liveBook · Manning

Constraint

As explained in Build High Value Software Systems

A condition or limitation that restricts how work can proceed within a delivery system.

Faster coding does not necessarily mean faster delivery. A constraint is a condition that limits what work can proceed or how quickly it can move through a delivery system. Some constraints are deliberate boundaries, such as a required deployment check. Others are capacity limits, such as a review queue or shared staging environment. Making either kind visible helps prevent work from piling up unnoticed. section 1.1.2 section 6.4.1

Ask what kind of constraint you are facing:

  • A boundary defines an acceptable outcome. A prompt to add endpoint validation still leaves decisions open unless it specifies the rules, error messages, and failure status code. Without those boundaries, AI may invent behavior that someone must review and own. section 3.3
  • A safety check prevents a known mistake. A CI gate can reject a database migration without a rollback operation; a deployment pipeline can require a runbook link. section 12.4
  • A bottleneck limits throughput. If every change must use one staging environment, opening more branches only grows the queue. Limit concurrent deployments and sequence integration tests while addressing the environment's capacity. section 6.4.1
  • An undocumented rule can constrain behavior without anyone recognizing it. Observe runtime behavior, then check tests and code before treating a surprising result as a defect. section 7.1

When the pipeline becomes the constraint

A team adds integration tests and security scans, increasing pipeline runtime from 12 to 42 minutes. Commits queue, feedback arrives late, and merge frequency falls. Cutting five minutes of coding will not fix the wait at validation. Layering or parallelizing checks can improve flow without abandoning necessary testing. section 6.5.2

To find a throughput constraint, trace one feature from request to deployment. Record waiting time, active time, queue depth, and batch size at each stage. Treat the longest sustained structural delay as a candidate, make one small change, then measure again. If the delay moves, map the new constraint rather than continuing to optimize the old one. section 6.6

Table 6.1 Practical strategies for elevating the constraint
Type
Symptoms
Strategy
Example
Code review
Aging PRs, long review wait time
Reduce demand
Cap PR size at 200 lines
Code review
Reviewers spend too much time on formatting and flagging internal standards
Remove trivial work
Automate linting, formatting
Code review
Limited reviewer bandwidth
Increase capacity
Dedicate a review block daily in the calendar
CI runtime
Long build queue
Parallelize
Split integration tests across runners
CI runtime
Flaky failures
Remove instability
Stabilize or remove flaky tests
Deployment approval
Manual gate delay
Replace gate with policy as code
Automated compliance checks
External dependency
API blocking progress
Decouple or mock
Contract testing with stubbed services

Compare symptoms with targeted responses: smaller pull requests for review waits, parallel test runners for CI queues, or contract tests when an external API blocks progress. section 6.4.1

The same reasoning applies to running systems. If services that share a database slow sharply under load while a service that does not use it remains steady, investigate the database before optimizing every service. First use its existing capacity better and reduce unnecessary demand; increase capacity if needed, then look for the next bottleneck. The practical question is not “How can everyone work faster?” but “What currently limits the flow, and what should protect or change it?” section 11.4

How Build High Value Software Systems covers it

How deep it goes
Develops a measurement-and-intervention procedure for bottlenecks and gives concrete prompt and pipeline safeguards, without prescribing one universal constraint.
Examples
Illustrates constraints with vague AI prompts, a congested staging environment, a slow CI pipeline, and a shared database.
What it assumes
Assumes familiarity with commits, pull requests, and deployments while providing a step-by-step method for tracing a feature.
What you can do after
Diagnose/debug with it by tracing delays, testing a targeted change, and measuring whether the constraint moves.
How it connects
Links constraints to Theory of Constraints, work in progress, value stream mapping, standard work, and CI/CD.

Also explained in

Compare how 29 books explain constraint →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Build High Value Software Systems. The section links point to the parts of the book it draws on.