What Is Scalability?

With Atif Ansar, Alexander Budzier, and Daniel Lunn

The Oxford English Dictionary defines “scalable” and “scalability” in the following manner:

“scalable, adj. Able to be changed in scale. rare. 1977 Jrnl. Royal Soc. Arts 125 770/1. Such lasers are scalable since large volumes could be pumped uniformly.”

“scalaˈbility n. the property of being scalable.1978 Sci. Amer. Nov. 44/2. It took demonstrations of the scalability of the technology and tests of improved beam focusing … to catalyze an effort that led to support by the AEC.”

We note, in particular, the unsatisfactory definition of the word scalable. What does the ability to be changed (or equally change without external force) in scale mean? In practice, scalability is most commonly discussed in the field of computer science, system architecture, and software programming (Hill, 1990; Gunther, 2007). This is insufficient for our purpose, which is understanding megaproject management in general.

In academic literature, an understanding of whether, and under what conditions, something has the ability to be changed in scale has been deeply informed by advances in mathematics and particularly the field of fractal geometry. Mandelbrot’s research identified two primary dimensions of scale: temporal and spatial. A grain of sand, a pebble, a rock, a cliff, and the coastline of say Britain represent a continuum of finer to coarser grades of the spatial scale (see Mandelbrot, 1967).

We define scalability as the ability of a thing to effortlessly transition back and forth from the very micro to the very macro spatial, temporal, and relational scales.

Similarly, the price movements of a stock price over a scale between five seconds to five years represents finer to coarser grades of a temporal scale. Finer and coarser grades of scale can also be thought of in terms of degrees of zoom-in (microscopic) and zoom-out (macroscopic).

To Mandelbrot’s spatial and temporal scale, we introduce a third scale: “relational.” The relational scale not only refers to the number of end-users (e.g., few or many) but also to the heterogeneity of end-users and their tailored need (see Ansar, 2010; Ansar, 2012 for a more extensive discussion on a spatial, temporal, and relational multi-scalar framework).

Based on this understanding of scale, we define scalability as the ability of a thing (e.g., a system, system of systems, process, or network) to effortlessly transition back and forth from the very micro to the very macro spatial, temporal, and relational scales. Effortlessness connotes minimum friction in terms of the time, cost, etc. it takes to build up or remove capacity. Over longer time scales, effortlessness implies ability to quickly upgrade without losing compatibility. However, if scalability is understood and practiced as mainly the ability to scale up, with scant attention to scaling down — which is the dominant approach today in both the academy and practice (Sutton and Rao 2014) — then this in and of itself adds fragility. Here we therefore understand scalability as the ability to change in both directions, i.e., both up and down.

If scalability is understood and practiced as mainly the ability to scale up … then this in and of itself adds fragility

In contrasting big and scale we arrive at a key insight. Big typically possesses a degree of slack, which Weinstock and Goodenough (2006), call the “ability to handle increased workload (without adding resources to a system).” A big power plant, for example, rarely operates at full capacity. If demand were to increase from one segment of the day to another the power plant can be ramped up to meet some or all of the added demand. However, this limited slack is different from true scalability. Thus, although the big power plant might be able to meet some incremental demand in a spatially, temporally, and relationally narrow field, it cannot effortlessly be scaled up (or scaled back down) to resolve a national or global energy crisis. Linking big, scalability, and fragility we therefore advance the following proposition:

Fragility arises when big is forced into doing what was best left to the scalable.

— — —

The above is an excerpt from Atif Ansar, Bent Flyvbjerg, Alexander Budzier, and Daniel Lunn, 2017, “Big Is Fragile: An Attempt at Theorizing Scale,” in Bent Flyvbjerg, ed., The Oxford Handbook of Megaproject Management (Oxford: Oxford University Press), pp. 60–95. Free pdf with the whole article, with full references, here: https://bit.ly/3zSRx76