At Sureify, I built a responsive, flexible component library which lets designers drag-and-drop bespoke charts into their screens. Charts are one-to-one with nivo.rocks, vastly improving dev hand-off. ###### Advised by Arjan van der Vlies, Anders Houston, Patrick Schaenzel. --- Problem statement ## A library by designers, for designers Our design team needed a way to present data consistently throughout its product lineup. Up until this point, designers had been using placeholders, mocking up ad-hoc charts. But their unrefined nature made it hard to communicate to stakeholders, as they would get tripped up on the small details. > *“We need a figma library that enables us to consistently design how data is communicated.”* > *- Anders Houston* ![drag-and-drop](https://schultzdavidg-portfolio.s3.us-west-1.amazonaws.com/images/dv/dv-drag_drop.png) --- User needs ## Understanding my stakeholders The first thing to get me on the right path was understanding who my stakeholders were, and what they needed. ![component controls](https://schultzdavidg-portfolio.s3.us-west-1.amazonaws.com/images/dv/dv-component_controls.png) ### → Designers As primary users, designers needed to be able to insert a chart in their design, and adjust it to their use-case. This meant the library needed to be: - Intuitive (easy to use out-of-the-box) - Responsive (flex into different viewports) - Modifiable (sense of control over the tool) ### → Developers As secondary users, developers needed to be able to easily translate designs to code. Discussions led us to nivo.rocks, an open-source React library for charts. This library serves as the foundation for the one in Figma. ![nivo.rocks screenshot](https://schultzdavidg-portfolio.s3.us-west-1.amazonaws.com/images/dv/dv-nivo.png) --- Implementation ## Breaking the limits of Figma Figma, for all its strengths, is not exactly equipped for responsive data visualization. The core problem: **layering**. Though auto-layout is powerful, it’s modeled after CSS flexbox—which fundamentally does not support layered elements. This means, to achieve responsiveness, I needed to find a way to combine auto-layout with constraints. ![line chart anatomy](https://schultzdavidg-portfolio.s3.us-west-1.amazonaws.com/images/dv/dv-chart_anatomy-line.png) ### Nailing the alignment Even after figuring out a way to stack layers, how do you keep your data, grid, and labels aligned? In a line/area chart, the grid extends to the edges of the container. This is because points don’t occupy horizontal space. Bar charts, on the otherhand, *do* fill up space—which means the grid needs to align itself by The solution I found was to force sub-components to maintain the same dimensions at all times. This can be done by putting them in a constrained noclip frame. ![example of poor grid alignment](https://schultzdavidg-portfolio.s3.us-west-1.amazonaws.com/images/dv/dv-grid_alignment.png) ### Variants vs. direct-selection There are three forms of modification to the chart components: - **Direct selection**, where a user can select an element and modify it’s color, text content, etc. - **Component controls**, which are primarily used to modify large structural changes (e.g. visible axes, number of bar segments, etc.). - **Show/hide** a given element (e.g. number of bars, every other gridline, etc.). ### Naming conventions & documentation With many different layers and sub-components in a chart, users need some way to differentiate between what they “can” and “cannot” touch. - **PascalCase** (capital letters) indicate a component that is intended to be editable. - **kebab-case** (dashes) indicate components which should not be touched. These are the “under-the-hood changes”. ![screenshot of documentation](https://schultzdavidg-portfolio.s3.us-west-1.amazonaws.com/images/dv/dv-docs.png) Additionally, I wrote extensive documentation. It covered topics such as: - **Component architecture;** components get complex, fast. I wrote two pages here: one outlining how to manipulate components, and one to guide further updates. - **Visual encoding;** the first question when visualizing data is, what kind of chart should we use? To answer this, we should consider data types: amounts, proportions, distributions, xy relationships, etc. - **Color theory;** is it more important that colors are pretty, or that they are informative? We settled on the latter, emphasizing the use of sequential, categorical, and diverging data sets. - **Dev hand-off;** how does a chart go from conceptualization to production? As this question is relevant to designer, developer, and product stakeholders, this page is written to address each perspective. ### Presets make it easy For most use-cases, a user doesn’t need full control over the shape of their data. By establishing presets for: the shape of data; labels + styles; date ranges + density—I could hide complexity under-the-hood—decreasing cognitive overload. --- Reflections ## Things I'd do differently I’m very proud of the system I developed. I was a junior at this time, and this project forced me to progress my technical skillset through a lot of trial and error. But there are a few things I’d do differently, and they all coalesce around focusing my time into things that actually matter. The responsiveness and flexibility I achieved was met with a lot of positive reception, but in practice I observed that designers would often just... break the component, and apply changes manually. They didn’t need all the bells and whistles I provided! They just needed a foundational structure. So this project set the groundwork for two important skills: resilience and creativity towards technical problems, and a stronger appreciation for how I align my energy with what my team needs.

david schultz

david schultz