Glossary
Tap menu for chapters & tools
Shared language
Glossary for leads
Short definitions you can use in design reviews, 1:1s, and hiring debriefs. Linked back to the chapters that use them.
ADR Architecture Decision Record
A short write-up of a significant technical choice: context, options, decision, and consequences. Used so teams stop re-litigating the same debate from memory.
DACI Driver, Approver, Contributors, Informed
A decision-role model. The Driver pushes the decision forward, the Approver has final say, Contributors give input, Informed people get the outcome. Prevents “everyone decides” fog.
RACI Responsible, Accountable, Consulted, Informed
A classic ownership matrix for work items. Similar spirit to DACI, often used for process and project roles rather than a single decision.
DORA metrics Delivery performance signals
Four measures used in software delivery research: deployment frequency, lead time for changes, change fail rate, and time to restore service. Useful for diagnosing system health, easy to game if turned into naive targets.
WIP Work in progress
How many items are started but not finished. High WIP usually means thrash and slow delivery. Lowering WIP often finishes more than adding people.
Architecture fitness function Automated check of architecture intent
A test or monitor that protects a design rule (dependency direction, schema compatibility, latency budget, module boundaries). Architecture that is not checked drifts.
SLO / SLI Service level objective / indicator
An SLI is a measured signal (availability, latency). An SLO is the target you commit to for that signal. Leads use them to trade features against reliability with eyes open.
STAR Situation, Task, Action, Result
A structure for behavioral interview answers. Keep Action about what you did, and Result concrete when you can.
PIP Performance improvement plan
A time-bound, written plan when performance is below the bar. Should follow prior feedback, include clear outcomes and support, and end with a real decision. Partner with HR.
Two-way door decision Reversible choice
A decision you can undo cheaply. Move fast. Contrast with one-way doors (hard to reverse): public APIs, major migrations, hires at the wrong level.
Psychological safety Truth can travel
The shared belief that people can raise risks, admit mistakes, and challenge plans without punishment. Not the same as “being nice.”
Strangler pattern Replace legacy gradually
Grow a new system around an old one and route traffic piece by piece, instead of a big-bang rewrite. Favored when risk of full replacement is high.
Blast radius How far failure spreads
The scope of users, systems, or teams hurt if something fails. Leads use it when ranking incidents, debt, and rollout plans.
Idempotency Same request, same effect
A property of operations where retries do not double-apply side effects. Critical for payments, webhooks, and flaky networks.
Conway’s law Org shapes system
Teams design systems that mirror their communication structure. If you want different architecture, you may need different team interfaces.
Dual track IC and manager paths
Parallel career paths for senior technical leadership and people leadership, so growth is not “manage or stagnate.”
Stream-aligned team Team on a flow of change
A team aligned to a continuous stream of work for a user segment or business domain. Aims for end-to-end ownership of outcomes to reduce handoffs.
Platform team Internal product for other teams
A team that provides self-service capabilities and paved roads so stream-aligned teams can deliver with less cognitive load. Best run as a product with users and SLOs, not only a ticket queue.
Enabling team Temporary capability builders
A team that helps others learn missing skills or adopt practices, then steps away. Without an exit criteria, enabling quietly becomes permanent ownership.
Complicated-subsystem team Specialist ownership
A team that owns a complex specialty (for example a billing core or ML runtime) so stream teams can consume it through a clear interface without absorbing the full complexity.
X-as-a-Service Interaction via clear service
An interaction mode where one team provides a service with documentation, SLOs, and support, so others consume without continuous collaboration.
Cognitive load (teams) How much a team must hold in mind
The amount of domain, tooling, and process knowledge a team must maintain to deliver. Topology design aims to keep load within what a team can handle.
Capacity planning Match demand to real supply
Estimating planned work against available engineer-weeks after interrupts, PTO, and ramp. Used to sequence work and justify headcount with scenarios.
Focus factor Share of time for planned work
The fraction of calendar time that can go to planned product and tech work after meetings, on-call, and drag. Often modeled around 0.6-0.75 for healthy teams.
Stability reserve Unplanned capacity buffer
Capacity deliberately not committed to roadmap so incidents, support, and small fixes do not force heroics or silent scope cuts.
OKR Objectives and Key Results
A cycle-based goal system: qualitative Objectives paired with measurable Key Results. Best used for a few team outcomes per quarter, not as a task list or a full KPI dashboard.
KPI Key performance indicator
An ongoing health or performance signal for a system (product, delivery, reliability). Watched continuously with thresholds; not 'completed' like a project. Distinct from OKRs, which are time-boxed ambitions.
North star metric Primary value outcome
The single outcome that best captures the value your product creates for users. Often paired with input metrics that lead it. Changes rarely compared to quarterly OKRs.
Two-pizza team Size for low coordination cost
A sizing heuristic (popularized at Amazon): keep a team small enough that two pizzas could feed it, so communication paths and shared context stay manageable. A signal to split or modularize when coordination brokers appear.
Counter-metric Guardrail against gaming
A paired measure that catches bad behavior from optimizing a single KPI or KR (for example, deploy frequency paired with change fail rate).
End-to-end ownership Own outcome path, not just tasks
A person or small pair owns understanding, plan, build, validation, and communication for an outcome, with agreed checkpoints. Opposite of task-only assignment with leader-owned how.
Micromanagement Owning the how after assigning the work
Steering daily details, requiring approval for small choices, and measuring activity over outcomes. Different from setting quality bars, priorities, and risk-based reviews.
Handover contract Written interface for delegated work
A short agreement covering context, outcome, non-goals, constraints, interfaces, checkpoints, support, and escalation so ownership is clear without constant steering.
Feedback loop Signal → compare → change → remember
A closed cycle where a sensor detects reality, the team compares it to intent, someone decides and acts, and the learning is stored. Open loops collect signal without action.
Loop latency Time from action to learning
How long it takes to learn whether a change helped users, delivery, or reliability. High latency forces bigger bets and slower correction.
Closed loop Signal that produces change
A feedback process with a named owner and a visible action when the signal is off target. Opposite of dashboards, surveys, or retros that never alter priorities or behavior.
Handover checklist Pre-flight before stepping back
A yes/no list confirming outcome, non-goals, owner, RACI matrix (one Accountable per activity), teach-back, decision rights, checkpoints, interfaces, support boundaries, escalation, and done metrics before a lead reduces involvement.
Handover metrics Evidence that ownership is real
A short set of measures such as time-to-first-progress, checkpoint hit rate, lead interrupt rate, blocker age, re-decision rate, and quality counter-metrics used to see if delegated work is truly owned.
Servant leadership Lead by enabling others
A leadership stance that prioritizes the growth, clarity, and effectiveness of the team. In engineering, it means removing blockers, sharing context, coaching ownership, and holding standards without micromanaging or martyring yourself.
Hero lead Lead as single point of delivery
A pattern where the lead takes the hardest work and decisions so the team stays dependent. Fast in a crisis, fragile as a system. Opposite of end-to-end team ownership.