Dojo

A web app for office layout design and analytics built using Javascript, D3.js and Python. The app enables office tenants to evaluate their existing office spaces, gain insights about what is working well and what isn't, and get recommendations about changes they can implement to make their office work better.

The future of office real estate

As the owner of the World Trade Center and several other large commercial buildings in New York City, Silverstein is one of the biggest players in the city’s commercial office market, leasing space to many of the country’s most innovative and fastest-growing companies.

Despite the high quality of their buildings and their excellent reputation in the industry, Silverstein's traditional leasing business model is now at risk from emerging business models such as co-working which are changing the expectations of Silverstein's clients toward their office space.

Screenshot of Dojo Explore interface which filters a large set of possible floorplan layouts according to a defined set of metrics

Measuring performance

Dojo gives Silverstein's tenants the upper hand by giving them real insights into how their space is being used, and suggestions for actions that can be taken to improve the space's efficiency and utility for the employees.

Dojo's UI gives office managers and HR personnel a quick view of employees and where they are seated in the office. But it goes way beyond seating charts. The app's backend uses custom spatial analytics algorithms built in Python to analyze the space's use patterns, predict areas of congestion, and recommend seating arrangements based on the existing team and communication structures in the office.

Screenshot of Dojo Plan interface which allows users to manage their office space, including assigning individual employees to specific work spaces in the office, and evaluating the effects those changes have on the performance of the office.

Catalyst for change

By combining tools for day to day office management with powerful algorithms to generate and evaluate a large variety of layout options, Dojo gives office managers and HR professionals a powerful new platform for managing their spaces now and into the future.

Product architecture

Dojo's backend is powered by a combination of spatial and organizational algorithms written in Python that are used together to analyze existing workspaces and make suggestions for changes to make the workspace better. The app's UI is written in Javascript and uses the D3.js library to render 2d floor plans and other graphics.

Towards spatial intelligence

The key to Dojo's ability to plan and analyze spatial layouts is a catalog of components which acts as a set of standard building blocks that can be used to create and manipulate the spatial arrangement of various parts of the office.You can think of this as defining a programming language for office spaces where the catalog components are the letters and words and the algorithms dictate the grammar that gives them meaning.

Programs within an office are categorized across two axes: Open vs. Enclosed which defines the architectural setting and Individual vs. Collaborative which defines its furnishing and use

Based on these categories, 8 unique spatial typologies are defined which set standards for size of spaces, how they can be arranged, and which spaces can be interchanged with others

These spatial blocks can be used to quickly lay out any given office space and can be swapped to easily explore different layout options

Computing space

The foundation to most of the spatial analytics behind Dojo is the discretization of the space into a finite set of triangular elements through a dynamic remeshing and relaxation process. To create the mesh, all walls and other edges are first divided into points at a regular interval (for example 1'-0"). These points form the initial vertices of the mesh, which is then triangulated using a technique like Delaunay. Then, the mesh is iteratively optimized by adding and moving vertices (except for the original ones which are fixed) while targeting mesh faces of equal area and with equal length edges.

Triangulated base mesh for sample floor plan

A circulation simulation run by tracing shortest paths through connected faces of the base mesh using Dijkstra

Circulation and congestion heatmaps can be generated by counting the number of shortest paths going through each mesh face and then smoothing the values across neighbor faces. See this paper for a related technique.

The base mesh provides a regular discrete sampling of the space which can be used to measure a variety of different properties such as traffic, daylight, and views to outside