Glossary

Technical terms explained briefly

terms: 476
Areas: 8
updated: Nov 13, 2026

MonWedFri

Sep

Oct

Nov

All terms

476 terms from A to Z, each explained briefly. Search for a term or jump to a letter.

A

A record: An A record maps a DNS name to an IPv4 address. It does not configure the website or define HTTPS, paths or an HTTP redirect.
AAAA record: An AAAA record maps a DNS name to an IPv6 address. Suitable network and server configuration is needed to make the service reachable through it.
Acceptance criteria: Acceptance criteria state verifiable outcomes used to assess whether a requirement meets its intended need.
Acceptance test: A check against agreed criteria that a solution meets intended user and business needs.
Accessible name: An accessible name labels a control for assistive technology. It should explain the action and agree with the visible text.
Acquiring: Acquiring covers accepting and processing card payments on the merchant side. The acquirer has a different role from the card issuer.
Agentic workflow: A process combining AI-assisted decisions and tool actions, with fixed checks and variable paths through the work.
Agile software development: Developing software in verifiable increments and adapting the plan based on feedback.
AI agent: An AI system that selects its next steps within defined limits and uses tool results to guide further work.
AI bias: Systematic distortions in AI systems that can disadvantage particular cases or groups of people.
AI chatbot: An AI chatbot supports dialogue using AI. A clear answer is not the same as a business operation actually completed.
AI guardrails: Guardrails constrain and check AI application behaviour. Model instructions and technical enforcement perform different tasks.
AI hallucination: An AI hallucination is plausible-looking content that is factually false or invented. Fluent language is not evidence of correctness.
AI inference: Applying an already trained model to inputs to obtain a result.
AI readiness: AI readiness concerns the foundations of an AI project: objectives, usable data, skills, responsibilities and an operating workflow.
Alternative text: Alternative text conveys an image’s meaning or function in context. Purely decorative images need different treatment.
Anomaly detection: Methods that flag unusual observations relative to a defined reference or a learned pattern.
API: An API defines how software can use functions or data offered by other software.
Architecture decision: An architecture decision establishes a significant aspect of software structure. Context and consequences explain why the choice was made.
Architecture principle: An architecture principle provides a reasoned guideline for recurring decisions about system design.
Argon2: Argon2 is a family of memory-intensive password hash functions. The variant, memory use and runtime must suit security and operational requirements.
ARIA: ARIA adds information about interface roles, states and properties for assistive technologies. The stated behaviour must actually be implemented.
Artificial intelligence: Artificial intelligence covers methods such as language processing and pattern recognition. Suitability is assessed against a specific task.
Asynchronous programming: Waiting for a result without blocking the rest of the intended workflow.
Attribute-based access control: ABAC evaluates access using defined properties and rules. The origin and freshness of those properties are part of the design.
Audit log: An audit log makes important application events traceable. Actor, action, affected object and outcome need to be readable together.
Autoscaling: Autoscaling adjusts resources using rules. Metrics, startup times and the limits of dependent services determine whether the application benefits.
Average order value: Average order value needs an order-level calculation. An import with multiple line items per order must not inflate the order count.

B

B2B commerce: B2B commerce is trade between businesses. Digital workflows need to connect the company, ordering person and applicable terms correctly.
B2C commerce: B2C commerce is a business selling to individual consumers. The term describes the buyer relationship, not a particular shop technology.
Backend: The backend handles an application’s server-side rules and data. It is more than a database or an administration screen.
Backpressure: Backpressure adjusts incoming data to the receiver’s processing capacity through signals to slow down or pause.
Backup: A backup is a saved copy used to restore required data or programs later.
Backward compatibility: A new version continues supporting agreed existing usage without requiring changes to existing consumers.
Bare-metal server: A bare-metal server provides exclusive physical compute hardware. Clarify who manages the application, operating system and recovery.
Batch processing: Batch processing handles a defined set of data or jobs together in one run.
Billing address: The billing address is a separate address in an order or invoicing process. It must not be accidentally swapped with the delivery address.
Blue-green deployment: Blue-green deployment prepares a new version in a second environment. After checks, traffic switches from the current environment to the new one.
Branch: A named development line that Git tracks through a movable reference to a commit.
Breaking change: A change that prevents previously supported consumers or uses from working as agreed.
Browser cache: The browser cache stores suitable HTTP responses for later use. Reuse and revalidation follow the applicable rules.
Build process: The defined workflow that produces software artifacts from inputs such as source code and dependencies.
Bundler: A tool that groups connected modules and resources into files for delivery.
Business process automation: Business process automation connects recurring steps with rules, responsibilities and defined ways to handle exceptions.
Business rule: A business rule describes a business condition or decision that governs how an operation is handled.
Buy now, pay later: BNPL enables a purchase with later payment or instalments. The customer’s payment plan and settlement with the merchant are considered separately.

