79 Terms, Plain English
Software Development Glossary
Written for the person paying, not the person coding.
Every definition says what the term means and why it costs or saves money, because that is the part a founder or CTO is actually trying to work out.
Quick Answer
Updated August 22, 2026
What do software development terms like MVP, technical debt, and staff augmentation actually mean?
An MVP is the smallest version of a product that lets real users do the one thing that matters, so you can learn whether they want it — typically $15,000–$40,000 and 8–16 weeks. Technical debt is work deferred to ship faster, which accrues interest as every later change gets slower; the symptom is estimates growing for similar-sized work. Staff augmentation places external engineers under your management, so you keep accountability for delivery, while a managed team takes a defined scope and is accountable for it. This glossary defines 79 such terms with the commercial consequence attached to each.
Terms defined
79
Categories
8
Written for
Buyers, not engineers
Best for
- Founders reading a proposal full of unfamiliar terms
- Non-technical stakeholders joining a build decision
- Anyone comparing vendors who define the same word differently
Not best for
- Engineers looking for implementation detail
- Formal or academic definitions
- Language-specific or framework-specific API reference
11 terms
Engagement and commercial
- Burn rate
- Monthly net spend. In an engineering context, the useful version is the fully-loaded monthly cost of your delivery capacity — team plus infrastructure plus tooling — because that is what a roadmap is actually purchased with.
- Discovery
- A short, paid phase that turns a rough idea into a scoped, estimable plan — user flows, data model decisions, technical constraints, and a delivery sequence. Usually 10–20% of project cost, and the cheapest place to change your mind. How we build →
- Fixed priceFixed scope
- An agreed deliverable at an agreed price, with the supplier carrying the risk of underestimating. Typically carries a 15–30% premium for that risk. Honest only when the scope is genuinely knowable up front; on a moving scope it converts every discovery into a commercial negotiation. Dedicated team vs project outsourcing →
- Fully loaded cost
- The real annual cost of an employee including salary, benefits, taxes, recruitment, equipment, and management time — typically 1.3 to 1.5 times salary. The number to use when comparing in-house against an agency, because comparing salary against invoice is not a comparison. In-house vs outsourcing →
- IP assignment
- The contractual transfer of ownership of the code and designs to you. It should be signed before any technical work begins, not at handover. Without it, a supplier can arguably retain rights over work you paid for. Security and compliance →
- Managed teamDelivery pod · Product pod
- A supplier-run team, usually three to five people including QA and a named delivery owner, that takes a defined scope and is accountable for delivering it. Costs more per head than augmentation because the management and review are included rather than yours to provide. Staff augmentation vs managed team →
- Nearshore development
- Outsourcing to a nearby country with a small time difference — Eastern Europe for UK and EU buyers, Latin America for US buyers. Roughly double offshore rates, bought for six to eight hours of natural overlap rather than for capability. Offshore vs nearshore vs onshore →
- Offshore development
- Building software with a team in a distant, lower-cost market — most commonly India for English-language delivery, at roughly 60–75% below US and Western European rates. What you trade is spontaneous same-hour collaboration, which is why overlap hours should be contractual rather than assumed. Offshore vs nearshore vs onshore →
- Runway
- How many months a company can operate before running out of money. It matters in a software decision because it determines whether you can absorb a two-to-four month hiring gap, and how much of the budget must be reserved for what you learn after launch.
- Staff augmentation
- Placing external engineers inside your existing team, under your management, working your backlog. You keep accountability for delivery; the supplier only supplies vetted people. It is cheaper per head than a managed team and fails badly if you have nobody to direct the work. Staff augmentation vs managed team →
- Time and materialsT&M
- Billing for hours actually worked rather than for a fixed deliverable. You carry the scope risk, which means no risk premium in the rate and no change-request negotiation when priorities move. The right shape when the product is still being discovered. Dedicated team vs project outsourcing →
9 terms
Delivery and process
- Code audit
- An independent review of a codebase covering architecture, dependency health, security posture, test coverage, and maintainability, ending in a written risk list. Usually one to three weeks, and the standard first step before acquiring, inheriting, or rebuilding software. Software audit services →
- Code review
- A second engineer reading a change before it merges. The cheapest quality control that exists, and the first thing missing when work is bought as raw hours. Its absence is the strongest predictor of a codebase that will need rescuing.
- Handover
- The transfer of a codebase to you or another team: repository access, documentation, credentials, deployment knowledge, and a walkthrough. Should be defined in the contract, because an undefined handover is where inherited projects become expensive.
