Overall keys to success: well-planned work flow that uses staff time efficiently phased approach that produces outcomes (not necessarily final ones) at intervals (important for funding since each phase can be a fundable unit, also for institutional support) staff retention and jobs that attract talent and provide a clear path to a next job (i.e. not dead ends); jobs that reward with opportunities to learn and be creative, perhaps in lieu of monetary competitiveness technical implementation plan that takes into account the realistically available support and technical talent, over the long term: make sure that whatever is built can be maintained with the people you'll have some kind of funding...

When we look at the landscape of digital projects, we can trace some very rough characterizations that may affect how the project's strategic planning and work flow are focused Undertakings that tend to be funded through grants strategic emphasis on producing new, fundable agendas work flow has to match grant cycle and has to accommodate the fundraising effort itself tend to be conceptualized as publications or as projects: i.e. having a definite term or projected outcome Projects that tend to be funded through institutional commitment of staff time (e.g. faculty release time or as a library project) strategic emphasis on producing work that is relevant to faculty research, that draws on their expertise and can fit into their reward structure work focus emphasizes efficiency but with a long-term view: sustainable processes, institutional memory, development of a vibrant organizational culture may be projects or publications, but often are conceptualized as centers: i.e. having some long-term institutional stability Projects that start as one of the above and then conclude as publications (e.g. with the Gale Group, Alexander Street Press, or an academic publisher) strategic focus on completion work flow emphasizes efficiency but not necessarily reproducibility

At this point we can start to answer more practical questions: what kind of work flow makes sense? what expertise do we need? what level of quality is useful?

Really a question of what order to do things in to minimize entropy, energy consumption, and pollution

There are a number of basic steps that most digital projects have to go through in developing a digital resource:

data ingestion (basic capture of images, text, sound; creation of contextual information such as notes, commentary) data analysis and encoding (adding analytical value to the basic data through encoding, metadata) error correction and quality checking/fixing (color correction, proofreading, validation, image cleanup assembly (i.e. putting the data into the publication system) testing

Issue of format conversion: you want to minimize the number of times you convert, and the lossiness of conversion processes keep data in the most information-rich format for as long as possible at the same time, need to accommodate different skill levels, people who are able to read and work with different formats and some formats are better for some things than for others: e.g. for writing prose, for adding metadata, for doing encoding

Issue of error correction: you want to catch errors in as few passes as possible, and you don't want to do things that risk reintroducing error after you've done a careful error correction process.

It may be useful to identify and contrast two (of the many) possible approaches to illustrate some differences: think of these as endpoints on a continuum rather than as opposed alternatives

The first is what we might call a craft approach: Once we've gone through the document analysis process and decided what to represent... Transcribe and encode the project's content in a detailed way to begin with Do careful error checking and review Develop an interface that exploits the representation

Note that there's significant expertise and training required at each step in this process...

We can also imagine taking a very different approach where you capture comparatively little information up front: just the basic transcription

Emphasize automation, or cheap, fairly untrained labor (for instance, off-shore transcription

Note that these labor forces can still produce fairly good TEI encoding, but they can't make the kinds of informed scholarly decisions about variants and difficult readings that we were imagining above

Use simple interface tools to put the materials online so that audiences can access them, do basic word searching

Later on, more information can be added

Note that some digital repository models are starting to envision this as a layering: Digital library does the initial capture, OCR, transcription via a vendor, whatever Then provides tools that allow experts to add markup: either directly or via out-of-line markup that lives outside the digital repository Not yet a practical reality at scale, but an interesting approach for the future The phased approach defers the deployment of expertise and also (potentially) decouples scholarly expertise from technical expertise: scholars may not need to engage with the encoding

In between these endpoints, we can imagine many midpoints: depending on the nature of the material: does it lend itself to automated capture? depending on the available labor pool: do you have access to smart students who can be a cheap, well-trained labor force? depending on your training/management staff: do you have the wherewithal to train and manage an encoding staff at the necessary level of complexity, consistency, quality?