Data Stories - Interview of Pei Ying Loh

Peiying: Thank you, Tom, and thanks for having me here. My name is Peiying, and I'm the head and co-founder of Kontinentalist. It's a data-driven editorial studio based in Singapore, with a focus on Asian culture and trends. We're committed to using data to drive our content and change people's minds about how they talk and see Asia. We feel that data is something that delights and surprises people. So that's where we are at the moment. We publish a number of stories online, so feel free to check it out.

Tom: On that note, I've written a "Getting Started" free guide resource for Buried Signals, and the first place that I sent people to aside from the Pudding is Kontinentalist.

Peiying: Oh great, thank you so much. Honestly, it's crazy that people even talk about us and put us in the same vein or sentence (as the Pudding) these days. When we started out, we really looked up to the Pudding. I think we've read all of their tutorial notes, guidelines, and editorial guide to figure out how we can create something like this for Asian content.

Tom: Your team has a unique style that's very friendly and accessible. You really dive deep into a story, which I think is unique to you. Can you tell me a little bit about your personal approach, or Kontinentalist's approach, to visual storytelling on the web?

Peiying: Yes, I can. I'll talk a bit about my personal approach first, and then how it extends into how we practice visual storytelling at Kontinentalist. Culture, history, and heritage are big topics that I personally gravitate towards. Having worked at the Asian Civilisations Museum in Singapore for several years, I've learned and studied a lot about Eurocentrism and Orientalism. The gaze we cast upon Asian content or even just talking about current affairs in Asia has been skewed in the media landscape. For example, when we talk about Japan, we tend to focus on hyper-exoticization and hyper-technology. And when we talk about Indonesian culture, we tend to describe it in traditional terms or as an exotic travel destination. These are all pigeonholing Asian culture to some extent, and are often from a Western perspective. This also affects how Asians view themselves. At Kontinentalist, we aim to change the way we talk and describe Asian cultures and societies. We have a detailed editorial guide that avoids using certain tropes or cliches. Personally, I enjoy producing visual stories that delight people and surprise them, giving them a sense of joy and representation.

Tom: That's fantastic. Could you tell me about your workflow for developing a story once you have an idea?

Peiying: We started out like any other news or media publication with a linear workflow. We had a writer who searched for data, did research, and pulled together a story. We also had designers and developers who helped take those ideas to fruition and bring the story to life. However, in the last couple of years, some team members found this process limiting and felt excluded from it, especially designers and developers who couldn't participate in the ideation stages of a project. So, we completely switched things up and designed our entire workflow through a design sprint. The entire company came together for a week to decide how we would work on stories together.

Our stories generally start from a pitch stage, and as with all publications, we have different periods for pitching ideas. We usually brainstorm our ideas for a year at the start of a year and spread them throughout the year. We leave some room for ideas to drop in, especially when current trends, current affairs develop, and we have some important markers. The story needs to be focused on Asia, and we have some umbrellas that it needs to cover. For example, Asian culture, society, social injustice, inequality, technological trends, infrastructure development, and climate change. We have general brackets on our website that cover these topics, and we only publish what we consider to be evergreen topics, meaning they can stand the test of time.

Our stories take a long time to develop, and usually, each story is in the pipeline for at least two to three months, sometimes even longer. Therefore, we avoid publishing stories that are just for the news cycle, such as those related to Covid. We ask people who pitch to us, or even our own team, to have a clear central question that they're asking or trying to answer, the data and resources that they're looking at to answer this question. If they adhere to these criteria, then we move to the next stage.

Unlike other newsrooms, we ensure that there's enough material to even start an idea on the story. We then use a little bit that you can launch on our slack channel to look for collaborators on the team. People will chime in on whether they want to work on a topic based on these parameters, and the editor will also give their approval. Then we bring in our designers and developers, and everyone brings in their own inspirations of what they want to implement within the story. We then ideate a rough copy for how the story would be approached. From there, everyone takes a little homework to start writing the first draft or doing wireframes, or testing out pieces of code to see if the concept has a new link to run on. We generally move towards the first round of edits and the second round of edits before publishing the story. This is the general lifecycle of our stories.

Tom: I just have one follow-up question. What do you mean by three resources when you say someone needs to have a central question that they answer and free resources?

Peiying: Yeah, so I think for us, the central question gives us an insight into what is the sort of angle of the story and the purpose of the question, or the purpose of the story. And when we refer to free resources, it can be quite scant, so for example, a database or a data set that they found online. This can be one or multiple datasets or even a website. For example, personally, I've pitched a story about K-dramas and for me, my data resource is actually like a fan site where I'm probably going to write a script and scrape together data. So as long as we know the data exists in some shape or form, it passes our criteria, and then the next step is just about how we're going to get the data. But we do need to know that it's even collectible in some shape or format or it's in a CSV file or even in a book. Sometimes we do source data from published books as well.