C

Cache invalidation: Cache invalidation ends reuse of a stored copy under defined rules. The source record remains a separate concern.
Cache stampede: A cache stampede occurs when concurrent cache misses trigger the same expensive calculation or data query repeatedly.
Caching: Caching reuses stored data or results. Validity and the permitted user scope belong in the design.
Canary deployment: A canary deployment initially gives a new version limited traffic. Observed outcomes determine whether expansion continues or stops.
Canonical URL: A canonical URL is the preferred address for identical or very similar content; search engines assess several signals when choosing it.
Capacity planning: Capacity planning connects expected usage with performance targets and resources. Measurements and load tests check assumptions, which are updated over time.
Cart abandonment: Items are added to a cart without completing the purchase. Analysis needs a time window and a rule for matching later orders.
Catalogue access permission: Catalogue access permission governs access to a catalogue and its contents. A visible link or a successful sign-in alone is insufficient verification.
Change data capture: Change data capture records database changes so that downstream applications can process them.
Change request: A change request records a proposed change to an agreed project baseline for assessment and decision.
Chargeback: A chargeback reverses a card payment through a dispute with the card issuer. It follows a different process from a merchant-initiated refund.
Checkout: Checkout leads from the cart to order submission. The current selection, required details and order summary need to agree.
Circuit breaker: A circuit breaker stops calls to a failing service and later allows limited probe calls to check recovery.
Clean Architecture: Separating business rules from external details and directing source-code dependencies towards the core.
Cloud computing: Cloud computing provides IT resources over a network on demand. Service and deployment models determine which tasks remain with the company.
CNAME record: A CNAME record points a DNS alias to another name. It is not an HTTP redirect and does not configure the destination server.
Code review: Focused examination of a source change for design, behaviour, clarity and appropriate checks.
Code splitting: Dividing program code into separately loadable parts and planning when they load.
Commit: A recorded project state with metadata and, where applicable, links to earlier states in Git.
Compiler: A program that translates source code into another target representation.
Component library: A component library brings together reusable interface elements. Behaviour, variants and the right version also matter when using them.
Component variant: A component variant is a defined form of a shared building block. It differs from a temporary interaction state.
Composable commerce: Composable commerce connects specialised commerce components. Selection and replacement need suitable interfaces and shared business rules.
Computer vision: Methods for extracting information from images and video, such as classifying objects, locating them or distinguishing image regions.
Concurrency: Organising multiple tasks so that their processing periods can overlap.
Confidence score: A confidence score assesses a model prediction according to a particular method. It does not grant business approval.
Confusion matrix: A table comparing reference classes with model predictions and counting correct and incorrect assignments.
Container: A container runs application processes in an isolated environment. It can simplify deployment, while backups and operational responsibilities remain necessary.
Container image: A container image is the package used to start containers. It includes software and dependencies, but does not automatically include current business data.
Container registry: A container registry stores and distributes images. Tags can change; a digest identifies content but does not establish approval of the software version.
Content Delivery Network: A CDN distributes web content across locations. Public files can be cached there; personalised portal responses need their own cache rules.
Content management system: A CMS lets editors maintain digital content. Content structure, drafts and publication are configured for the needs of the website or application.
Content Security Policy (CSP): A CSP can restrict permitted script sources in the browser. A report-only policy observes violations but does not enforce those rules.
Context engineering: Context engineering prepares information for individual model calls: relevant instructions, messages, sources and current states.
Context window: The context window limits how many tokens a language model can consider in one processing step. It is not a permanent knowledge archive.
Continuous delivery: Continuous delivery prepares software changes for release through a repeatable process. Approval for production can remain a deliberate decision.
Continuous integration: Continuous integration regularly merges changes into a shared codebase. Automated checks can reveal certain conflicts earlier.
Contract test: A test of agreed messages between a consumer and a provider at an application interface.
Conversion rate: Conversion rate expresses defined goal actions as a share of an explicitly named base.
Core Web Vitals: Core Web Vitals describe loading experience, responsiveness and visual stability. The measurement environment and underlying data should be reported with the result.
Correlation ID: A correlation ID links related messages within an operation. It can cover several events and does not replace duplicate detection.
Crawl budget: Crawl budget describes which URLs Google can and wants to fetch. Server capacity and crawl demand work together.
Crawling: Crawling is the automated fetching of web pages and resources by a bot, for example for processing by a search engine.
CRM: CRM organises work on customer relationships. A quotation example explains its use and the basic distinction from an ERP.
Cross-Origin Resource Sharing: CORS controls browser access to responses across origin boundaries. It replaces neither API login nor permission checks.
Cross-selling: Cross-selling suggests additions to a selected offer. In software, the add-on, technical prerequisite and existing access rights must fit together.
Cross-site request forgery (CSRF): CSRF exploits a signed-in session for an unwanted request. A session cookie alone does not prove that the user intended the action.
Cross-site scripting (XSS): In XSS, external content becomes executable code in the browser. The key issue is how an application inserts input into its pages.
CRUD: CRUD stands for the basic data operations Create, Read, Update and Delete.
CSS: CSS describes the presentation of structured content, including layout, colours and typography. It must fit interaction and actual content.
Current-state analysis: Current-state analysis examines how processes and systems actually work today, including exceptions and unresolved questions.
Customer retention rate: Customer retention needs a fixed starting group. A SaaS report must count customers distinctly and separate new accounts from reactivated ones.
Customer-specific assortment: A customer-specific assortment defines which items are offered to a customer. Selection, prices and access rights need to fit together in the shop.

