Empiricism and Scrum values · Lesson 1

Complex work and empirical control

9 min read

Lesson objectives

  • The learner distinguishes an assumption-led decision from one grounded in inspectable evidence and designs a small next step that exposes the next risk.
Reading tools

Complex work and empirical control

Imagine a team building a clinic appointment service. Marketing believes customers want text reminders; reception staff believe changing an appointment is the real problem. Either view may be useful, but a person's position does not make a claim about customer behavior true. In complex work, user needs, operating practices and technical constraints interact, so the team may not know in advance which design will work. Scrum helps the team work in short cycles, produce a usable Increment in each Sprint, inspect the result and adapt its plan. A cycle does not remove planning. It makes the plan responsive to what the team learns.

The Scrum Guide calls this foundation empiricism: knowledge comes from experience and decisions are based on observation. It differs from following a detailed plan while postponing learning until the end. Before a Sprint, the team might state a hypothesis: ‘If patients can change appointments themselves, calls about rescheduling will fall.’ A plausible hypothesis is not an established fact. The team can select a small Increment that lets it examine the claim, perhaps offering self-service changes for a limited class of appointments and recording successful and failed attempts. The evidence includes what worked, what broke and what users report; it is more than a count of completed tasks.

Empiricism in Scrum rests on transparency, inspection and adaptation. Transparency means that the state of work and the Increment are clear to those making decisions. If the team reports the change feature as Done when it works only in a test environment, stakeholders inspect a misleading picture. Inspection compares the product and progress toward goals with observable evidence at a useful time. Adaptation changes the work or plan when that inspection reveals a consequential gap. A meeting that displays a failure rate but continues the same plan without discussion has not completed a learning loop. Later lessons examine the Scrum events that provide regular inspection opportunities; here the focus is the reasoning that connects them.

Not every opinion deserves an expensive experiment. Ask which decision depends on the assumption, what smallest useful Increment could reveal something about it, and what signal would actually change the decision. For appointment changes, the team might examine the share of people who start and complete the change, calls about rescheduling, and reasons for failure. Those signals need context: calls may fall simply because total appointments fell. The team therefore compares rates with known denominators and listens to some users. Scrum does not promise that one metric proves causation. It makes assumptions and results available for discussion and adjustment.

The Product Owner is accountable for maximizing product value and ordering the Product Backlog; Developers determine how to create the Increment during a Sprint; the Scrum Master helps everyone understand the framework and empirical work. These are complementary accountabilities, not a chain of approvals for an experiment. If evidence prompts a stakeholder to request a new direction, the team should neither ignore the evidence until an annual plan expires nor let that stakeholder dictate a technical solution. The Product Owner discusses value and ordering, the team makes the effect on the Sprint Goal visible, and its plan is adapted within Scrum's boundaries. Understanding the accountabilities keeps adaptation from becoming disorder.

When answering a situational question, separate an observed fact, a possible explanation and a proposed action. ‘Eighteen of fifty attempts failed’ is an observation. ‘The interface is confusing’ is an explanation that needs examination. ‘Add longer instructions’ is one possible solution, not a conclusion forced by the number. A stronger response makes the condition visible, tests a relevant explanation and allows a change that respects the Product Goal and Increment quality. The cause might lie in technology, appointment policy or user understanding. Do not choose one before gathering suitable evidence. This is a way to reason with Scrum, not a fixed numerical recipe for every product.

Open a small decision record before testing a service policy. State what you need to learn and what evidence could change your choice. In lending, reservation rules may not explain why some readers abandon renewal. Observe one step after making its explanation clear while holding the lending conditions stable. Changing the message, policy and timing together makes the resulting difference harder to interpret. Select a change suited to the question, define the observation period, and identify the sample. This is a practical learning technique, not a scientific experiment guaranteeing causation. Inspect effects on staff as well: fewer reader calls may accompany more manual processing. Record both so the Product Owner sees benefit and cost together. Return to the original uncertainty after the trial. Did it resolve the decision, or identify another useful test? An inconclusive result remains useful when its limits and next action are clear.

Lesson summary

The learner distinguishes an assumption-led decision from one grounded in inspectable evidence and designs a small next step that exposes the next risk.