Spanning 6,362 days of data, covering seven locations and six weather metrics, this project transformed the windows all around our studio into a large-scale data visualization. Each of the 31 windows features a unique composition of generative tile patterns. Together, they narrate the environmental and organizational story of C°F. While grounded in pure data, the final output is a visual experience, structured, layered, and open to interpretation.

The underlying idea

The basic idea was to create a timeline spanning the studio’s history from its founding 17 years ago to the exact day we moved into the new studio. This timeline would begin at one end of the building, wrap around it chronologically, and conclude at the final window. We incorporated weather data from cities connected to our studio: de Bilt, where it was founded; Utrecht, our current location; and Chicago, the location of our satellite office. On the timeline we marked significant events for the studio such as key projects or awards.

In terms of the design, the ambition was to create something aesthetically pleasing and intriguing that provided hints of the underlying data. Once explained, it should be easy to comprehend, but it was not meant to be a literal data visualization that could be read directly.

Approach

Our challenge was to transform an enormous dataset into a cohesive, meaningful, and attractive visual piece. We worked with temperature, wind, sunlight duration, shortwave radiance, rainfall, and snowfall, collected from cities as different as Utrecht and Dubai. Our goal was to represent this temporal and spatial spread in a way that would spark curiosity at a glance but also reward closer inspection.

The initial concept was built around traditional tile systems. We imagined using standard tile sets to create patterns that, when repeated and arranged, would form a larger composition. But as we began prototyping, we quickly saw the limitations of this approach. The results felt static, overly repetitive, and disconnected from the nuances in the data. In addition, it was difficult to grow the number of cities in the data over time; we needed something more expressive.

One of the initial standard tile approaches

We continued to use individual tiles as the building blocks, but we started challenging the way in which they were composed. We explored layering, stacking, and rotating shapes in ways that broke the uniformity of conventional patterns. Using basic geometric forms gave us the flexibility to experiment with overlaps and visual density and allowed new figures to emerge naturally from the system. It needed to be bold enough to create clear rhythms when viewed across the facade, but detailed enough to spark curiosity when standing right in front of it.

Rather than plotting every data point on a line or bar, we used a generative system that treated each tile as a unit of time. Every tile represents a specific day, location, and metric. No two are exactly alike. When these tiles are placed across the window grid, they form patterns that feel rhythmic, organic, and structured.

The result was not just a timeline, but a series of line charts with each layer corresponding to a weather metric. As tiles accumulate, they create a stacked visual flow across the surface. Because this system naturally introduced a chart-like logic (time on the x-axis, value on the y-axis), we introduced a set structure for the tiles. Each metric, in each city, was assigned a unique tile shape. This approach allowed us to embed both visual identity and structural consistency within the composition. When the tiles are placed according to their normalized values, they maintain visual coherence while making the data legible and distinctive across both time and location.

These patterns act as both charts and timelines, displaying seasonal variation, climate anomalies, and long-term trends. When viewed in full, the composition evokes motion, flow, and the passage of time. It becomes a dynamic data portrait of both the environment and our agency.

Mapping of the windows and planning of the pattern

Layers and logic

The base layer of the system is built from normalized data, scaled between 0 and 1 for each metric, where 0 represents the lowest value across all days for all cities, and 1 represents the highest. This allowed us to treat diverse variables, like intense heat in Dubai and sunlight in Copenhagen, with equal visual weight.

This created a grid with time running horizontally, and metric values running vertically. To make cells of suitable dimensions, multiple days (for example, July 2, 3, 4, and 5) and a range of values (for example, between 0.20 and 0.24) were combined into a single cell. When determining what to plot, all values that fall within this range for these days were selected. From this selection, we chose the metric that experienced the most significant change compared to the day before. This approach effectively focused on change, which is often the most memorable aspect of weather.

Test prints checking different parameters

On this foundation we layered moments of significance. Historic weather records were marked using a special set of textures that stand out visually. Agency milestones were also highlighted, tied to exact dates in the dataset. These tiles are three times larger than standard ones, creating clear focal points in the pattern’s flow and adjusting surrounding tile density to accommodate them.

The result was a layered data landscape: some windows feel calm and rhythmic, while others appear more chaotic and dense. These variations reflect not just climate conditions but also the evolution of our agency through time.

Grid adjustment is visible by simply adjusting the size of the bigger tiles

Tools and process

To get there, we built our own tools. We started with the Weather Data Explorer, a browser-based tool using the Open Meteo API and p5.js. It allowed us to explore values, extremes, and trends across cities and years. We could zoom in on specific days, or zoom out to see patterns over time.

Next came the Grid Explorer, a layout tool that let us prototype how tiles would behave at real-world window dimensions. We used it to test spacing, rhythm, and layering. We developed the next iteration of the tool, Panel Maker, to test and refine the sets of tiles we needed more effectively. By exporting SVGs, we could quickly move into Figma and collaborate with the wider team on design refinements.

Overview of the used tools during the project

In-between coding and Figma, sketching played a big part in team alignment.

In early prototyping, we generated every tile programmatically. Later, we shifted to designing tiles as SVGs, which we then integrated into our system. This hybrid approach gave us more control over visual quality while still maintaining the flexibility of a generative system. Once finalized, the tiles were mapped to the grid system and prepared for physical production.

To move beyond the screen and get an impression of how the patterns would look when applied, we created test prints for multiple windows. This helped us find the right scale and detailing for the patterns, as well as determine the exact positioning for elements like our logo, and a legend on the front door. Additionally, we did a small test print on the actual foil to assess print quality and the effect of transparency on the design.

Test print in foil

Final tile set used in the Window Panel Maker

Visual language

The design system borrows heavily from C°F’s own visual identity, which prides itself on being geometric, mostly monochromatic, and having a minimal approach to its boldness. We worked with basic, stable forms (circles, lines, arcs, etc.) that reflect clarity and structure, while challenging the constraints of traditional tile systems. Instead of repeating static units, we let complexity emerge from the stacking and layering of different metric shapes.

Visual detail explorations

At first glance, the installation reads as a single flowing composition. Up close, it reveals micro-patterns, overlaps, and tiny anomalies. These artefacts, which are unexpected shapes formed by data collisions, became some of our favorite features. They reflect the messy, noisy, beautiful reality of weather data and make the piece feel more dynamic.

System and pattern explorations

Detailed view of different shapes and textures

Reflection

This initiative was more than a design piece for us; it acted as a celebration, a portrait, and a wayfinding system. We wanted to create something beautiful, but also something that would spark questions, start conversations, and connect the viewer to both place and time.

In this way, the windows became a stage where data history, agency milestones, and climate trends intersect. As with the weather itself, the patterns are always shifting, always open to interpretation, always layered with meaning.

Final pattern installed on the windows