It's "just work": A lesson in managing hard projects

It's debatable who said it first, but years ago my colleagues at The New York Times taught me a phrase that I still use often when managing complex projects:

It's "just work."

The way I've chosen to interpret it, the idea is that there are two phases to any hard project. During the first phase, you have no idea whether some part of it is even possible. It's a speculative, exploratory phase tinged with both excitement and dread, where the probability of failure is difficult or impossible to know.

In product and technology, maybe you're trying something entirely new, building novel technology, or scaling something beyond where it's been scaled before. In journalism, maybe you don't know if the a source will talk to you, if you can get your hands on that critical document, or if your reporting will support the angle you've envisioned in your head.

During this phase, the risk is that you're spending time on something that just won't work out. No amount of additional manpower can help you overcome that risk. What gets you through it is persistence, ingenuity and, sometimes, luck.

Once you cross that threshold, you enter the next phase — where the remainder of the project is "just work."

That's not to diminish "the work" as easy or trivial — often it's the most important part of the project. But at this point, you're beyond the speculative. You know that what you want to do can be done.

Finishing "the work" becomes a function of time and energy. The probability of success, at this point, is more knowable. The challenge turns into one of discipline and execution.

Of course projects can also toggle back and forth between those phases, but the basic idea still holds: Sometimes you need creativity and ingenuity and luck; sometimes you need process, structure and resources.

I suppose this comes to mind because when I was coming up with a lane for Local Angle, it felt more natural to focus on the first phase rather than the second.

Partly that's because the news industry is getting pretty good at execution. Thanks to the News Product Alliance and others, we're adopting modern product development practices and recruiting some truly wonderful product professionals and engineers into news organizations. This was not broadly the case even a decade ago.

The first part, though, is still a challenge. It's more organic. It depends more on individuals and small teams and rewards generalists over specialists. It's harder to systematize. There have been some notable exceptions, but I still don't think we've found a model that enables that kind of messy but pragmatic innovation in a repeatable way across the industry.

If you believe, as I do, that innovation is not a function of resources, this is actually a hopeful thing. It means that large, well-resourced organizations don't have a monopoly on game-changing ideas. Small and mid-sized newsrooms can play, too. Maybe not every day — that's where resources come in — but certainly often enough to make a difference.

The other lesson I take from this is that the best projects have both phases. If too much of your work feels like, well, "just work," you might consider taking a flyer on some more risky ideas that scare you a little.

And if you need help with that, don't hesitate to reach out.

Outside Angles

Some things I found interesting this week:

• The shape of stories: A fascinating experiment in using embeddings to detect semantic shifts in stories. I've tried using these to estimate the "complexity" of news stories, which could be an interesting feature to feed traffic or conversion models.
• Development Speed is not a Bottleneck: "If you look at what actually matters (growth, revenue, retention), knowing the right thing to build is the most common bottleneck. And we can't reliably know what will work from the outset."
• A high-velocity style of software development: A lot of great news developers I know use many of these tricks, often unconsciously. Some great tips here on doing pragmatic development on deadline.
• Why RDF is the natural knowledge layer for AI systems: "But something remarkable happens when you add a knowledge layer between your data and your AI. When that same data is transformed into a knowledge graph, accuracy more than triples. The improvement is dramatic."