D

Data consistency: Data consistency means that data follows defined rules and related states do not contradict each other.
Data drift: A change in the distribution of model inputs relative to an explicitly chosen reference.
Data mapping: Data mapping defines how source data enters a target structure. Field names, meaning, units and error handling all need to fit.
Data migration: Data migration moves selected data to a new system. Mapping, validation and the transfer of responsibility for writes belong to the process.
Data minimisation: Data minimisation limits personal data to what is adequate, relevant and necessary for a defined purpose.
Data retention policy: A data retention policy defines the purpose, start, duration and end of storage. Deletion jobs and copies must be addressed in the implementation.
Data schema: A data schema describes the structure and rules of data. It defines fields and data types, for example; provenance and freshness are additional questions.
Data synchronisation: Data synchronisation repeatedly reconciles agreed data between systems. Direction, delay, conflicts and safe retries need explicit rules.
Data validation: Data validation checks data against defined requirements such as format, required fields or permitted value ranges.
Data version: A data version identifies a particular state of data. Whether identifiers encode a sequence must be defined; a schema version is a different concept.
Database: An organised collection of data that software can access selectively and apply changes to.
Database index: A database index supports suitable lookups. It does not accelerate every query and adds storage and data-change overhead.
Database lock: A database lock coordinates concurrent access to data or database objects so incompatible operations do not proceed unchecked.
Database query: A database query requests data under defined conditions. Selection, result scope and ordering should fit the work being done.
Database transaction: A database transaction groups related changes so they are committed together or rolled back together.
DDoS protection: DDoS protection aims to mitigate distributed overload attacks. Covered services, network capacity and effects on normal traffic need checking.
Dead-letter queue: A dead-letter queue holds messages set aside for investigation and controlled reprocessing.
Dealer portal: A dealer portal gives affiliated dealers digital access to agreed information and workflows. Its scope and company assignment need to be clear.
Debugging: Systematically narrowing down an error’s cause through behaviour, data and program execution.
Decision brief: A decision brief prepares a specific choice by presenting options, rationale, consequences and unresolved assumptions.
Deep learning: Machine learning with neural networks containing multiple processing layers.
Definition of Done: The shared description of the quality conditions a completed product Increment must meet.
Delivery capability: Delivery capability describes whether a specific request can be fulfilled. A stock value alone does not fully answer questions of quantity, destination and time.
Dependency injection: A component receives dependencies from outside instead of creating concrete implementations itself.
Design handoff: A design handoff passes a design on for implementation. States, behaviour and outstanding decisions need to be clear.
Design pattern: A reusable solution structure for a recurring software design problem, with its own trade-offs.
Design system: A design system combines design rules, reusable components and usage guidance. It also needs maintenance and traceable changes.
Design token: A design token stores a named design decision. Multiple components can use it, allowing a value to be changed in a defined place.
Destination system: The destination system is intended to receive data on a particular transfer path. Reachability, receipt and completed processing are distinct states.
Digital accessibility: Digital accessibility enables people with disabilities to perceive, understand and operate digital services.
Digital asset management: DAM organises digital media files with metadata, versions and usage rules. An application can then find the appropriate approved file.
Digital wallet: A digital wallet uses saved payment methods or a wallet balance for payment. Device support and the shop integration must work together.
Direct-to-consumer: D2C is a brand’s direct sale to consumers. Shop and ERP integrations must distinguish sales channel, customer type and access rules.
Discount code: A discount code unlocks a reduction during the purchase under defined conditions. Validity, eligible items and combination rules need to align.
DKIM selector: The DKIM selector and signing domain point to the public verification key. It helps map keys to specific sending routes.
DMARC: DMARC links a successful SPF or DKIM check to the visible sender domain. Policy and reports help with its technical oversight.
DMARC aggregate report: A DMARC aggregate report summarises a reporting mail system’s domain checks over a period. It helps examine sending sources and misconfigurations.
DMARC policy: A DMARC policy describes the requested handling of failed domain checks. It repairs neither sending sources nor their alignment.
DNS zone: A DNS zone is an administrative unit for DNS data. Its boundaries follow delegation; not every subdomain automatically becomes a separate zone.
Docker: Docker provides tools to build, distribute and run containerised applications. The image, running container and operating environment serve different purposes.
Document chunking: Splitting a document into smaller coherent sections for search and model context.
Domain: A domain is a named part of the DNS namespace. The name does not create a website; hosting, email and the application still need to be connected.
Domain Name System: DNS supplies data about names, such as a web server’s address. Name resolution, the server connection and application behaviour are separate steps.
Domain-driven design: Developing software models from the business domain together with the people who know its rules.
DomainKeys Identified Mail: DKIM links signed email components to a domain. A valid signature establishes neither encryption nor the person behind the message.
Dropshipping: With dropshipping, a supplier sends the sold item directly to the customer. Shop orders, supplier orders and updates need to match.
Duplicate detection: Duplicate detection finds records that might describe the same entity. A match is initially only a suspected duplicate, not an automatic decision to merge.
Dynamic pricing: Dynamic pricing adjusts prices to changing conditions. Its data, permitted limits and handling of existing orders need clear rules.

E

E-commerce: E-commerce describes trade through digital ordering channels. Payment and delivery need not take place online as well.
E-commerce merchandising: E-commerce merchandising controls product selection and presentation in shop software. Filters, sorting and promotions need explicit precedence rules.
Edge computing: Edge computing processes data closer to its source or use. Which work runs locally and how it interacts with central services still need to be planned.
Electronic Data Interchange: EDI transfers structured business documents between systems. Format, partner profile, data mapping and responses need to agree.
ELT: ELT stands for Extract, Load, Transform: raw data first enters the target store and is then transformed there for the required analyses.
Email deliverability: Deliverability is about whether expected email actually reaches the inbox. Check sending status, authentication and actual receipt separately.
Embedding: An embedding represents content as a numeric vector. The model and comparison method determine which similarities become visible.
End-to-end test: A check of an agreed workflow across the system parts essential to it.
Enterprise Service Bus: An Enterprise Service Bus mediates centrally between applications: it can route messages, convert data formats and adapt communication protocols.
Entity tag: An ETag identifies a particular representation version of an HTTP resource and supports conditional reads or changes.
Envelope encryption: Envelope encryption separates the data key from the key protecting it. The encrypted file and wrapped data key need to be managed together.
Envelope-From: Envelope-From identifies the SMTP sender for transport and error feedback. It must be distinguished from the visible sender address.
ERP: ERP connects core business processes such as purchasing, inventory, orders and finance through shared data.
Error handling: Defining how a program handles detected errors and responds appropriately.
ETL: ETL stands for Extract, Transform, Load: data is read from a source, transformed before loading and then transferred into the analytics store.
Evaluation matrix: An evaluation matrix compares options against common criteria and records assessments together with their basis.
Event ID: An event ID identifies an event within an agreed scope. A repeat delivery needs to be distinguished from a new event.
Event tracking: Event tracking records selected actions on websites or in apps. A clearly defined trigger explains what a counted event actually establishes.
Event-driven architecture: In event-driven architecture, components report events that other components can respond to independently.
Eventual consistency: Eventual consistency permits temporarily different data states that should converge under the intended operating conditions.
Exit strategy: A prepared route to switch, take over or retire an IT service in an orderly way.
Explainable AI: Approaches that make an AI system’s results or operation understandable to the intended recipients.
Exponential backoff: Exponential backoff multiplies the waiting interval between retries, up to a configured limit.

