A few years ago, I found slides to a presentation by Eric Ries, founder of IMVU and an early practitioner of the Lean Startup movement. Ries refers to the OODA loop–a combat operations model developed during the Korean War by then USAF pilot Col John Boyd. This post captures all of the relevant links and videos
In this post, I will explore Continuous Deployment and Competitiveness using Lanchester Strategy, a warfare model developed during the First World War by the legendary British inventor Frederick William Lanchester.
To begin, let’s briefly look at Lanchester Strategy as it applies to the simplest form of combat (aka single combat). If two evenly matched combatants square off, they each have an equal chance of winning (50%). Two competing salespeople pitching evenly matched products to a customer also each have an equal chance of winning that battle. This also applies to two competing banner ads, etc.
Whoever possesses superior skills, weapons or training will have an edge over their opponent. If we think of the product being sold as the weapon, then the salesperson who possesses a superior product will also have an edge. Let’s see how this plays out.
A product on January 1 of year 0 undergoes a certain amount of improvement over the course of a year. Improvement includes both new features as well as changes to meet evolving customer expectations and behaviors. An example of evolving customer behaviors is the expectation that objects on a computer screen can be manipulated using hand gestures, or some other expectation that arises from their use of another product or service.
So let’s start on January 1 year 0 with two evenly matched products; A and B. Product A will follow a traditional 6 month release cycle. Product B will use Continuous Deployment. Let’s assume that, the development teams are equally capable; each team can make their respective product 100% better at the end of 1 year. Six months in, team A releases their first update, making their product 50% better. Team B, meanwhile, releases an update every week, each of which improves the product by 1.92%. At the end of 4 weeks, product B is 7.68% better than product A, an advantage that expands to 15.36% at the end of 8 weeks, 23.04% after 12, and 30.72% better after 16 weeks. Now, let’s assume that better directly translates into sales–for every 1% difference in quality, there is a 1% difference in sales. Thus, after 1 week, product B will sell 101.92 copies for every 100 sold of product A.
Now, let’s see how this all plays out over a six months. We’ll assume that both products sell for $1, and both companies start out selling 100 units. At the end of six months–right when product A releases it’s next version–product A will have made $2,600 and product B will have made an astonishing $3,224, a difference of $624 or 24%! Of course, there are numerous simplifications and caveats; most importantly that customers have to internalize the difference in value in order for their propensity to purchase one over the other.
Since we made the sale price 1$, the total sales amount also equals total units sold, which equals the number of customers acquired–a 24% difference in favor of product B!
From a cash flow perspective, continuous deployment has the same benefits as just in time manufacturing–income and manufacturing expenses are more in harmony. Let’s see how that plays out in our model. We’ll assume that both companies have equal bi-weekly payroll of $75. Over the course of 6 months, that amounts to $975 in payroll–not one penny of which company A begins to recoup until it releases its next version. Company B, meanwhile, started recouping its development expenses much sooner. This monetary advantage rapidly accumulates, and, generally speaking, can not be made up.
But there is more to Lanchester Strategy that applies here. Lanchester Strategy is actually concerned with numerical force superiority; the number of troops that need to be brought to bear against an enemy of a given size. Lanchester derived mathematical formulae for both single and stochastic combat conditions, and concluded that, to ensure victory in single combat required an approximately 3:1 force ratio, while stochastic combat required an approximately square root of 3:1 (1.7:1) force ratio. In the business world, market share is the determinant. If we assume two companies both spend their dollars equally well, then the company spending more dollars will gain greater market share, and, as Lanchester Strategy implies small numerical advantages parlay into large advantages later on, all else being equal, company B’s advantage will accrue, creating an increasing imbalance with respect to company A; the rich get richer, faster, and company A falls even further behind than our model indicates.
Let’s wrap up by briefly mentioning quality, which Joseph Juran defined as “fitness for use”. Continuous Deployment helps us greatly increase quality as follows. Product development is a continuous effort to improve a product’s fitness for use by its ever-evolving customers. The best warfare equivalent I could think of is blitzkrieg. Blitzkrieg begins with probing infantry operations along a line of attack. These are, by design, low-cost experiments. Forces are rapidly redeployed to create an overwhelming advantage wherever a weakness is found. If we think of continuous deployment as a series of probing operations, it gives us the means to identify–not enemy weaknesses–but customer hot buttons. A/B testing and other techniques help us rapidly refine our offering to maximize customer receptivity and satisfaction. Even if it turns out that the solution will require significant investment, we’ve cost-effectively identified its salient features while at the same time improving our chances of satisficing the customer with whatever we implemented in the short run. Since fitness for use is a key factor in a customer’s decision to stay or defect, continuous deployment provides a mechanism for reducing churn. Reducing churn is another mechanism for accumulating competitive advantage, just like improving cash-flow and product effectiveness as described above.