Hugs Method post

Every wonder how much you hug? Or how often your friends hug? A little curiosity can go a long way…

Data Collection

I don’t know of any sources for hugging data, so I had to start by collecting hug data. Knowing that my own dataset wouldn’t be too interesting, I wondered what my friend’s hugging patterns were. In particular, I have quite a few married friends, so I thought it’d be interesting to compare couples hugging habits. Specifically, who hugs more, and at what time of day?

I designed a hug tracking sheet which I then gave hard copies to nineteen couples (including my husband and I) and asked them to each track their hugs for three days. I’ll refer to these couples as “respondents” from here on out. Maximizing ease of use was important, so I designed the sheet in a such a way that it could be folded down to the size of a phone.

A clip of the survey I designed for couples to track their hugs

There were several data points that I was interested in collecting: time of hug, the gender of the person who was hugged, their relationship to the respondent, and how many times they hugged that person in that hour.

I also asked the couples to disclose how long they had been married, how many children they had, and their age.

My hypothesis was that spouses hug each other most often in the morning and evening. I also guessed that females would hug more than males, and couples with children would hug more than couples without children.

After three days I collected the survey sheets from the respondents, and spent my Saturday morning inputting data (all 576 rows of it) into a GoogleSheet.

Data Exploration

I started by sketching my ideas and experimenting with different ways to display the information.

Sketches for how to visualize the hug data
More sketches. I decided to explore a histogram-like visualization (left image) for the hug data.

While sketching, I would sometimes get distracted with ideas or trying to include too much data all at once. I kept going back to my hypotheses and the root of this project, which was to compare within a couple who hugs more and at what time. Having that vision helped me create a design that optimizes those aspects of the dataset.

Once I had the data entered in a spreadsheet, I was able to explore in Tableau. It was helpful to see what my data looked like, and I was able to better understand some unusual data points.

Data exploration in Tableau

After exploring, I translated my sketches into Illustrator, to experiment with how I wanted the visualization to look.

First draft of visualization

This version was confusing. There were too many lines and floating symbols. I realized that I was failing to represent the volume of hugs: I needed to use density to convey the quantity. I decided to replace the dots with bars to create something more akin to a histogram, with the relationship-type symbols embedded inside.

Building the frameworks

In Processing, I started by trying to build the different pieces of my visualization separately. I’m still learning to code, so this was tricky. I found that I would have to go back and forth between my spreadsheet and code to convert or massage data.

It took a good deal of math and coding cleverness before I was able to have the computer draw what I wanted. There were plenty of (interesting) glitches along the way.

Nearly there…

Eventually—after experimentation and algebra—I was able to draw my data the way I wanted to. I exported the charts from Processing and tightened them up in Illustrator. Still concerned about readability, I user tested my visualization with several peers.

Aggregate of the hug data
An example of a couple’s hug dataset

In the feedback I collected, I saw that people were struggling to understand how to read the data. There were a few aspects of my design that looked slick, but ended up being distracting and hindering understanding.

In particular I really loved the aesthetic of the overlapping bars. The purpose of the chart was to show when all of the couples I surveyed hugged the most. While the transparency does have some “hot spots” of where people hugged more often, it was too difficult to read and didn’t reveal the true density of hugs.

For the individual datasets, the time tick marks were hard too see and distracting. There were also some inconsistencies: a small stroke was used to split hugs within the same gender (color), but not used between the different genders. The key on the left—intended to orient viewers to understand the quantity of hugs and the which spouses data they were looking at—was also ambiguous and confusing.

Final result!

Improving the visualization from here wasn’t too difficult. For the aggregate data, I stratified the data points instead of having them overlap. The true form of the hug data emerged, which turned out to be more interesting than I expected.

On both the aggregate and couples datasets I severely reduced the timeline. The quantity of spouse hugs was what I wanted to highlight, with time as a secondary element. So the time reveals itself when moused-over. I found this to significantly reduce cognitive load, while preserving the option to drill into the data.

As a speedy way to display all nineteen couples datasets, I made a gif that quickly flips through all of the couples charts.

Final visualization of the aggregate hug data
A couple’s dataset, with a rollover revealing the time and details of a specific datapoint
All nineteen couples datasets

Additional charts

With such a rich dataset, I wasn’t quite done. I created a few other charts to supplement the main visualization that I had so far. This semester I enrolled in a statistics class, so I wanted to use my newfound knowledge of boxplots! I drew the boxplot and scatterplots in Processing, and tidied them up a bit in Illustrator. This was the quickest part of the whole process, which was nice.

What I would do differently

This project was a great learning experience. It was my first deep dive into data, and I had a lot of fun! That being said, there are a few things that I learned along the way and would do differently.

  • Collect data digitally. Next time, ask respondents to fill out a form online or input their own data. I spent several hours inputing the handwritten data into a spreadsheet. Probably was my least favorite part of the process.
  • Leave more comments more in my code. Coming back to the project after not working on it for a few days was sometimes difficult because I would get lost in my code. I have a lot of room for improvement here, and want to work on having neater and more organized code.
  • Don’t design too early in the process. It was hard to avoid starting with design. I had visions of what I wanted this to look like in my head, and sometimes I would waste time in Illustrator rather than coding it in Processing. I need to be better at planning and making sure my charts are meeting my project goals.

See the full project on my Behance portfolio.