F

F1 score: The harmonic mean of precision and recall for the same positive class.
FaaS: Function as a Service runs deployed functions in response to events or calls. A platform provides and manages the execution environment.
Faceted navigation: Faceted navigation narrows product lists using several attributes. Visible filters and structured product data need to describe the same selection.
Feature flag: A feature flag switches planned application behaviour through configuration. It can separate technical deployment from making a feature available.
Fine-tuning: Fine-tuning adapts a pretrained model through further training. Supplying current documents at runtime is a different operation.
FinOps: FinOps brings engineering, finance and business teams together on technology cost and value decisions. It relies on understandable usage and cost data.
Firewall: A firewall controls network traffic according to rules. An allowed connection replaces neither secure software nor permission checks within the application.
Focus indicator: A focus indicator shows which control receives keyboard input. It differs from a mouse pointer or a marker for the selected item.
Focus order: Focus order determines the route through focusable elements. It should preserve meaning and operability, including during dynamic changes.
Forecasting model: A model that uses available information to estimate a future quantity for a defined period.
Foreign key: A foreign key protects a defined reference between records through matching key values.
Framework: A framework gives an application a technical structure and shared rules. The tasks it handles depend on the chosen framework.
Frontend: The frontend is an application’s user-facing interface. Helpful browser feedback does not replace authoritative server-side checks.
Fulfilment: Fulfilment covers the operational preparation and shipping of ordered goods. Picking, packing, handover and delivery need traceable events.
Full-stack development: Full-stack development covers frontend and backend work. The term describes scope, not automatically team size or staffing.
Function calling: A language model requests a described function with structured arguments; the connected application executes it.

G

Generative AI: Generative AI produces content such as text, images or code. Quality is assessed against its intended use and available source information.
Git: A distributed version control system with local repositories and recorded change history.
GitOps: GitOps combines a versioned desired state with automatic retrieval and continuous reconciliation. Configuration files in Git alone cover only part of this process.
Global Trade Item Number: A GTIN identifies a trade item across businesses. Internal item codes, serial numbers and barcodes have different roles.
GraphQL: GraphQL lets applications request fields through a typed API schema. The server remains responsible for data access and permissions.
Grounding: Grounding links an AI answer to verifiable evidence. A source link alone does not show that every claim is supported.
Guest checkout: Guest checkout permits a purchase without mandatory account sign-in. Contact and delivery details needed for the order are still recorded.

H

Headless CMS: A headless CMS manages content separately from presentation and supplies it through APIs. A website or app needs its own rendering logic.
Headless commerce: Headless commerce separates the storefront from the commerce system. Interfaces connect its data and commerce capabilities.
Hotfix: An urgent, targeted correction for a problem in a production system.
Hreflang: Hreflang identifies related language or regional versions of a page for search engines.
HTML: HTML describes the structure and meaning of web content. It identifies elements such as headings, tables and input fields.
HTTP Strict Transport Security: HSTS tells a browser to remember HTTPS for future visits to a host. Duration and subdomain coverage require a deliberate operational decision.
HTTP/2: HTTP/2 carries several HTTP exchanges on one TCP connection. Potential efficiency gains do not replace measuring actual page and server performance.
HTTP/3: HTTP/3 carries HTTP over QUIC. Benefits and reachability also depend on the network and endpoints; a universal speed improvement is not established.
HttpOnly cookie: HttpOnly prevents direct cookie access by page scripts. Matching requests can still carry the cookie; session and permission checks remain necessary.
HTTPS: HTTPS protects HTTP communication using TLS. Protected transport alone does not establish correct form processing or a user’s access rights.
Human-in-the-loop: Human-in-the-loop actively involves people in an automated process. Useful review requires understandable information and ways to intervene.
Hybrid app: Combining a web interface with native device integration in an installable application.
Hybrid cloud: Hybrid cloud connects different cloud environments. The joint application needs defined data flows, access rules and behaviour when connections fail.
Hybrid search: Combining lexical search and vector search into a shared result list.
Hydration: Hydration connects existing browser HTML with its application logic. Visible content and working interaction require separate checks.
Hypertext Transfer Protocol: HTTP governs requests and responses between client and server. The method, status and content must fit the agreed application behaviour.

I

IaaS: Infrastructure as a Service supplies computing, storage and networks. The operating system and the company’s own applications remain the team’s tasks.
Idempotency: An idempotent operation has the same intended effect on system state when repeated identically as when performed once.
Idempotency key: An idempotency key identifies retries of the same intended request. Scope, retention and API rules determine its effect.
Identity and access management: IAM combines management of digital identities with their access rights. Accounts, role changes and removal of old permissions belong together.
Indexing: During indexing, a search engine analyses content and adds information about it to its search index.
Indexing status: Indexing status describes a URL’s inclusion in a search index. It is distinct from reachability and visibility for a particular search.
Infrastructure as Code: Infrastructure as Code describes infrastructure in files that can be versioned and reviewed. Managed resources only change when those changes are actually applied.
Integration test: A check that selected software parts work together as intended through their interfaces.
Interactive prototype: An interactive prototype responds to selected actions. It can reveal a path through an interface without implementing the real processing behind it.
Interface contract: An interface contract describes the agreed data exchange. Data structure, responses and their meaning need to fit together for both sides.
Internal product search: Internal product search finds items in the shop from a query. Customer vocabulary and maintained catalogue data need to match.
Inventory management: Inventory management brings quantities, locations and states together. Physically present goods are not automatically available for further sales.
Inventory reservation: An inventory reservation assigns a quantity to an order or purpose. It becomes unavailable for other operations even though it has not been shipped.
Investment costs: Costs of building a new or expanded IT capability, as considered in project planning.
IP address: An IP address provides addressing on a network. It differs from a domain name and alone says nothing about the application reachable at that address.
iPaaS: iPaaS is an integration platform delivered as a cloud service. The team configures connections, data mappings and error handling on that platform.
IPv4: IPv4 addresses network destinations using 32-bit addresses. A DNS record and a reachable address do not replace configuring and checking the service.
IPv6: IPv6 uses 128-bit addresses. A portal offered over it needs matching DNS, network, server and access-rule configuration.

J

JavaScript: JavaScript is a programming language for application logic, in browsers and elsewhere. Its role differs from HTML and CSS.
JSON Schema: JSON Schema describes the structure and constraints of JSON data. A validator checks these rules; conformance alone does not establish business correctness.

K

Key rotation: Key rotation introduces new key material. Older data and backups need a plan for the key versions they still require for decryption.
Keyboard operability: Keyboard operability makes tasks possible without a mouse. Actions must be reachable through a clear sequence of controls.
Keyboard trap: A keyboard trap leaves keyboard users stuck within a region. Deliberately containing focus in a dialog needs a keyboard-operable exit.
Keyword: A keyword is a word or phrase used in search and content planning. In search ads, it is a targeting term and need not match the actual query word for word.
Knowledge distillation: Training a student model using behavioural information from a teacher model or an ensemble of models.
Kubernetes: Kubernetes manages containerised applications against a declared desired state. Its value also depends on configuration and operational responsibilities.

