System Framing, Needs and Requirements · Lesson 2

From Need to Verifiable Requirement

5 min read

Lesson objectives

  • Turn a need into a verifiable requirement with threshold and context.
Reading tools

A stakeholder need describes a problem or benefit from a person's perspective. A requirement specifies a property the system should exhibit under conditions that can be checked. In the pump maintenance service, a resident needs assurance that a report has not disappeared. The call centre wants a case number as soon as a report is registered. Neither statement automatically means use a text message. Messaging is one possible implementation and may fail where connectivity is weak. Ask what evidence is actually needed: could a visible case number suffice, and what happens when a channel is unavailable? NASA's technical requirements guidance calls for statements with enough information to permit later verification. We use that engineering principle for the fictional service without treating NASA's agency procedure as a rule for this municipality.

A practical structure states an operating condition, observable action, beneficiary, threshold with units, and evidence method. An initial case requirement might be: when a report containing contact information and a fault description is saved, the service issues a unique case identifier and displays it within five seconds for at least 95 of 100 attempts under a test load of 50 concurrent requests. The figures are assumed design targets for training, not general industry standards. We can run a controlled load, measure screen display time and check identifier uniqueness. Immediately is not a failure threshold. Nor does measuring an API response prove a promise about what the user sees. The measurement boundary must match the beneficiary's experience.

Separate the function from a chosen implementation. The requirement to retain an incomplete report and allow its location to be completed leaves room for architecture decisions. A demand to use database product X is a design constraint and needs its own justification if imposed. Keep stakeholders distinct as well: display time matters to the reporter while location validity before dispatch matters to the crew. Link each requirement to its originating need, owner and threshold rationale. When the threshold lacks evidence, label it as an initial value for stakeholder testing rather than presenting it as established fact. A usability trial may change its value. The measure and test method should remain explicit, so later reviewers can see exactly what changed and whether the requirement has been met.

Quality requirements need operational meaning. Reliable service is vague until we name the function, observation period and measure. One possible training target is: during a two-hour test window under 50 concurrent requests, every accepted report remains in the case store with its identifier and no identifier is duplicated. A separate recovery test is needed after connectivity loss; a load test by itself cannot establish recovery. We should not call this a production availability target before setting service hours and outage measurement. Prioritize requirements by consequence of failure. Losing a report or sending a crew to the wrong place calls for stronger evidence than a button colour. A review also checks feasibility and conflicting thresholds: a shorter response target might require costlier connectivity or a different storage design.

Lesson summary

Turn a speed request into a testable threshold and distinguish acceptable location format from physical location accuracy.