Best-in-Class Is a Dog's Dinner: The Integration Tax Nobody Talks About

The alternative to building your own technology has traditionally been to buy the best available solution for each component of your business process. The best CRM. The best cashflow modeller. The best client portal. The best compliance tool. Pick the winner in each category, plug them together. In theory, this is elegant. In practice, it produces what we once described as a dog's dinner, and the industry needs to stop eating it. The problem is not with any individual component. That client portal might genuinely be five percent better than the alternatives. But when you step back and look at the whole picture, you have a bag of bits that needs plugging in, configuring, integrating, and managing — not just to get it started, but continuously, indefinitely. The enterprise software world learned this lesson expensively through the best-of-breed versus integrated suite debates of the 1990s and 2000s. MIT Sloan found that companies with integrated technology architectures generated significantly higher revenue and profit margins than those with fragmented environments. There is a mathematical reality here that most firms miss: integration complexity grows combinatorially. Three tools means three integration points. Five tools means ten. Ten tools means forty-five. Each connection point is a potential failure point, and you are responsible for all of them. There is an instructive parallel in high-fidelity audio. Audiophiles who assemble systems from the finest individual components — turntable, preamp, speakers, each from a different manufacturer — often find the system sounds worse than a well-designed integrated system. Impedance mismatches, cable interactions, resonance issues. The components were optimised in isolation but never designed to work together. The same applies to your technology stack. W. Edwards Deming put it well: the performance of a system is not the sum of the performance of its parts. It is the product of their interactions.