# The Origins of Efficiency **Brian Potter**  --- _Discovery is necessary but not sufficient. What matters is whether you can produce it at scale, reliably, and cheaply._ Fleming discovered penicillin, but the discovery nearly didn't matter. In 1941, the United States didn't have sufficient stock to treat a single patient. By the end of 1942, fewer than 100 patients could be treated. By September 1943, supply was sufficient for the entire Allied Armed Forces. Between 1943 and 1952, the price of a 600-milligram vial fell from $200 to $1.30. By some estimates, antibiotics have extended average human lifespan by 23 years. What changed everything wasn't the science. It was figuring out how to manufacture it efficiently. Potter's argument is that efficiency is the engine powering human civilisation. Finding ways to produce goods and services with fewer resources is the force behind some of the largest consequential changes in history. This isn't about making things cheaper. It's about making the impossible possible at scale. --- **A production process is a series of steps through which inputs are transformed into finished product.** Five factors characterise any step: the transformation method itself, the production rate, the inputs and outputs, the buffer size (how much material is waiting in process at any given time), and variability (fluctuations in speed, quality of output, breakages). Work in process is material that's been paid for but hasn't yet been sold: an investment that has yet to yield a return. There are six ways to make a process more efficient. Change the transformation method to one that requires fewer resources. Increase the production rate to exploit [[Scale]], where per-unit costs fall as volume rises. Reduce variation so the process works every time. Decrease input costs through cheaper materials, lower wages, or less waste. Reduce buffer sizes to minimise tied-up capital, an argument for [[Speed]] as an efficiency lever. Delete entire steps that don't add value. The theoretical limit is a process with no buffers, no variability, no unnecessary steps, inputs as cheap as possible, operating at maximum scale with transformation methods at the limits of what technology allows. --- **Individual technologies follow S-shaped curves.** Slow initial progress, rapid improvement in the middle, diminishing returns as physical limits approach. Production technology follows overlapping S-curves: a new method is developed, improves until it hits a ceiling, then gets replaced by one with a higher ceiling. Early on, a new technology often performs worse than the established one along the most important measures. But it's superior along some other axis that matters to a particular set of customers, giving it a foothold. Over time, it improves along all dimensions and eventually displaces the incumbent. Nails made up 0.4% of US GDP in 1810, roughly the same fraction as computers today. They got cheaper because new production methods required fewer inputs, ran more reliably, and operated at higher rates. Between 1770 and 1841, Britain's cotton cloth output increased by a factor of over 100. Efficiency improvements in textile production didn't just make cloth cheaper: they restructured the entire economy. --- **Production methods are surprisingly unportable.** A process doesn't stand alone; it's embedded in a particular context of resources, knowledge, assumptions, and relationships. Much of what's required to run a process is tacit knowledge: skill-based know-how that can't be easily written down or transferred. It might live in working relationships among staff, or in the tricks foremen and operators devise to keep things running but conceal from supervisors. This is exactly the accumulated advantage that constitutes [[Process power]]. Adopting new production technology isn't like buying a new car and getting a better ride. It's like trading the car for an aeroplane: you have to learn how the new system works and develop skills that take experience to acquire. The creation of new technological capability tells you nothing about whether that capability can be achieved cheaply. What matters is production at scale, reliably, and at cost. ---