will encourage people to submit their work. I.e., allow different file formats, give quick initial feedback, get good reviewers, etc.
• Use editorial process to improve articles. This relies on having good reviewers.
• Make articles easier to find. Make sure that articles are well indexed by relevant scholarly search engines.
• Make articles free and open access or at least have a liberal copyright policy. You want readers to be able to find your articles. Allowing authors to post copies online or making articles open access will facilitate this. Of course, author publishing fees will discourage submission.
• Find a niche. As @significance mentions, it can be helpful to find a niche. I.e., rather than being the 4th ranked journal on a more general topic, your journal can be the best or 2nd ranked journal or a more refined topic.

Questionable approaches

• Delay the time between online publication and final paginated publication. I'm not sure if this works, but the idea is that by the time a paper is counted in the impact factor metric, it has already accrued citations.
• Think about types of papers that might attract more citations (e.g., reviews, special issues, etc.) and use this citation potential as a criteria for acceptance.

Generally unethical approaches:

• Encourage authors to cite the journal. Of course, at a basic level, citing articles from a journal is an indication that the article is on topic, but this has the potential to be abused. See this discussion of coercive citation practices.
• Encourage editors to cite articles from the journal in their other work irrespective of relevance.
• Collaborate with editors from related journals to encourage citations of each others journals

You're asking two different, but related questions:

1. "How do journals improve their impact scores?"
2. "How do journals go from obscurity to highly respected?"

Those are related, but distinct questions.

For impact scores, they can begin targeting highly cited/citable papers. For example, some medical/public health journals allow for the publication of a "cohort profile" that goes over the basic recruitment strategy, underlying demographics, etc. of a cohort of patients. This paper will be cited continually by the papers that come out of that cohort, taking advantage of "...the details of recruitment may be found elsewhere [cite]."

One can also deliberately promote citations, for example by having lots of commentaries focusing on one or more papers in that journal. The International Journal of Epidemiology did this to great effect - while they were already publishing good papers, they were now publishing good papers and an accompanying commentary from someone fairly visible in the field.

In terms of obscurity to highly respected - by publishing and attracting high quality papers. That includes having the editorial team try to get good papers in their journals, sponsoring sessions at conferences with accompanying publications, etc.