Tom: And that actually takes me to one of my questions, which is, do you have any sort of favorite tools and hacks that you use regularly or that you think have had a big impact on how you produce stories?

Peiying: Yeah, actually, personally for me, I don't consider myself to be extremely skilled at tools because I do not know how to code as well, and most of the people on the team, we sort of stumbled into data storytelling. We weren't trained by any means in information design or infographic design, or even web development, so we all kind of stumbled into it.

Which is why I don't think I have any hacks. If anything, I think I love learning things through mistakes or if while going through data I find that it's really becoming a chore, then I will try to Google for some shortcuts that I may use, like certain formulas I can write in Excel or use to try and shorten the process. That's where I usually discover my so-called hacks, but otherwise, our process is still pretty manual. And my favorite tools to sort of analyze my data are usually Flourish and Google Sheets.

A big tool we started to use very frequently when we first began is actually Mapbox. So we used a lot of Mapbox to visualize geographical data. That's like a big part of how we started, and when we first started, Mapbox had also, I think, just been around for a couple of years, maybe only one or two years, so even the tools had not been fully developed, so we really enjoyed the high degree of customization that we could get with Mapbox, being able to go in on a very granular level to different zoom levels to show people the landscape. And we felt that, also, like maps, give you a certain sense of empathy and relatability. Also, that sense of wonder and discovery. So Mapbox is a big tool for me, and I really like how in recent years, it's added a ton of features. So you even have, I think now you could change projection types. You can zoom all the way out and look at it as a globe. You can even add that and make it look like a very realistic 3D map, you know, and things like that.

Tom: It's interesting that most of your team stumbled into data storytelling and learned on the fly.

Peiying: Yes, about 80% of the team did. For example, our lead designer studied fashion in university but now is well-versed in UI and data storytelling. Our current editorial lead is a renowned playwright in Singapore, but decided to venture into data storytelling. Even our front-end developer who built most of our website's stories, studied Engineering Science and later did a Masters in Urban Science and Planning before discovering data storytelling through a course.

Tom: It's great that you empower people to learn and grow in the field. Do you think having people with completely different backgrounds affects your storytelling?

Peiying: Definitely. Sometimes, technical limitations can spark creativity. For example, we have a small number of developers, so we focus on visually stunning pieces that don't necessarily need to be interactive. We look up to the South China Morning Post, who creates stunning, well-executed graphics that appear interactive but are actually just graphics. We try to learn from them and use techniques that take less time.

Tom: It's also easier for more people to work on.

Peiying: Yeah, like a static graphic rather than an interactive. So I think a lot of these things influence what we decide to put in the story, and what sort of approach we take. We don't avoid coding things now, but we're more purposeful. We decide to add interactivity or some sort of engagement function to our stories.

Tom: So, diving deeper into the workflow and the process, let's say you have the first draft of the copy for the story, and it includes some visualization techniques that are intended to be used. It goes through feedback. How do you decide for this article what visualization techniques or interactions you're going to use? Are there some things that usually go through your mind, is there a checklist or a process?

Peiying: Yeah, that used to be decided when the writer submits their first draft and the editor looks through the proposed visualizations and assesses whether they make sense. But these days we actually have a reverse process. The writer takes their cue from what the team is proposing for the visuals. We're trying to make the process more collaborative, so even before we begin to put words on a page, we've really decided what sort of key visuals or key impact, and visualizations we would be putting into a story. The writer would put in the descriptors and content pieces that will bring those to life. The editor still makes a pass on proposed visualizations and assesses whether or not it's fruitful. Some general questions we ask ourselves when looking at these proposed visualizations first, do I understand what this visualization is trying to say? Can I understand it within the first thirty seconds of looking at it? The second one is we ask what is the one key message of the visualization? We actually find that writing alt text for visualization helps with this a lot, because we need to be able to describe through a screen reader what the visualization is trying to say to somebody who cannot read or see the visualization for themselves. So, essentially, if you have to distill your visualization to a single sentence, what would that be? And we find that provides a very clear picture of what would be the best visualization to achieve that message. Even if the intent or the purpose is a question, for example, if you want to invite exploration of the data, that also determines how we want to use the visualization. If it's an exploratory dataviz, we want them to draw their own conclusions, then that visualization would need to be a bit more visually complicated or a bit more interactive so that they can do a self-discovery element. We have that in one of our stories about the evolution of Chinese names. At the end, you can actually try your own name to see which generation you're in the dataset.

