Digital projects, like all software endeavors, are easily derailed. Developing a site or application is initially seductive — the discovery phase presents a green field where all frustrations about your existing or missing capabilities can be magically erased by the New Thing. The early vision is grand — the stakeholders are picturing the end result not against a platform or service they have seen, but against a perfect unicorn. Spirits are high; people are engaged.

Requirements are the painful beginning of a process of understanding what’s possible. There’s what’s technically possible, and what’s possible given business owners’ goals, budget, and realistic maintenance capability. Tough compromises are made — in a best case scenario, rapid prototyping can improve the result. Content strategy may or may not come up, and let’s hope it does. It’s a discipline helpful for curtailing impassioned pleas for six-minute welcome videos and for preventing people without the bandwidth to update a Twitter feed from signing on for weekly 500-word blog posts.

Then a full design phase kicks in, and stakeholder engage fully in imagery, color palettes, and line leading. Hopes are once again high, and PhotoShop goes a long way to erase the sting of features lost in the requirements phase. The joy of the Bright and Shiny Object is in full effect.

During the build, compromises are made; inevitably, some degree of requirements shifts. The technology supports the main use cases, but developers managing cross-platform delivery may have to make hard decisions about the fringe. Even in an eight-week sprint influenced by Agile, stakeholders are exhausted.

Enter the final 10%. The final 10% is what separates a just-OK user experience from a terrific one. It’s closely related to the effort Ben Lerer pointed to in the NYT yesterday. The final 10% means making sure you’ve taken care of the tedious details that ensure your project has meaningful search results; delivers analytics to inform future iterations (and not just fill inboxes); plays well with social media; and that content syndicates neatly where it’s supposed to.

The final 10% isn’t sexy — it’s stuff like delivering small fixes to the administrative interface that will cumulatively make the difference between adoption and rejection, or checking that the adaptive design is breaking just right in the 84,563 flavors of Android. The final 10% isn’t capital-V Vision like the discovery phase or beauty like the design phase, but it’s a big predictor of digital project success.