The language of real-time data.

Plain, practitioner definitions across real-time, alternative, market, and satellite data - and machine learning.

A working vocabulary for desks and teams that trade and train on data: what each term means, why it matters, and how it shows up in real-time data. Every definition links to the terms it depends on, so you can follow an idea rather than look up one word at a time.

The vocabulary is where this industry hides its ambiguity.

Every data vendor says real time, and almost none of them mean the same interval. The same is true of freshness, provenance, coverage, and quality: they are all words a buyer is expected to interpret generously. A glossary is the cheapest form of honesty available to a vendor, because a definition written down can be held against the product later.

So these are practitioner definitions rather than marketing ones. Where a term has a precise meaning, we give it: event time is not processing time, point-in-time history is not a corrected archive, and an at-least-once delivery guarantee is not exactly-once no matter how it is described in a datasheet. Where a term is genuinely contested, we say that too.

Each definition links to the terms it depends on, which is what makes this usable rather than alphabetical. A reader who arrives at latency can follow it to round-trip time, to event time, and to what a backtest needs from both, without leaving to search for the next word.

How a definition connects4 pages

The term you landed on

Real-Time Data

Data delivered the moment it is observed, with minimal delay between an event in the world and its arrival at the systems that act on it.

Linked from inside the definition

Data Latency

The delay between an event occurring and the corresponding data being usable - the core quality metric of any real-time dataset.

Streaming API

An interface that pushes new records to consumers the instant they arrive, rather than making them poll for updates.

Data Freshness

How recently a dataset was updated relative to the world it describes - closely related to latency, but measured as staleness rather than delay.

Every definition links to the terms it depends on, so a reader who arrived for one word can follow the argument instead of leaving to search for the next one.

Three ways in.

You met a word and want it defined

Use site search or the category sections below. Every definition opens with a one-line answer before it explains anything.

You are evaluating a data vendor

Read event time, delivery guarantees, point-in-time, and data provenance. Those four decide whether a dataset can be trusted in research.

You are reviewing us on security

Start at post-quantum cryptography and hybrid key exchange, then the post-quantum page for what our transport negotiates.

Real-Time Data

Data delivered the moment it is observed, with minimal delay between an event in the world and its arrival at the systems that act on it.

Data Latency

The delay between an event occurring and the corresponding data being usable - the core quality metric of any real-time dataset.

Data Freshness

How recently a dataset was updated relative to the world it describes - closely related to latency, but measured as staleness rather than delay.

Time-Series Data

A sequence of observations indexed by time - the native shape of prices, sensor readings, and most real-time data.

Data Provenance

The documented origin and processing history of a record - where it came from, when it was collected, and how it was transformed.

Event Time vs Processing Time

The difference between when something happened and when a system got around to handling it - the distinction every real-time claim rests on.

Watermark

A marker in a stream stating the point up to which data is believed complete, so downstream consumers know when a window can be closed.

Tail Latency

The slow end of a latency distribution - the 99th percentile and beyond - which is what actually breaks a strategy or a pipeline.

Clock Synchronization

Keeping distributed machines agreed on the time, so timestamps from different regions can be compared and ordered.

Edge & Delivery

Edge Computing

Running computation close to where data is produced or consumed, instead of sending everything to a distant central region first.

Edge Node

A machine positioned close to where events surface, holding connections open so an event is received rather than fetched later.

Point of Presence (PoP)

A network location where a provider hands data off to a client's systems - the last hop, and usually the one that decides delivery latency.

Event-Driven Architecture

A system design where components react to events as they are published, rather than asking each other for updates on a schedule.

Round-Trip Time (RTT)

How long a packet takes to reach a destination and return - the floor under any request-and-response interaction.

Delivery Guarantees

What a stream promises about each record reaching you: at most once, at least once, or effectively exactly once.

Alternative Data

Non-traditional datasets - satellite imagery, web activity, card transactions, geolocation - used to form a view before it reaches the mainstream.

Web Scraping

Automated collection of publicly available web data at scale - pricing, listings, hiring, reviews - turned into structured datasets.

Consumer Transaction Data

Anonymized, aggregated card-spend and receipt panels used to estimate a company's revenue before it reports.

Geolocation Data

Aggregated location signals - foot traffic, dwell time, movement - used as a physical-world proxy for demand and activity.

Nowcasting

Estimating the present or very near term - this quarter's revenue, this month's activity - from high-frequency data, before the official number.

Sentiment Analysis

Extracting tone or polarity from text - news, filings, social posts - to turn unstructured language into a numeric signal.

Data Licensing

The terms under which a dataset may be used - what you may derive from it, redistribute, train on, and keep after the contract ends.

Market Data

Tick Data

The most granular market data - every individual trade and quote, timestamped to the microsecond or finer.

Order Book Data (Level 2)

The full ladder of resting bids and offers at each price level - market depth beyond the single best quote.

Market Microstructure

How price and liquidity actually form from the mechanics of orders, spreads, and trading - the physics beneath the tape.

Point-in-Time Data

Data that preserves exactly what was known at each moment in history, with no later revisions leaking backward.