L

Language model token: Tokens are processing units for language models. They do not reliably correspond to whole words and differ from design tokens.
Large language model: A large language model processes language using learned patterns. Fluent text proves neither factual accuracy nor access to current business data.
Lazy loading: Loading resources when the intended user step needs them rather than immediately.
Least privilege: Least privilege limits permissions to the required task and data scope. Necessary work remains possible while unnecessary powers are removed.
Legacy system: An existing system still in use, with technical and business dependencies accumulated over time.
Linter: A tool that examines code for particular problems using configured rules.
LLM evaluation: LLM evaluation tests model outputs against defined tasks and criteria. One good example does not establish overall quality.
Load balancer: A load balancer distributes connections or requests across several destinations. Distribution rules, state handling and failure checks determine its behaviour.
Long-tail keyword: Long-tail keywords are search phrases with relatively low individual demand. They are often specific; length alone proves neither volume nor business value.

M

Machine learning: Models learn patterns from data and apply them to new inputs.
Mapping table: A mapping table stores specific correspondences, such as identifier pairs from two systems. It makes a confirmed link reusable.
Master data: Master data describes core business entities such as customers or products. It is reused across processes and can still change over time.
Master Data Management: Master Data Management organises the identity, quality and maintenance of important master data. Technology, business ownership and change rules work together.
Memory leak: Memory that is no longer needed stays allocated while the program runs.
Message acknowledgement: A message acknowledgement tells the broker that a delivery has been received or processed according to the agreed rule.
Message broker: A message broker accepts messages and routes them to recipients according to defined rules.
Message queue: A message queue buffers messages until a consumer takes them for processing.
Microservices: An architecture of bounded services intended to evolve, operate and deploy as independently as practical.
Middleware: Middleware provides shared software services. In an integration, it can handle communication and data adaptation between applications.
Migration costs: Effort and expenditure for the planned move of data or functions to another environment.
Minimum viable product (MVP): An MVP is a limited product or offer used to test a concrete assumption about customer behaviour.
Model card: A model card documents an AI model’s purpose, foundations, evaluation conditions and known limitations for a particular version.
Model Context Protocol: MCP standardises exchanges between AI applications and servers providing tools, context data or templates.
Model drift: A decline in a model’s predictive quality as data or relationships change during use.
Model quantisation: Representing model weights and, depending on the method, intermediate values with lower numerical precision.
Model routing: Choosing a suitable model for each request from a defined set of permitted candidates.
Modular monolith: An application deployed together with explicitly bounded internal modules and interfaces.
Monitoring: Monitoring continuously collects and evaluates selected system states and functions. What it tells you depends on what is actually observed.
Monolith: An application whose core is deployed as one unit; clear internal modules are still possible.
MTA-STS: MTA-STS publishes requirements for TLS connections to a mail domain. Mail servers, certificates and the policy actually applied belong together.
Multi-factor authentication: MFA uses different types of authentication factors. Asking for the same factor type several times does not create a multi-factor check.
Multi-tenancy: Multi-tenancy supports separate customer or group scopes in one solution. A tenant is not automatically one person or one database.
Multichannel commerce: Multichannel commerce means selling through several channels. Buying journeys can remain separate even when some data is connected.
Multicloud: Multicloud uses services from several cloud providers. This can give access to specialised capabilities but does not create automatic failover.
Multimodal AI: AI using several information types, such as text, images or sound, in inputs or outputs.
Must-have requirement: A must-have requirement is essential to the agreed delivery scope. Without it, the intended outcome is not acceptable within that scope.
MX record: An MX record names a mail server for a domain. Its preference value orders possible destinations; it confirms neither a mailbox nor delivery.

N

N+1 query: In an N+1 pattern, a list query is followed by another fetch for each result. The structure matters, not merely a high call count.
Named entity recognition: NER marks named entities in text and assigns types such as person, organisation or place.
Native app: An application built with its target platform’s own user-interface and system capabilities.
Neural network: A model of connected computational nodes with adjustable parameters.
Node.js: Node.js runs JavaScript outside the browser. The runtime provides technical capabilities; the program supplies the actual business logic.
Noindex: Noindex controls inclusion in search engines that support the instruction. The crawler must be able to read it; it does not replace access protection.
Non-functional requirement: A non-functional requirement describes a solution’s quality or constraints. It can be as mandatory as a functional requirement.
NoSQL: An umbrella term for database systems with different, not primarily relational data models.

O

OAuth 2.0: OAuth enables limited access to protected resources. Permission to use an API does not automatically prove a signed-in person’s identity.
Object-relational mapping: Connecting program objects to data in a relational database.
Omnichannel commerce: Omnichannel commerce connects sales channels for continuous customer tasks. Data and workflows need to support the move between channels.
Online marketplace: An online marketplace gives several merchants a shared selling platform. Product, selected seller and specific offer need to remain connected.
Online shop: An online shop is a digital sales channel with product selection and ordering. The shop system is its underlying software.
Open source: Software whose licence meets the freedoms and conditions defined for open source.
OpenAPI: OpenAPI describes HTTP interfaces in machine-readable form, including paths, parameters and responses. The description must match the running service.
Operating costs: Expenditure to plan for the ongoing operation of an IT solution.
Optical character recognition: Optical character recognition makes writing in images machine-processable. Reading an amount is not the same as checking an invoice against business rules.
Optimistic locking: Optimistic locking detects on save whether a record has changed since the version that was read.
Order approval: Order approval is an internal authorisation under defined rules. It confirms neither payment nor delivery nor acceptance by the ERP system.
Order history: Order history shows previous orders within an agreed account scope. Old prices, current availability and current processing remain separate information.
Order management system: An order management system links information and processing steps for an order. Individual lines can remain in different states.
Order template: An order template stores a reusable selection of items. The template and an order created from it have different states and permissions.
Overfitting: Excessive adaptation to details of training data that reduces a model’s ability to generalise to new examples.