Another important question we ask is whether the visualization is the most sensitive or emotionally resonant way to present the data. For example, in a story on sexual violence and #MeToo in Singapore, we had a visualization where we wanted to describe the places where sexual assault survivors and perpetrators were, and we actually had data on that.

We initially had the idea of presenting the places where sexual assault survivors and perpetrators were through a word cloud visualization. However, we felt that this approach did not show enough respect or sensitivity to the topic. Instead, we reworked the visualization to be a tree map with illustrated visuals of the spaces without people in black and red tones to signify danger, emphasizing that even traditionally safe spaces like homes or schools could be places of danger for survivors.

Lastly, we think about which parts of the story are most complex and need to be visualized. If a particular aspect of the story is complex, then we want to find a way to visually represent it in a clear and easy-to-understand manner, so we write down what we believe the team would suggest as a visualization for this. Then we produce it, and then we test it internally or externally to see if within thirty seconds it's understandable. Then we try to gain feedback to adjust if necessary, and always make sure that it is a powerful visualization that can be distilled into one sentence.

Tom: and the checklist?

Peiying: We have a checklist at the end when we publish the story. It can be very technical, for example, ensuring that we have taken care of everything, such as meta descriptions, URL slug, and so on. One of the things we like to do is put a link to our story in a general chat with every team member. We invite everyone to look at it and give feedback, so we do implement a lot of last-minute fixes and changes. Obviously, it's to account for bugs, different screen sizes, and different browsers, but also, I think it's to get a sense of what people who did not work on this story think about it. We look at different teams, and it's not like every story is assessed by everyone in the team. Some people on our team may not have seen it before, so they would be looking at it with fresh eyes and be able to give us that feedback. Even for the editors, sometimes when you look at something for long enough, you forget that many people are reading this piece without the same context and understanding that you have.

Tom: Do you have an audit phase to measure whether a story was a success? Can you explain how you measure the success of a story?

Peiying: At the start, we struggled with assessing the impact of our stories because we're based in Singapore, a small country with a small population, and the media space is already saturated with different publications covering everyday news from various angles. Since our work takes a lot of time and we don't publish as often, our viewership is generally low. For a long time, we were stressed out by the low viewership and thought that fewer people encountering our pieces made our work less meaningful. However, we realized that this was the wrong mindset to have towards our pieces, and that the viewership wasn't as important as what people took away from a story. We changed the way we looked at our pieces and the metrics that we use to measure them. We monitor the number of people who come to our page on a regular basis, which is pretty constant. For us, we really care about read time, and we check that by looking at the average read time for our pieces. We generally aim for at least a four to five minute read time, which means that at least readers are engaging with some of the vision of a story. Some of our pieces are longer, up to 15-20 minutes, but we don't expect everyone to read everything to the end. We try to ensure that we punctuate enough visuals that readers get a sense of what the story is trying to capture. For us, a five-minute read time is satisfactory, but some of our pieces have an average read time of even up to ten minutes. This is where we know someone is really reading it carefully, referring to it for research purposes, or when the evergreen element kicks in. Once we publish a story within that first month window, we usually get a high volume of viewers. However, a lot of our pieces consistently get read on a lower volume basis every month, and that's where the evergreen element comes in. People bookmark them when doing their own stories or when doing reports, and we've seen some of our articles get cited in reports by the UN. We keep an eye out for these elements to understand how our research is building engagement with the public. We've also decided to move content onto Instagram. Most of our audience is actually Gen Z or millennials and they have short attention spans. They prefer visuals and don't want to click on a link to read another article. So we've produced micro-stories with ten squares that are published on Instagram. They capture social media metrics and allow us to engage our audience better.

Tom: That's interesting. Do you think people read articles less now?

Peiying: It's a big question, and it does make us confront a couple of things. Journalists tend to think that every issue is important and everyone must and should read it. However, we find that there is a gap with our audience and what they actually want to read. It's not that people are frivolous and don't care about important issues, but the way that we're engaging them is wrong. That's the gap we've been trying to close. We've been producing micro-stories, as most of our audience is Gen Z or millennials who have short attention spans and prefer visuals. We've found that producing micro-stories that capture social media metrics allows us to engage our audience better.

Tom: you mentioned that the average time is four to five minutes, do you try to encapsulate your stories in those first few actions so that someone who doesn't read the full fifteen minutes will still get the essentials?

