ExperTech InsightsAn Organization Builds a Resilient Foundation in Earlier Stages, not Later Ones
An organization fails to build a sturdy architecture, because it overlooks mistakes in the most impactful areas: earlier stages. Errors flow downstream, compounding as they move further along. To find upstream defects, read the sections below.
Focus on the Process, not the Outcome
An organization should focus on the process rather than the outcome. Over time, methods determine results; luck decides single cases. In Thinking in Bets, Annie Duke argues that luck’s influence evens out, while the effect of a process doesn’t.
Focus on Actions, not on Words
To understand current processes, an organization should concentrate on actions rather than descriptions. Reports are misleading in a way that behavior isn’t, because they are based on human cognition. In Thinking, Fast and Slow, Kahneman details the ways people are unreliable narrators. Their stories are further altered by incentives. In The Best and the Brightest, Halberstam describes the manner in which they shape reports. An organization should trust verified actions, not descriptions.
Focus on Rules and Standards and Where Each One is Applied
To find problems with an enforcement-architecture, an organization needs to understand the governance framework that is actually applied. That scheme has significant influence on people’s actions. In Knowledge and Decisions, Sowell states that incentives and constraints are better at predicting behavior than intentions are. Yet, the applied architecture isn’t always clear. An organization only receives feedback from outcomes. The same action can be incentivized by objectives and a governance framework. When the two don’t align, confusion occurs. Multiple root causes send an inconsistent signal. A behavioral architecture is a more stable indicator.
Focus on The Coordination of Policies, Techniques, and Underlying Logic
To detect inconsistencies early, an organization must focus on the coordination of its policies, techniques, and underlying logic. The coherence of those components is difficult to assess based on damaging outcomes. The impact of incoherent pieces is highly variable. In The Principles of Product Development Flow, Reinertsen shows that incompatibility can lead to anything from inefficiency to disaster. Suboptimal results are difficult to assess, because they can be attributed something other than incoherence. Disasters force an organization to examine alternate explanations, leading to the discovery of a process’s misalignment. A clear signal is sent by a catastrophe. To receive an early warning, an organization must focus on orchestration.
Focus on the Areas Where Judgement, Data, and Algorithms Are Applied
To diagnose decision flaws, an organization must focus on the application of judgment, data, and algorithms. Formulas are the best tool for integrating information. In Noise, Kahneman, Sibony, and Sunstein show that algorithms integrate information more consistently than people do. A formula makes the same decision from identical inputs, while expertise doesn’t. Yet, algorithms do have limitations. In Red-Blooded Risk, Brown argues that a model can be directionally correct but not perfect. Even with the optimal mix of components, an organization still incurs risk due to pricing and measurement constraints. Data can be particularly tricky to optimize, because it solves many problems but not all of them. Numbers are, in certain contexts, far more reliable than judgment. In The Signal and the Noise, Silver states that data becomes less useful as the size and precision of its set decreases. With a sufficiently large and precise collection, it is superior to judgment; otherwise, it isn’t. That assessment can be difficult for organizations to make. They have suboptimal mixtures of which they are unaware. Decisions are typically the only signal for flaws, and they are a noisy one; luck factors into every outcome. The application of judgement, data, and algorithms is a clearer indicator.
Focus on Evolution and Iteration, not on Specification
Iterate; don't specify everything up front. Deploy-learn-adjust cycles teach lessons only execution reveals, faster and cheaper. The development of knowledge follows an evolutionary path rather than a designed one. In The Structure of Scientific Revolutions, Kuhn describes a process where scientists use a paradigm to guide their work, encounter limitations, and adjust it to overcome shortfalls; when the model hits a wall, the process searches for and eventually finds a superseding paradigm, which scientists use for future work. They learn lessons only obtainable through application and do so faster and cheaper. Unlike detailed specifications, which are big batches, the scientists operate in smaller chunks. In The Principles of Product Development Flow, Reinertsen states that smaller bundles accelerate feedback, revealing errors earlier and improving cycle time. Evolution is the only suitable choice to develop an organization’s processes.