Statistics South Africa (StatsSA) publishes around 250 statistical releases a year — GDP, inflation, unemployment, poverty lines, census data — covering the full sweep of South African life. But the numbers alone rarely reach ordinary people. That’s the problem Kevin Parry’s team exists to solve.
Parry is deputy director of the data visualisation team at StatsSA, and joined The Outlier’s Out to Lunch webinar in April to talk through how a team of five turns government statistics into something the public might actually read.
Five people, 250 releases
The team consists of three content developers, a social media specialist, and a manager. Their job is to work alongside StatsSA’s subject matter experts — demographers, economists, statisticians — and produce accessible outputs: data stories, infographics, media presentations, radio audio bites, and video content. They don’t replace the official PDF releases; they sit alongside them, with the explicit goal of driving traffic back to the underlying data.
The team’s roots go back to around 2005, but the real turning point came with the release of 2011 census data in 2012. Then-statistician-general Pali Lehohla was a strong advocate for data visualisation, and StatsSA partnered with the late Hans Rosling’s Gapminder Foundation to present the census data in an unusually vivid and engaging way. That success convinced management to formalise what had until then been ad hoc work.
How they decide what to visualise
With far more data than they could ever cover, the team draws on three sources of inspiration. Social media surfaces common questions from the public — recurring themes that suggest a story might be worth telling. News events and international calendar days (International Beer Day gets a regular outing using inflation data) provide topical hooks. But the most important source is the subject matter experts themselves.
“What the subject area experts tell us isn’t what appears on the front page of the official release,” Parry explained. “It’s often something within the pages, somewhere in one of the tables or in the Excel sheets.” When an economist flags something they find genuinely surprising in a new data release, that’s usually where the team starts.
Final sign-off always rests with the subject matter experts, not the visualisation team — a deliberate choice. Parry was candid about the risk on the other side: a visualiser who doesn’t fully understand a dataset can easily misrepresent it, drawing conclusions that aren’t actually in the data. Between the 2001 and 2011 censuses, for instance, the disability question was rephrased to meet international standards, making direct comparisons invalid. That kind of nuance won’t appear in the raw figures.
Tools: mostly Excel and PowerPoint
The answer that surprises people most is the toolset. Around 90% of the team’s work happens in Excel and PowerPoint — titles, logos, annotations and all. The team also uses a few other tools for more specific chart types, in particular Raw Graphs for less conventional formats like Sankey charts, and QGIS and ArcGIS for maps. The team also holds Adobe Suite licences and uses Illustrator and Photoshop occasionally.
The workflow tends to run in reverse: the team spots an interesting chart type elsewhere, then asks what StatsSA data it might suit. That’s how they discovered Sankey charts and Voronoi diagrams — Parry’s personal favourite, which he described as a more visually appealing alternative to the overloaded pie chart.
What’s next
The team hasn’t yet explored programmatic tools like R or Python in any systematic way, though one team member is experimenting. AI has been discussed but not adopted — Parry’s concern is precisely the nuance problem: an AI generating a visualisation from raw figures won’t know that two census questions aren’t directly comparable, and could easily produce something misleading without flagging it.
Looking ahead, Parry pointed to interactivity as the most likely area of development — moving beyond static JPEG images towards visualisations users can explore. A dedicated web portal for StatsSA’s visual output is another possibility. And as internal demand for the team’s work grows, so too might the team itself.
Why it matters
Asked why StatsSA should visualise its own data rather than leaving it to journalists or researchers, Parry gave a straightforward answer: accessibility and accountability. The data is produced with public money, it exists to support evidence-based decisions, and StatsSA has a mandate — reinforced by international statistical charters — to make it usable. The visualisation team is how they honour that commitment for audiences who will never open a 200-page PDF.