Peiying: Yes, it's something that our team has been trying. We've actually run a couple of experiments. We don't do this for every story, but in any piece of writing, we do try to make sure that in the opening part of the article, you get a sense of the direction and tone of the story. We've also implemented Hotjar on our site, and we tried A/B testing writing a little summary at the top of our story. Some pieces have a bullet-pointed summary at the top that allows you to take away the essentials of the story, and we ran an A/B test to see if people care about the summary. The conclusion was that people don't care very much, at least for visual stories, and they prefer to scroll through a story. I think that was very interesting for us because we considered whether we should implement something like that, and that was the takeaway from the experiment. However, I think it really varies from piece to piece. For us, we find that if somebody has the patience to come through a story, they will. If they don't, we have these ten-slide summaries for Instagram, and we find that most of our readers just read the Instagram squares. The majority of our readers read those squares, and that is good enough. Usually, by the fourth square, they're getting a good sense of the story. If they really need to know more, then they would move on to the website to read the full thing.

But by that time, our goal has been achieved. Essentially, on social media, we're also thinking about moving this to Twitter. Unfortunately, not many Singaporeans use Twitter. The number one social media platform is still Instagram. That's where we're at, but we are thinking of expanding into other markets like Manila, or even Jakarta, for example. They may have a different preferred social media platform, so we're also looking at how we can translate some of this material for those platforms.

Tom: That answered my next question, which was if I want to get a TLDR, I can go on Instagram, and then you've achieved your mission without compromising the visualization for people who are really curious about the subject and want to dive in. So you have both options you cater to. You're covering both audiences, which is great. While super interesting to hear about the Hotjar, the summary is just not of interest.

Peiying: Yes, I think there are other tools you can use, but for us, Hotjar has a pretty nice free tier, and it really measures the hot spots on the page. For those people who invested a lot of time creating a hyper interactive visual, you get a sense of where people are clicking with it and playing with it, and that's something you can implement without much cost. Because I think you just put a little script on the page and it will grow. You know, the first sort of had users or something like that. So that's been really useful for us.

Tom: Now, you mentioned that some pieces have an average time of ten minutes, which to me is mind-blowing. I didn't even think that was possible. I think the average read time for a lot of articles is like two or three minutes. Are there any success patterns for those stories with the longest read times? You had mentioned in one of our talks that the accumulation of backlinks is a good measure of engagement with the story. So, in short, on those most powerful stories, no matter what metric we use to determine whether or not they're powerful, have you noticed any common success patterns?

Peiying: Um, I haven't actually thought about this before, but it's a good question. If I were to make a guess or hunch of why some of those pieces have a very long read time, for me, I suspect it's because those are the pieces that have the most meaning to a conversation. For example, a topic on sexual violence and signaling. It's a really important topic, and what we did with the data was to demystify certain narratives and ideas. But it's also one that generally people already know and understand. So without having to read through every visualization, they sort of get what we're trying to say. So those do have an average of four or five minutes, and sometimes a month average of six or seven minutes, but they're not that long. The ones that really get to ten minutes, sometimes even fourteen minutes, are the ones where we've done unique research that adds to that conversation in a way that no other report has done. For example, a story that recurrently gets that sort of read time is a story that I wrote a couple of years ago on how Southeast Asia is managing its waste problem.

Generally what the article is proposing is that it's not about how many straws or how much plastic material you recycle. It's also not about banning plastic or single-use plastic. It's really about how waste is disposed of, and that is the biggest challenge for Southeast Asia because most of us put our waste in landfills, and in some countries, these landfills become huge and become entire habitats and ecosystems of their own when they're poorly managed. This is particularly bad in Indonesia and the Philippines, for example. I think this is a surprising data point for the climate change conversation or for the environmental protection conversation because so much of the efforts that businesses and even activists are making are focused on reducing single-use plastic, reducing or avoiding plastic straws, and things like that. They forget that they live in an ecosystem of trash, and there's a life cycle of where your trash goes to, but they forget about their trash essentially after they throw it in a bin.

And I think that article brings forward a lot of these questions that people don't usually think about, and that piece has a lot of impact for both policymakers, actors, and also researchers in general. I've also received comments from people who are searching this topic and who reached out to me to ask for how I gathered the data and things like that.

That's one example. Another example that often gets a lot of readership as well is about foreign domestic workers in Singapore. So one of our freelancers went through this very detailed survey data that a nonprofit agency had conducted with foreign domestic workers in Singapore. That story is really lengthy too, but she really takes you into the specifics of the questions that were asked and sort of percentages of responses that were given. This is a topic that I think a lot of people have a certain stereotype about, but they don't know it well, and I think our story was able to add value again to that conversation with research, so that people feel more informed when they're talking about this. It's also a topic that mainstream media doesn't particularly care for, for various reasons. So those pieces where there is real human interest and raw research material on the web, then we get a long read time.

Tom: Yeah, and you know that resonates. This is the fourth time that I asked the question in an interview, and consistently I get the same answer, which is that no amount of visualization can compensate for the interest there is in a story.