A time series analysis API

Time Door is an API for statistical insights into time series data

Matrix Profile
2019-12-31 16:00:00
2019-12-31 16:16:39

Matrix Profile computed on 1000 data points with the Time Door API in 1047.26ms.

Insights into change

Time Door makes it easy to integrate advanced time series analysis capabilities into your systems by providing an accessible, well-documented API.

Time series data

if missing values

Missing value imputation

if data assumptions

Data transformation

Statistical computation



Applied arguments

if missing values

Imputed values

if data assumptions

Applied data transformations

Statistical computation results

Each response contains the applied arguments, every imputed missing value and every data transformation – enabling full reproducibility of the results.

Designed for real-world data

Time series data is often incomplete, non-stationary, and fast. Time Door is designed to handle all these properties:

  • Every endpoint provides advanced value imputation methods to replace missing values
  • Endpoints with data assumptions provide (automatic) data transformations to deliver valid results
  • Endpoints that compute moving statistics can process hundreds of new data points per second, making them ideal for fast data streams
  "data": {
    "collective_anomalies": [
        "t": "2020-01-01T00:00:15Z",
        "mean_change_strength": 2.876,
        "variance_change_strength": 0.1659,
        "size": 10
    "point_anomalies": []

Response snippet of the Collective And Point Anomalies endpoint. One collective anomaly with a size of 10 data points starting at t = 2020-01-01T00:00:15Z has been detected.

Algorithmic variety

Time Door provides an expanding set of time series analysis algorithms for:

  • detecting anomalies
  • finding recurring patterns
  • discovering how recurring patterns change over time
  • identifying early warning signals of critical transitions
  • finding serial dependencies
  • spectral analysis

Adaptive usage pricing

Test our API with 100K ms of free computation time and pay only for what you use afterwards.