The Subtraction Instinct

Why your best improvement might be a deletion

Leidy Klotz is a professor of engineering at the University of Virginia, which means he has spent his entire career learning how to design and build things. None of that training helped him one afternoon in his living room, where he was building a Lego bridge with his three year old son, Ezra. The bridge had a problem. Its two support towers were different heights, so the span between them would not sit level. Klotz did what came naturally. He turned around to grab another block to add to the shorter tower. By the time he turned back, Ezra had already solved the problem. The toddler had simply removed a block from the taller tower.

A three year old had beaten a professor of engineering to the better answer, and the professor was self aware enough to wonder why. Both moves would have leveled the bridge, but subtraction was faster, cheaper, and more elegant, and it never even occurred to him. Was his additive impulse a personal quirk, or was it something deeper about how human minds search for solutions?

The research

Klotz took the question, and the Lego bridge, to his University of Virginia colleagues, and ended up collaborating with behavioral scientist Gabrielle Adams, along with Benjamin Converse and Andrew Hales. The result was a 2021 paper in Nature with a title that gives away the ending: People systematically overlook subtractive changes. Across eight experiments, when people were asked to improve something, whether an object, an idea, or a situation, they overwhelmingly searched for things to add and rarely considered what they might remove.

The experiments were wonderfully varied. Participants edited essays, revised travel itineraries, redesigned a hole on a miniature golf course, and stabilized Lego structures. In one Lego task, participants had to secure a roof above a figurine so it could hold a masonry brick, and they could do so either by adding supports or by removing a single block. Adding pieces cost them money while removing was free. Most people paid to add anyway. Critically, the researchers found that subtractive solutions were not being considered and then rejected on the merits. They simply were not coming to mind at all. When the instructions explicitly reminded people that removing pieces was an option, or when people had a few practice rounds to explore the problem, subtraction rates jumped. And when participants were distracted or under cognitive load, the additive default grew even stronger.

That last pair of findings is the most important part of the study for leaders. The bias is not destiny. It responds to prompting. We will come back to that.

Organizations are addition machines

If individuals default to addition, organizations institutionalize it. Think about the asymmetry in how the two kinds of change are experienced at work. Addition is visible. A new initiative gets a name, a kickoff meeting, a staffing plan, a champion, and eventually a line in someone’s promotion packet. Subtraction is invisible. Nobody updates their resume with the meetings they cancelled, the process steps they eliminated, or the features they declined to build. Every proposal to add has an advocate who benefits from its existence. Every proposal to remove threatens someone who built, owns, or depends on the thing being removed. Layer on our well documented loss aversion, the tendency to feel losses roughly twice as strongly as equivalent gains, and the deck is thoroughly stacked.

There is also a signaling problem. Addition demonstrates effort in a way subtraction never can. A consultant who delivers a thick deck looks more valuable than one who delivers a single page, even when the single page is worth more. A new executive who launches three programs in the first quarter appears decisive, while one who quietly shuts three down appears to be shrinking the empire. When boards, bosses, and performance reviews all implicitly ask what you added this year, rational people add. The result is that the additive bias in each individual head gets amplified by the incentives around it, and the two reinforce each other in a loop.

I have written before, in my list of eponymous laws, that the complexity of a system increases with each new feature. Klotz and his colleagues supply the cognitive mechanism that sits underneath that organizational law. Systems do not bloat because anyone decides to bloat them. They bloat because thousands of well intentioned improvement decisions each defaulted to addition, and nothing ever pushed back. The roadmap only grows. The meeting calendar only fills. The policy manual only thickens. Left to its defaults, every organization becomes an accumulation.

The city that deleted a highway

What does it look like when someone overrides the default at scale? Consider Seoul. The Cheonggyecheon stream once ran through the heart of the city, but it was buried under concrete in the late 1950s, and by 1976 an elevated expressway ran above it, a proud symbol of Korea’s postwar industrial rise. By the early 2000s that highway carried 169,000 vehicles per day through downtown. So when mayor Lee Myung-bak, himself a former construction executive, proposed tearing the whole thing down and restoring the stream beneath it, transportation experts warned of congestion and chaos, and local business owners lined up in opposition. The logic of addition says you fix traffic by building more roads, not less.