P

PaaS: Platform as a Service provides a platform for custom applications. The team deploys its code; the provider operates the underlying environment.
Package manager: A tool for obtaining and managing software packages, their versions and dependencies.
Pagination: Pagination splits an API result list into retrievable pages. Ordering, continuation and changes during retrieval belong in the API contract.
Parallelism: Actually executing several computations at the same time.
Passkey: A passkey enables sign-in with a service-bound key pair. Device changes, account recovery and account permissions remain separate responsibilities.
Password salting: Password salting gives each new password hash its own random additional value. The salt is stored alongside it and does not need to be secret.
Pay by invoice: When customers pay by invoice, they receive goods or services and pay later by the agreed due date. The invoice and receipt of payment are tracked separately.
Payment gateway: A payment gateway transmits payment data between a shop and payment processing, returning the responses to the shop.
Payment service provider: A PSP supports businesses with electronic payments. Payment methods, processing and additional services depend on the offering.
Payment tokenisation: For card payments, a token can replace the card number. Permitted uses depend on the particular token system.
PCI DSS: PCI DSS sets technical and operational requirements for protecting payment card data. The particular assessment environment determines the scope.
Pilot project: A limited deployment with selected participants to learn about use and operation before expansion.
PostgreSQL: An open-source object-relational database system for structured storage and querying of data.
Precision: The share of actually positive cases among all cases the model marks as positive.
Pricing rule: A pricing rule links conditions to a price or calculation. Product, customer, quantity and validity need an unambiguous match.
Private cloud: A private cloud serves one organisation and can be operated externally. Exclusive use does not replace a clear agreement on operational responsibilities.
Process mining: Process mining examines workflows recorded in event logs. Case identifiers, activities and timestamps provide the basis for analysis.
Product Backlog: The ordered, continually evolving collection of improvements needed for a product.
Product bundle: A product bundle offers several products together. Its contents, price and available components must match the quantity sold.
Product catalogue: A product catalogue organises range information for search and selection. Catalogue entry, inventory availability and channel publication are separate data.
Product configurator: A product configurator guides product choices and their rules. Its result must clearly describe the selection that is currently valid.
Product data quality: Product data quality describes how well information fits its intended use. Complete fields can still describe the wrong goods.
Product discovery: Investigating needs, constraints and risky assumptions before settling on a product solution.
Product feed: A product feed supplies structured item data to a target channel. Field mapping, stable identifiers and freshness need to work together.
Product information management: PIM organises product attributes, descriptions and language versions. A PIM system prepares this information for different output channels.
Product Owner: The person accountable in Scrum for product value, the Product Goal and effective backlog management.
Product variant: A product variant is a particular version of a product. Its attributes must remain consistent across selection, pricing, stock and ordering.
Progressive web app: A web application with added capabilities for installation, device integration and offline use.
Project charter: An internal document that sets out a project’s goals, boundaries, responsibilities and authorisation.
Project dependency: A project dependency describes how a task or decision is tied to other results, dates or resources.
Project milestone: An important point or event in a project whose achievement can be established.
Project scope: The agreed boundary of the outcomes and work a project is intended to include.
Project summary: A project summary briefly brings together an initiative’s purpose, boundaries, owners and key constraints.
Prompt: A prompt is the input to a generative model. A clear task helps but does not replace data connections or access permissions.
Prompt engineering: Deliberately designing and checking tasks, context and output rules for a generative model.
Prompt injection: Prompt injection attempts to redirect an AI application through inserted instructions. Retrieved documents can contain them too.
Proof of concept: A bounded investigation into whether an important assumption or technical approach is viable.
Proprietary software: Software whose licence does not grant general freedoms to modify and redistribute it.
Prototype: A prototype makes a design testable against a specific question. Its form and scope follow the uncertainty that needs to be resolved.
Public cloud: Public cloud describes a cloud model offered to a broad customer base. Access rules and configuration still determine who can see your application data.
Publish-subscribe: Publish-subscribe distributes published messages to interested subscribers without requiring the sender to contact each one directly.
Pull request: A proposal to integrate changes into a target branch, with context for shared review and discussion.

Q

QUIC: QUIC is a secure transport protocol over UDP with its own reliable streams. HTTP/3 uses QUIC; the terms refer to different protocol layers.
Quick order: Quick ordering shortens item entry, for example using product numbers. Validation, the cart and order completion remain separate steps.

R

Race condition: A result depends on the timing of overlapping accesses and can therefore be wrong.
Rate limiting: Rate limiting caps how frequently a client may make requests under a defined rule.
Re-ranking: Re-ranking reassesses search candidates already retrieved. Relevance to the query and the limits of the initial selection both matter.
Real-time processing: Real-time processing handles arriving events against a defined latency target. Event time, processing time and late data need explicit rules.
Recall: The share of actually positive cases that the model correctly identifies as positive.
Recommendation system: A system that selects or ranks relevant candidates from available items, for example using their features or interaction history.
Recovery point objective (RPO): RPO defines the maximum acceptable time gap between recoverable data and an outage.
Recovery time objective (RTO): RTO sets the maximum acceptable time from a service interruption to restoration of the required usable service.
Recurring payment: Recurring payments charge an authorised payment method at agreed intervals. Amount, payment attempt and service status remain separate.
Refactoring: Changing software’s internal structure while preserving its expected observable behaviour.
Refund: A refund returns all or part of an amount already paid. Returning money and returning goods are separate processes.
Regression test: A check for unwanted effects of a change on behaviour that is still required.
Reinforcement learning: A learning method in which a system develops an action strategy from actions and rewards in an environment.
Relational data model: Modelling data in tables with rows, columns and explicitly described relationships.
Reordering: Reordering uses an earlier order as the starting point for a new purchase. Prices, items and permissions need to be checked again.
Repeat purchase rate: Repeat purchase rate is the share of customers who buy again. A report needs an explicit period, customer identity and distinct orders.
Reproducibility of AI outputs: Being able to repeat AI generation under documented conditions and compare results against defined criteria.
Requirements analysis: Requirements analysis clarifies what a solution must achieve and why. Requests are examined for meaning, conflicts and verifiable outcomes.
Requirements catalogue: A requirements catalogue organises requirements with identifiers and verifiable details. It helps track scope, unresolved questions and changes.
Response time: Response time needs a defined start and end. Fetching an API response and reaching a usable view measure different sections.
REST: REST defines architectural constraints for distributed interfaces. Resources, stateless requests and links are part of the approach.
Retrieval-augmented generation: RAG combines information retrieval with generated answers. Access to sources does not remove the need for quality and permission design.
Retry jitter: Retry jitter varies waiting times randomly to spread retries and reduce simultaneous requests from many clients.
Retry strategy: A retry strategy defines which failures to retry, the waiting intervals and the limit on further attempts.
Return-Path: Return-Path records the SMTP reverse path in a header. It explains the technical route for error reports, not a person’s identity.
Reverse proxy: A reverse proxy accepts requests in front of an application and forwards them. Routing, connection boundaries and forwarded information need clear rules.
Robotic process automation: RPA uses software to perform recurring application steps, often through user interfaces. Clear rules and verifiable outcomes matter.
Robots directive: Robots directives give indexing or presentation rules to search engines that support them. The crawler and the delivery route also matter.
Robots.txt: robots.txt gives rules to cooperating crawlers; it provides neither access protection nor reliable exclusion from the search index.
Role-based access control: RBAC connects permissions to roles and assigns people to those roles. The data scope must be clear alongside the role name.
Rolling update: A rolling update replaces running application instances gradually. Old and new versions may work simultaneously during the transition.
Runtime environment: The environment in which a program executes and accesses the functions made available to it.

