At the office we like to say “20:80”, which is a shorthand for, “Do the twenty percent effort version that gets eighty percent of the result.” I don’t like it, but given that time is one of the most valuable resource in a start-up, I try to run with it. I want to do it better, and as is typical of me, I went in search of a framework.

Last week during games night, I raised this question to some team members,

“This 20:80 thing … doesn’t really fly with me. There must be a class of tasks/problems that deserve the 100% effort treatment. How do we know we are 20:80ing the right things?”

This conversation continued on and off in the week, during which a  couple candidate criteria emerged.

One was the strategy test, which is evaluating whether a particular task is core to the company’s strategy, or just part of the scaffolding that’s required to get there. This is good in theory, but it also seemed vague. As the company gets bigger, it becomes increasingly difficult for each member of the team to hold their entire company strategy in their head, let along trace each task back to the strategy.

Another was the competitive advantage test. Is the task crucial to differentiating the product from competitors. This is similar to the strategy test, and suffers from the same unwieldiness in application. If, say, the company’s core competitive advantage is machine learning, is it okay to 20:80 the entire design experience? I would argue no.

It is not clear that understanding just the product vision is enough to help us distinguish “20:80” tasks from “100%” tasks. In conversation this week, a teammate raised an alternative.

“If you chart the effort-to-impact graph, is it convex, or concave?”

This took me back to economics class. Essentially the convex curve test asks, “Is an extra unit of effort/polish going to yield an exponential return, or a diminishing return?” A task with an exponential return has a convex shape when plotted on the effort-to-impact graph.

Intuitively this seemed like a good approach. While it isn’t always clear that a task has a convex return, there are a certain class of tasks that seem to be clearly concave. In light of this framework, concave tasks are the ones that should be 20:80’d, and I ought to be more disciplined about time-boxing concave tasks.

It’s still not clear how to identify convex tasks though, let alone quantify and compare between convex tasks. Ideas?