- MVPMinimum viable product
- The smallest version of a product that lets real users do the one thing that matters, so you can learn whether they want it. Not a cheap version of the full product — a deliberately narrow one. Typically $15,000–$40,000 and 8–16 weeks through an Indian agency. Cost to build an MVP →
- Project rescueTakeover
- Taking over a stalled or failing codebase from another team. Starts with an audit — does it build, are dependencies alive, what does the data model look like — before any decision on stabilising versus rebuilding. Roughly half the time, stabilising is cheaper. Software audit services →
- Proof of conceptPoC
- A throwaway build that answers one technical question — can this integration work, will this model be accurate enough, can this handle the load. Explicitly not production code, and treating it as production code is a common and expensive mistake.
- Scope creep
- Requirements expanding during delivery without the timeline or budget changing to match. It is not caused by clients changing their minds — that is normal — but by changes being absorbed silently instead of re-priced or re-sequenced.
- Sprint
- A fixed period, usually one or two weeks, in which a team commits to a defined set of work and ends with something demonstrable. The value is the boundary: it forces prioritisation and creates a regular point where direction can change cheaply.
- Technical debt
- Work deferred to ship faster, which accrues interest as every later change becomes slower. Some is deliberate and sensible; the dangerous kind is unrecorded, because nobody can price it. The symptom is a team whose estimates keep growing for similar-sized work. Software audit services →
11 terms
Web and frontend
- Accessibilitya11y · WCAG
- Building so people using screen readers, keyboards, or with low vision can complete the same tasks. WCAG 2.2 AA is the usual target. Increasingly a procurement requirement, and cheaper to build in than to retrofit.
- Core Web VitalsCWV
- Google’s three field metrics for user experience: LCP (how fast the main content paints), CLS (how much the layout jumps), and INP (how quickly the page responds to input). They are a ranking input, and more importantly they correlate with conversion. Performance services →
- Cumulative Layout ShiftCLS
- A score for how much page content moves unexpectedly while loading. Good is under 0.1. Caused by images without dimensions, late-loading fonts and banners, and content streamed in out of document order.
- Design system
- A documented set of reusable components, tokens, and rules shared between design and engineering. Its value is that the tenth screen costs less than the first; without one, every screen is rebuilt slightly differently. UI/UX and design systems →
- Hydration
- The step where JavaScript attaches behaviour to server-rendered HTML in the browser. Heavy hydration is a common cause of a page that looks ready but does not respond to clicks for a second or more.
- Incremental static regenerationISR
- Serving a static page while regenerating it in the background on a schedule or on demand. Gives you static performance with content that stays fresh, without rebuilding the whole site for one change.
- Interaction to Next PaintINP
- How long the page takes to visibly respond after a user interacts. Replaced First Input Delay as a Core Web Vital in 2024. Usually limited by how much JavaScript is competing for the main thread.
- Largest Contentful PaintLCP
- Time until the largest visible element finishes rendering. Google’s threshold for good is 2.5 seconds. The usual causes of a bad LCP are an unoptimised hero image or an element whose visibility is gated behind JavaScript.
- Progressive enhancement
- Building so that core functionality works without JavaScript, then layering enhancement on top. Improves resilience, accessibility, and crawlability at essentially no cost when designed in from the start.
- Server-side renderingSSR
- Generating a page’s HTML on the server so it arrives complete rather than being assembled by JavaScript in the browser. Matters commercially because search crawlers and AI crawlers largely do not execute JavaScript — an unrendered page can simply be invisible to them. Next.js development company →
- Static site generationSSG
- Rendering pages to HTML at build time rather than per request. The fastest and cheapest way to serve content that does not change per user, and the default we choose for marketing and content pages.
13 terms
Backend and data
- Caching
- Storing a computed result so it does not have to be recomputed. Cheap to add and easy to get wrong: the hard part is never storing the value, it is deciding when it becomes invalid.
- Connection pooling
- Reusing a limited set of database connections across many requests. Databases have hard connection limits, so a service without pooling falls over under exactly the traffic spike you wanted it to survive.
- Event-driven architecture
- Services communicating by emitting and consuming events rather than calling each other directly. Decouples systems and absorbs load spikes; costs you the ability to reason about a request by reading one call stack.
- GraphQL
- A query language letting clients request exactly the fields they need from a single endpoint. Genuinely useful when many clients have divergent data needs; overhead without benefit for a single web app.
- Idempotency
- The property that performing an operation twice has the same effect as performing it once. Essential anywhere money moves: without idempotency keys, a retried payment webhook charges the customer a second time.