Survivorship Bias

The distortion from studying only the entities that survived - dropping delisted and bankrupt names inflates historical results.

Look-Ahead Bias

Using information in a historical test that nobody could have had at the time - the single most expensive error in quantitative research.

Backtesting

Evaluating a strategy or signal against historical data to estimate how it would have performed before risking capital.

Alpha

The excess return of a strategy above its benchmark - the edge attributable to skill or information rather than market exposure.

Machine Learning

Signal

The predictive, actionable information extracted from noisy raw data - what remains after the noise is removed.

Ground Truth

A direct, verified observation of the real world, used as the reference against which models and estimates are checked.

Training Data

The dataset a machine-learning model learns from - and usually the single biggest lever on how well the model performs.

Natural Language Processing (NLP)

Techniques for turning unstructured text - news, filings, transcripts - into structured, machine-usable data.

Named Entity Recognition (NER)

Identifying and tagging the real-world entities in text - companies, tickers, people, places - so unstructured content becomes linkable.

Entity Resolution

Linking records that refer to the same real-world entity across different sources, spellings, and identifiers.

Model Drift

The gradual decay of a model's accuracy as the world shifts away from the data it was trained on.

Signal-to-Noise Ratio

How much of a dataset carries information relative to how much is clutter - the measure that decides whether volume is an asset or a cost.

Retrieval-Augmented Generation (RAG)

Giving a language model current, sourced material at query time instead of relying on what it memorized during training.

Model Grounding

Anchoring a model's output to verifiable external facts, so what it produces can be checked rather than only believed.

Feature Engineering

Turning raw data into the inputs a model actually learns from - the step where most of a model's real performance is decided.

Data Engineering

Data Pipeline

The chain of steps that moves data from source to consumer - ingest, transform, and deliver - reliably and on schedule.

Streaming API

An interface that pushes new records to consumers the instant they arrive, rather than making them poll for updates.

Webhook

A server-to-server callback that pushes data to a URL you register when an event fires - push delivery without a held-open connection.

Data Normalization

Reconciling data from many sources into one consistent schema, units, timezone, and set of identifiers.

Deduplication

Removing duplicate and near-duplicate records so each real event is counted once, not many times.

Data Quality

The fitness of a dataset for use, across accuracy, completeness, timeliness, and consistency - for a data vendor, it is the product.

Data Lineage

The traceable map of where data came from, how it was transformed, and where it flowed - the backbone of debugging and audit.

Backfill

Populating a dataset's history - loading past data for a new or changed source - while preserving point-in-time correctness.

REST API

A request-and-response HTTP interface - the standard way to pull a defined range of history, as opposed to receiving records as they happen.

API Key

A credential identifying the caller on every request - and, handled properly, the boundary of what that caller can reach.

Idempotency

The property that handling the same record twice produces the same result as handling it once - what makes retries safe.

Backpressure

A consumer telling a producer to slow down - the mechanism that makes a stream degrade predictably instead of collapsing under a burst.

Data Observability

Monitoring the data itself - freshness, volume, distribution, schema - so a source failing silently is detected before a consumer acts on it.

Schema Evolution

Changing the shape of a dataset over time without breaking the systems already reading it.

Data Governance

The rules and records covering who may use a dataset, for what, and how its handling can be demonstrated later.

Security & Transport

Transport Layer Security (TLS)

The protocol that encrypts data in transit and authenticates the server - the layer every API call and stream rides on.

Post-Quantum Cryptography

Cryptography built on mathematics that a large quantum computer is not known to break - the replacement for today's key exchange and signatures.

Harvest Now, Decrypt Later

The attack of recording encrypted traffic today in order to decrypt it years from now, once the cryptography has been broken.

ML-KEM (FIPS 203)

The NIST-standardized post-quantum key encapsulation mechanism, based on module lattices - the replacement for classic key exchange.

Hybrid Key Exchange

Combining a classic and a post-quantum key exchange in one session, so the session is only broken by an adversary who defeats both.

Traffic Analysis

Learning from the pattern of encrypted communication - who talks to whom, when, and how much - without reading any of the content.

Forward Secrecy

The property that compromising a long-term key does not expose past sessions, because each session's keys were ephemeral.

Geospatial & Satellite

Satellite Imagery

Overhead images of the Earth captured from orbit, turned into structured measurements of physical activity.

Remote Sensing

Measuring an object or area without physical contact - typically from satellites or aircraft, across many parts of the spectrum.

Geospatial Data

Data tied to a location on the Earth's surface, enabling analysis by place and joins across otherwise unrelated sources.

NDVI (Vegetation Index)

A remote-sensing index that measures vegetation health from red and near-infrared light - a proxy for crop condition and yield.

Synthetic Aperture Radar (SAR)

An active radar imaging technique that sees through cloud and darkness - useful where optical satellites are blind.

AIS Vessel Tracking

Ship-position data from marine transponders, used to track trade flows, port congestion, and commodity movements in near real time.

Real-time data, at edge speed.

A live evaluation measures EdgeOrigin against your coverage requirements: decision-ready real-time data, delivery latency into your systems, and record-level provenance.

Book a demo Contact sales