S

SaaS: Software as a Service means using an application operated by its provider. The company configures users, data and the available settings.
Same-origin policy: The same-origin policy limits browser access between different origins. A different page path alone does not create a new origin.
SameSite cookie: SameSite affects cookie inclusion in cross-site requests. The choice must fit login and navigation flows; CSRF protection remains necessary.
Schema change: A schema change alters database structure, such as tables, columns, data types or constraints on permitted data.
Schema registry: A schema registry centrally manages data schemas and their versions. Compatibility checks help data exchange but do not replace checking meaning.
Scope creep: Scope creep is gradual expansion of project scope without adequate assessment and corresponding changes to the plan.
Screen reader: A screen reader makes digital content available through speech or Braille. Meaningful structure and labels support understandable interaction.
Scrum: A framework for working together incrementally on complex product challenges.
Search intent: Search intent is the goal behind a search query. Wording provides clues, but does not give certain knowledge of an individual’s intentions.
Search suggestion: A search suggestion offers a possible query while someone types. Selection, submission and the later results list need to work together clearly.
Semantic HTML: Semantic HTML uses elements according to their meaning. A real button has properties that a styled text block does not acquire automatically.
Semantic search: Search that finds and ranks results by their semantic relevance to the query.
Semantic versioning: An agreed version scheme expressing intended compatibility of changes to a defined public API.
Sender Policy Framework: SPF checks a sending server’s authorisation to use an SMTP domain. The visible sender line and receipt are separate questions.
Sentiment analysis: Estimating the attitude expressed in text, such as positive, negative or neutral; it does not replace assessment of the underlying business issue.
Server-sent events: Server-sent events carry text notifications from server to client over HTTP, for example about the progress of a background job.
Server-side rendering: Server-side rendering produces HTML on the server. Freshness, personalisation and speed also depend on data access and caching.
Server-side request forgery (SSRF): SSRF makes an application send unwanted network requests from its server. What matters is which destinations the server actually reaches.
Serverless: With managed serverless services, the provider operates the execution infrastructure. Code, access and error handling remain application responsibilities.
Service level indicator (SLI): An SLI is a clearly defined measure of an aspect of service quality that matters to users.
Service level objective (SLO): An SLO sets a target for a defined service-quality measure over a stated period.
Shipping address: The shipping address describes a transaction’s delivery destination. Shop and interface must distinguish it from billing details and the current address book.
Should-have requirement: A should-have requirement matters but is not essential to the delivery scope under consideration. An accepted workaround can initially allow use.
Single Source of Truth: A single source of truth provides a shared, reconciled data foundation. Authoritative sources, mappings and freshness must be traceable.
Single-page application: A single-page application updates views within a loaded web document. It may still have many views and separate URLs.
Skip link: A skip link bypasses repeated navigation and leads to the main content. The destination must also work for subsequent keyboard navigation.
SKU: A SKU identifies an internal product or variant. Two pack sizes show why the correct match matters for ordering.
Software architecture: Fundamental decisions about a software system’s structure, interactions and quality goals.
Software dependency: A software dependency is another software component a program needs. Indirectly introduced packages can also affect operation.
Software library: A software library provides reusable functions or components. An application uses it to solve a particular subtask.
Software licence: The terms under which rights to software are granted.
Software release: A software release is an identified software version made available for a particular use. Publishing it does not automatically switch every system to it.
Source code: The readable, editable description of a program from which its behaviour is produced.
Source system: The source system supplies data for a particular transfer path. That role alone does not establish who authoritatively maintains all business information.
Speech-to-text: Converting spoken language in audio data into written text.
SPF alignment: SPF alignment links the successfully checked SMTP sender domain to the visible sender domain. The DMARC mode determines the required comparison.
SQL: A language for defining, querying and changing data.
SQL injection: SQL injection occurs when external data changes a database query’s structure. Binding parameters handles values, not access permissions.
Stakeholder: Stakeholders are people or groups who influence a change or are affected by it. Participation and decision-making authority need to be distinguished.
State management: Defining where current application state lives, who changes it and which views are derived from it.
Statement of work: A statement of work explains the requested scope: outcomes, boundaries, contributions and prerequisites. It helps put quotes on the same basis.
Static code analysis: Examining code, or a representation of it, without running the application’s workflows to do so.
Static site generation: Static site generation creates pages as files in advance. Changes to source data must also reach the output served to users.
Strong customer authentication: SCA combines independent evidence to authenticate the payer. Successful authentication must be distinguished from payment status.
Structured data: Structured data describes content in a machine-readable format and explicitly identifies its meaning.
Subdomain: A subdomain sits within a parent DNS namespace. It can name a service; its administration and security still need to be planned.
Subscription commerce: In subscription commerce, customers receive products or services on a recurring basis. Delivery or usage plans and billing need to match.
Supervised learning: A learning method that uses inputs together with their corresponding target values or classes as training examples.
System integration: System integration links applications, data and work steps. Interfaces alone are insufficient: identifiers, field meanings and outcomes need to agree.

