The Vicious Cost of Feature Creep
Don’t let featuritis sink your ship.
Many ideas in product management (and the startup world in general) trace their roots to the unlikely soil of U.S. military technology.
Truth be told, I know little about defense. But I do know that DARPA, the Navy, and contractors like Booz Allen Hamilton and Lockheed Martin have played pivotal roles in modern technology throughout the Western world. Correspondingly, they’ve invented and (literally) field-tested many of the planning methods often found in Silicon Valley, and in the private sector more broadly.
From the simplest Gantt chart—popularized by Allied Forces in WWI—to the most intricate PERT CPM—from which we get the phrase “fast-track”—U.S. military strategists have introduced some of the most influential tools still used in products launches *cough*, marketing campaigns *cough cough*, and talent recruitment *cough cough cough* every day. (Also the Internet, but who’s counting.)
So it’s always ironic when—all too often—the military botches a planning and development project for mission-critical materiel.
For instance, the Air Force’s failure to do — well — anything right with the new F-35 has been subject to widespread media ridicule these last few months. That being an ongoing battle, I’m not taking sides here. 😏
But 5 hours ago, the Associated Press announced that the U.S. Navy is finally ready to command its own newest and largest warship, a “stealth-guided land-attack missile destroyer” (read: big fast hidey boat) called USS Zumwalt… arguably 15 years late. The AP’s announcement includes contributions from Zumwalt captain-to-be James Kirk (yes, his name is really Captain James Kirk), alongside folks who worked on the project, like electrician John Upham. Says Upham: “We’ve overcome lots of obstacles to get to this point.”
Well yeah, you betcha. Here’s the salient point, as lifted from the AP’s announcement:
“[Zumwalt’s] growing cost forced the Navy to reduce what was originally envisioned as a 32-ship program to just 3 ships. The loss of economies of scale drove up the cost of the individual ships.”
If you’re keeping score, that’s a whopping 90% deficit in planned output, presumably because relative costs rose ≥9×. 😲 In decision theory — cousin to the more popularly known game theory (itself of military origin)—we call this 90% the cumulative regret. And yes, “regret” is the best word: in machine learning, we model it with a loss function. Appropos.
Such failures result from feature creep (a play on “mission creep”, yet another military planning term). Most people intuitively get the dangers of feature creep. Sometimes it’s a headache; other times it’s an existential threat. But the uniquely vicious cycle broached in AP’s Zumwalt announcement is less often anticipated: creep increases costs; costs asymptotically crash economies of scale; crash further increases relative cost; which reduces total productive output. That second point is the key: yes, creep per se increases cost, but the point at which it crashes economies of scale is the beginning of a vicious cycle that costs all the more for it.
So we don’t just need systems for planning our products, we also need systems for managing them. This management must be well-directed to be useful. Thus, one of the less obvious but more important heuristics might be to seek to minimize cumulative regret with respect to economies of scale. If you’re building hardware, this lesson is clear. If you’re building software, you should adapt this lesson to your pipeline (probably around generalizability, your impact-effort matrix and technical debt). In all, you’ll introduce more vigilant and sustainable options in your feature planning, and will sure-up your product strategy as a whole.
Now as they say in the Navy, fair winds and following seas. ⚓
—Enjoy this post? Click the ♡ button below and tweet me.
—Looking for work? I’m CTO at Volley.com and we’re hiring!