- Index
- A data structure that lets the database find rows without scanning the whole table. The single most common cause of a slow query is a missing or unusable index, and the fix is usually minutes rather than a migration. Hire PostgreSQL developers →
- Multi-tenancy
- Serving many customer organisations from one application while keeping their data isolated. The model — shared tables, schema per tenant, or database per tenant — is one of the most expensive architecture decisions to reverse later. Cost to build a SaaS app →
- ORMObject-relational mapper
- A library that maps database rows to objects in code — Prisma, Drizzle, Mongoose. Speeds up ordinary work and hides query cost, which is why most ORM performance problems are really unreviewed generated queries.
- QueueJob queue · Worker queue
- A buffer holding work to be processed asynchronously. Keeps slow operations — emails, exports, image processing — off the request path, so a user never waits for something that did not need to be immediate.
- Rate limiting
- Capping how many requests a client can make in a period. Protects against abuse, runaway integrations, and cost blowouts on paid downstream services. Needed on any endpoint reachable without authentication.
- REST API
- The dominant convention for HTTP APIs: resources addressed by URL, actions expressed with HTTP verbs, state passed as JSON. Simple to cache, document, and debug, which is why it remains the default for most products. API design services →
- Webhook
- An HTTP callback a third party sends when something happens in their system. Reliable handling requires signature verification, idempotency, retries, and a dead-letter path — providers will send duplicates and will retry.
- Zero-downtime migration
- Changing a database schema while the application keeps serving traffic, usually via expand-and-contract: add the new structure, backfill, dual-write, switch reads, then remove the old. Each step is reversible, which is the point.
7 terms
Mobile
- App Store review
- Apple and Google’s approval process before an app or update goes live. Rejections are routine and mostly procedural once you have seen them; the cost is the days lost, which is why release experience matters more than it sounds.
- Background location
- Tracking a device’s position while the app is not in the foreground. Both mobile operating systems restrict it aggressively to protect battery, which makes reliable continuous tracking genuinely hard engineering rather than a setting. Cost to build a delivery app →
- Cross-platform development
- Building one codebase that ships to both iOS and Android, usually with React Native or Flutter. Roughly halves the cost of a product app against separate native builds. The wrong choice for games and heavy AR. Cost to build a mobile app →
- Flutter
- Google’s cross-platform framework, written in Dart, which renders its own widgets rather than using platform controls. Strongest where the interface is heavily custom and animation quality is part of the product. Hire Flutter developers →
- Native module
- Platform-specific code written to expose a capability a cross-platform framework does not wrap. Usually needs writing twice, once per platform, which is why an unusual hardware requirement can add materially to a mobile estimate.
- Over-the-air updateOTA
- Shipping JavaScript-level changes to an installed app without a store submission. Useful for fixes and copy changes; constrained by store policy, and no substitute for a proper release process.
- React Native
- A framework for building native mobile apps in React and TypeScript. Strongest when you already have React on the web, because business logic and engineers are shared across both. Hire React Native developers →
9 terms
Cloud and DevOps
- Blue-green deployment
- Running two production environments and switching traffic between them on release. Makes rollback instant, which changes deployment from a risk to be scheduled into something routine.
- CI/CDContinuous integration · Continuous delivery
- An automated pipeline that runs tests, builds once, and deploys the same artefact to each environment. Turns a manual, frightening release into a routine one, which is what actually raises a team’s shipping frequency. DevOps services →
- ContainerisationDocker
- Packaging an application with its dependencies so it runs identically everywhere. Removes the entire class of problems that begin with "it works on my machine", and is a prerequisite for most modern deployment platforms.
- Infrastructure as codeIaC · Terraform
- Defining servers, networks, and services in version-controlled files rather than by clicking in a console. Makes environments reproducible and changes reviewable, and is the difference between an outage you can roll back and one you cannot.
- KubernetesK8s
- A container orchestration platform. Powerful and genuinely necessary at scale, and a permanent operational overhead most products under a few million daily requests do not need — ECS or a managed platform is usually the better answer.
- NAT gateway
- An AWS component letting private resources reach the internet, billed both hourly and per gigabyte processed. One of the most common causes of a surprising cloud bill, and frequently replaceable with a cheaper design. Hire AWS cloud engineers →
- Observability
- Being able to understand what a system is doing from its logs, metrics, and traces without adding new code. The practical test is whether you can diagnose an incident at 2am from what is already being recorded. Observability and monitoring →
- Serverless
- Running code without managing servers, billed per invocation. Excellent for spiky, event-driven, short-lived work; at sustained load it typically becomes more expensive than an equivalent container, and that crossover is worth calculating. Serverless and edge services →
- Vendor lock-in
- Dependence on a provider whose pricing or roadmap you do not control, where leaving would cost more than staying. Rarely worth avoiding at all costs; always worth knowing the size of before you build on something.