Seoul subtracted anyway. Demolition began in 2003, and by 2005 the expressway was gone, replaced by a 3.6 mile stream and greenway corridor through the center of the city. The predicted “carmageddon” never arrived, in part because the city paired the removal with serious investment in public transit. According to a Landscape Performance Series case study, the project contributed to a 15.1% increase in bus ridership between 2003 and the end of 2008, the corridor now attracts an average of 64,000 visitors a day, and summer temperatures along the stream run 3.3 to 5.9 degrees Celsius cooler than on a parallel road a few blocks away. Mathematicians would not have been surprised. Dietrich Braess showed back in 1968 that adding capacity to a road network can actually slow everyone down, and removing it can speed things up. Seoul deleted a highway and got back a river, a park, cooler streets, and a more functional transportation system. It remains one of the most celebrated urban projects in the world, and it was, at its core, an act of subtraction.

Minus 2,000 lines

Closer to home for those of us in technology, there is a story from Apple that deserves to be told at every metrics review. In early 1982, managers on the Lisa software team decided to track engineer productivity by requiring everyone to submit a weekly form reporting how many lines of code they had written. Bill Atkinson, the creator of QuickDraw and by most accounts the most important programmer on the project, thought lines of code was a silly measure. His goal was to write programs that were as small and fast as possible. That very week, as Andy Hertzfeld recounts on Folklore.org, Atkinson had just finished rewriting QuickDraw’s region calculation engine with a simpler, more general algorithm. The rewrite made region operations almost six times faster, and as a byproduct it eliminated about 2,000 lines of code. When Atkinson reached the lines of code box on the form, he thought for a second and wrote: negative 2,000. After a couple more weeks, the managers stopped asking him to fill out the form.

I have covered Goodhart’s Law and the unintended consequences of metrics before, and this story is the perfect intersection of that theme and this one. The best engineering work done at Apple that week was not just invisible to the metric in place. It registered as negative productivity. If your measurement systems can only see addition, your organization will only ever add, and your best subtractors will learn to stop volunteering.

Subtraction as strategy

Fifteen years later, subtraction saved that same company. When Steve Jobs returned to Apple in 1997, the company was hemorrhaging money and selling a sprawling, confusing catalog of overlapping computers, printers, and gadgets. Jobs reviewed the entire product line and cut it down to a two by two matrix: a desktop and a portable machine, each in a consumer and a professional version. As he later told his biographer Walter Isaacson, in an account summarized by Entrepreneur, “Deciding what not to do is as important as deciding what to do.” In the fiscal year ending September 1997, Apple lost $1.04 billion and was, by Jobs’s own reckoning, about 90 days from insolvency. One year later it turned a $309 million profit.

I am generally wary of great-person turnaround stories, and Apple’s revival had many parents, including a badly needed cash infusion and years of operational discipline. But the shape of the opening move is important. The turnaround that eventually produced the iMac, the iPod, and the iPhone did not begin with a bold new addition. It began with mass deletion. The focus freed up Apple’s best people to work on a small number of things worth doing insanely well. Strategy, it turns out, is as much about what you refuse to do as what you commit to do.

Leading by subtraction

Remember the hopeful finding buried in the Nature paper: the additive bias responds to prompting. People find subtractive solutions when something reminds them to look. That is very good news for leaders, because prompts are exactly the kind of thing leaders control. You cannot rewire your team’s cognition, but you can build the reminder into your operating rhythm. A few ways to do that:

Open reviews with a kill list. Before any roadmap, budget, or planning discussion turns to what to start, spend the first fifteen minutes on what to stop. Make the subtractive question the literal first item on the agenda, because if it comes after the additive brainstorm, it will not come at all.

Pair every addition with a removal. For each feature shipped, retire one. For each recurring meeting created, cancel one. For each new metric on the dashboard, delete one. The one for one rule is crude, but that is the point. It converts subtraction from an occasional virtue into a standing constraint.

Make removal visible and rewarded. Celebrate the negative 2,000 lines moments in all hands meetings and performance reviews. If recognition only flows to launches, your people will rationally conclude that deletion is career poison, and the Atkinsons in your organization will keep their best work to themselves.

Schedule subtraction when the pressure is lowest. The research showed the additive default gets stronger under cognitive load, which means teams in crunch mode add bloat fastest exactly when they can least afford it. A recurring, protected subtraction review, quarterly works well, catches the accumulation before it compounds.

A challenge

Here is your assignment for this week. The next time you are asked to improve something, a plan, a document, a product, an org design, your own calendar, force yourself to generate at least one subtractive option before you allow yourself a single additive one. Then go one step further. Find one thing your organization is doing today that nobody would miss if it stopped, and stop it. Not study it, not sunset it over three quarters. Stop it. A three year old at a Lego table saw what a professor of engineering could not: sometimes the fastest way to level the bridge is to take a block away. Somewhere in your company sits your own minus 2,000 lines. Go find it.