T

Target architecture: A target architecture describes the intended structure of a system through its main building blocks, relationships and boundaries.
Target-state concept: A target-state concept describes the intended future process or system through tasks, rules and checkable expectations.
Technical debt: A metaphor for internal software shortcomings that make later changes harder or riskier.
Technical SEO: Technical SEO addresses the technical conditions that let search engines fetch, understand and index suitable pages.
Temperature in language models: A setting that influences the distribution of possible next tokens during sampling and therefore their selection.
Test automation: Tools perform defined testing tasks, such as execution and comparison with expected results.
Test coverage: The share of an explicitly defined set, such as executable statements or requirements, that has been tested.
Test data in machine learning: Data reserved for final assessment whose results were not used to select the model variant being assessed.
Test tenant: A logically bounded customer context of a multitenant application, configured for testing.
Text classification: Text classification assigns text to predefined categories. Categories and the handling of ambiguous messages must fit the task.
Text-to-speech: Producing synthetic spoken output from written text.
Throughput: Throughput describes work processed per period. It is neither the duration of one request nor the number of concurrent users.
TLS-RPT: TLS-RPT provides aggregate feedback on TLS connections between mail servers. Reports help investigate transport failures, not prove that mail was read.
Top-p sampling: Selecting the next token from a small set of likely candidates that together cover a probability threshold.
Total cost of ownership (TCO): TCO covers a solution’s total costs over a defined period, including implementation and any later exit.
Training data: The examples from which a model learns relationships or parameters during training.
Transactional inbox: A transactional inbox uses durable message records to protect local processing against duplicate deliveries.
Transactional Outbox: An outbox stores a business change and its outgoing event record together. A separate process publishes the message later.
Transformation rule: A transformation rule defines how input data becomes destination values. The calculation, prerequisites and error cases need describing together.
Transmission Control Protocol: TCP provides applications with a reliable, ordered byte stream. That transport service does not confirm that a request was processed in the business workflow.
Transport Layer Security: TLS protects data in transit between its endpoints. Application permissions, stored data and further connections need their own measures.
Tree shaking: Leaving unneeded parts of a set of modules out of the output.
TTL in DNS: The DNS TTL limits normal caching of an answer. A shorter new TTL does not retroactively shorten the validity of old answers already cached.
TXT record: A TXT record stores text in DNS. The relevant procedure determines how a service interprets it, for example when checking control of a domain.
TypeScript: TypeScript adds a static type system to JavaScript. It assists development but does not replace checks on actual incoming data.
Typing in programming: How a programming language handles kinds of values and suitable operations.

U

UI component: A UI component is a reusable part of a user interface. Its purpose determines the presentation and interaction behaviour it needs.
Underfitting: A model fails to capture the relationships needed for its task adequately and therefore produces weak predictions.
Unique key: A unique key constrains values or combinations so they cannot occur more than once within the defined data scope.
Unit test: A focused check of a small, defined part of software, such as a rule or function.
Unsupervised learning: A learning method that looks for patterns in data without predefined target answers for the particular training task.
Upselling: Upselling offers a more expensive alternative. In software, plan comparison, confirmed changes and actual enabled limits must agree.
Usability: Usability concerns use by particular people for particular tasks and conditions. An attractive interface alone is insufficient evidence.
Usability test: A usability test observes suitable people performing realistic tasks. It examines use and needs a clear scope and well-documented notes.
User Datagram Protocol: UDP transfers individual datagrams without its own guarantee of delivery or order. A protocol built on top must provide any required reliability.
User experience: User experience extends beyond individual screens. A portal example distinguishes the interface from the experience of the wider service.
User interface: A user interface enables interaction with a system. Labels, feedback and responses to actions matter alongside the visual layout.
User research: User research investigates tasks, needs and conditions of use. Observations and their interpretation must be kept clearly separate.
User story: A short description of a user role, a need and the goal behind it.
UTM parameters: UTM parameters label a campaign link with information such as source, medium and campaign.

V

Validation data: Data used to assess and select model variants during development.
Vector database: A vector database stores and searches vector representations. A similar match is not automatically a valid or permitted answer source.
Version control: A system that records changes to tracked files and makes earlier states comparable.
Virtual machine: A server VM runs a guest operating system. The operating model must clarify who handles the application, access and maintenance.
Virtual Private Network: A VPN connects participants logically over existing networks. Protected connections, permitted routes and rights in target applications need separate planning.
Virtualisation: Virtualisation provides software with a simulated environment. For servers, a hypervisor allocates physical resources to separate virtual machines.

W

WCAG: WCAG provides testable criteria for accessible web content. A conformance claim needs a defined version, level and scope.
Web application: A web application supports tasks using web technologies, often through a browser. Installation, offline operation and server connections depend on its scope.
Web Application Firewall: A WAF checks web requests against rules. It can filter unwanted traffic, but needs suitable settings and does not replace fixing errors in code.
WebAuthn: WebAuthn is the browser API for public-key sign-in. The server must validate the browser response before accepting authentication.
Webhook: A webhook sends an HTTP notification to another system when a specified event occurs.
Webhook replay: Webhook replay means sending a webhook notification again. Deliberate recovery and replay attacks require different checks.
Wireframe: A wireframe shows the structure of a page or screen. It supports discussion of content placement without proving a complete interaction flow.

X

XML sitemap: An XML sitemap gives search engines selected URLs from a website and may include additional information about them.

Z

Zero trust: Zero trust does not derive trust solely from network location. Resource access needs appropriate identity and permission checks.

#

3-D Secure: 3-D Secure supports authentication for online card payments. The flow may proceed without an extra prompt or require confirmation.

Ready for the next step?

Tell us what you have in mind. In a free initial consultation, we clarify the goal, the effort and the first step.

Request a quote Get in touch

Free initial consultation
No obligation
NDA on request