9 terms
AI
- AI agent
- A system where a language model plans and takes multi-step actions through tools rather than only producing text. Costs far more than a chatbot to build because anything that changes state needs permissions, confirmation, audit, and a spend ceiling. Cost to build an AI agent →
- Embedding
- A numeric representation of text that places similar meanings close together, letting you search by meaning rather than keyword. The storage layer behind most retrieval systems.
- Evaluation harnessEvals
- A labelled test set plus scoring pipeline that turns "does the AI feature work" into a number. Without one, prompt engineering is guesswork with a confident tone, and nobody can tell whether a change helped. Cost to build an AI agent →
- Generative engine optimisationGEO · AEO
- Structuring content so AI assistants can quote it: direct answers near the top, facts in tables, question-shaped headings, and honest statements of limits. Distinct from SEO because the goal is citation rather than a ranking position. Comparison guides →
- Hallucination
- A model producing fluent output that is not supported by its inputs. Mitigated by grounding answers in retrieved sources, requiring citations, and measuring accuracy — not by instructing the model to be accurate.
- Large language modelLLM
- A model trained to predict text, used to generate, summarise, classify, and reason over natural language. In product terms it is a component with a per-request cost and a measurable accuracy, not a feature that is simply on or off. Hire AI and ML engineers →
- Prompt caching
- Reusing the processed form of repeated context across requests so it is not paid for every time. One of the most effective levers on AI running costs when a large system prompt or document set is sent repeatedly.
- Retrieval-augmented generationRAG
- Retrieving relevant documents and passing them to a language model so its answer is grounded in your own content. Quality is dominated by retrieval, not by the model — most disappointing RAG systems have a search problem, not an AI problem. Cost to build an AI chatbot →
- Vector database
- A store optimised for similarity search over embeddings. Often unnecessary as a separate service — pgvector inside an existing PostgreSQL instance covers most product-scale workloads at no extra operational cost.
10 terms
Quality and security
- Data processing agreementDPA
- The contract governing how a supplier may handle personal data on your behalf — purposes, security measures, sub-processors, and deletion. Required under GDPR whenever a third party touches personal data.
- DPDP ActDigital Personal Data Protection Act
- India’s data protection law, governing consent, purpose limitation, and the obligations of organisations processing personal data. The relevant regime for products serving Indian users, alongside GDPR where EU users are involved.
- End-to-end testingE2E
- Automated tests driving the real application the way a user would. The most expensive tests to maintain and the only ones that prove a whole flow works, so they belong on the journeys that would cost you money if they broke. Hire QA engineers →
- Flaky test
- A test that passes and fails without the code changing. Corrosive because a suite people learn to ignore provides no safety at all. Often caused by timing assumptions, shared state, or a genuine race condition the test was right to catch.
- GDPR
- EU regulation governing personal data: lawful basis, minimisation, subject rights, and rules on transfers outside the EU. Working with a non-EU supplier is routine but requires a data processing agreement and standard contractual clauses. Security and compliance →
- Least privilege
- Granting every person and service only the access they need, and no more. The single highest-value security practice relative to effort, because it limits how much damage any one compromised credential can do.
- Penetration testPen test
- A controlled attack by security specialists to find exploitable weaknesses, ending in a report of findings by severity. Distinct from an automated scan, and usually required annually once you sell to enterprise customers. Security and compliance →
- Regression
- Something that used to work and no longer does. The reason automated tests exist: manual regression checking does not scale past a handful of releases, and it shrinks exactly when deadlines are tightest.
- Single sign-onSSO · SAML · OIDC
- Letting users authenticate through their organisation’s identity provider rather than a separate password. Frequently the gate on selling B2B software to larger companies, and typically $6,000–$15,000 to add.
- SOC 2
- An audited report on a company’s security controls, commonly requested by enterprise buyers. First-year cost is typically $20,000–$60,000 including auditor fees, tooling, and remediation. Worth doing when buyers ask, and rarely before. Security and compliance →
Go deeper
Where these terms turn into decisions
The guides that put numbers and trade-offs behind the vocabulary.
pricingSoftware development costPrice bands and running costs for every kind of project.ExplorecompareComparisons and alternativesAgency, freelancer, in-house, offshore — compared honestly.Exploreq&aAnswersEvery question we answer across the site, in one index.ExploreteamHire developersRate cards and availability across every role we staff.Explore