In Sesotho, dipuo means conversations. This page explains how those conversations are collected, how the numbers are checked, what the work cannot tell you, and how the figures are kept up to date.
Two kinds of work
Analyses are long form policy pieces. Each one reads official data, peer reviewed research, legislation and court records against what people are saying in public, and each ends with its full list of sources and a ready made citation.
Signal Posts document a moment in public conversation. Each collects public posts on one theme over a stated window, quotes them exactly as written, and adds a short reading of what they signal.
Both rest on mixed methods. Qualitative reading gives the meaning; official sources such as Stats SA, SASSA, DHIS2 and Treasury reports give the structure. Neither is treated as enough alone.
How Signal Posts are made
- Collection. Only publicly visible posts are used. Most come from X; some pages also draw on TikTok and Instagram. The collection window is stated on each page, for example 7 to 20 August 2026 for the load shedding post.
- Selection. Posts are chosen by purposive sampling, which means deliberately, not at random. The three tests are thematic relevance, engagement, and linguistic or analytical significance. The aim is to show the range of a discussion, not to build a one sided case. On the load shedding post, that meant including a voice that complicated the dominant frame.
- Coding. Posts are read closely and coded by theme before anything is written about them.
- Reading. Each post is followed by a short section called "What this signals". The reading is kept separate from the quote, so you can judge the quote for yourself.
- Checking against data. The conversation is set against official figures. Where the talk and the data disagree, or where no data exists, the page says so.
Ethics and safety
- Only publicly available posts are used, and no personal data beyond publicly visible usernames and post content is collected or stored.
- Posts are reproduced for commentary and research under fair dealing provisions.
- First person statements of suicidal intent are not reproduced. Documenting public discourse about suicide is not the same as publishing a specific person's crisis disclosure, so the Suicide Prevention Month page leaves one such post out, however relevant it was.
- Pages that discuss suicide carry helpline details, including the SADAG Suicide Crisis Helpline on 0800 567 567.
What this work cannot tell you
- It is not representative. Every Signal Post says it does not claim to be exhaustive or statistically representative. It is a qualitative snapshot.
- Social media is not South Africa. People on these platforms are not a random sample of the country, and selecting by engagement can give extra weight to the loudest and most shareable posts.
- A post is not a belief. It shows what someone chose to say publicly, on one day. It does not show what they think, or whether what they describe is true.
- It is a moment, not a trend. A Signal Post shows what was being said during its window, nothing more.
- Some data does not exist. Where a gap is found, the page names it. The load shedding post, for example, notes that municipal distribution failures and Eskom's national count are measured separately, and no public dataset combines them.
How numbers are checked
Every figure on The Numbers is traced to the source it names, with the period it covers and the date it was last checked.
- Measures are kept apart. The intimate partner femicide rate is not the rate for all killings of women. The official unemployment rate is not the composite labour underutilisation rate. Each label says which one it is.
- Derived figures are labelled. If a number is worked out from another, such as about 900 men a month from 10,861 male suicide deaths in a year, the description says so.
- Sources are named on the card, so you can check the original and not take this page's word for it.
Updates
Figures are refreshed when their sources publish new data. Statistics South Africa releases the Quarterly Labour Force Survey every quarter, and the labour figures on The Numbers and the homepage are updated to the latest release. Each figure shows the period it covers and the date it was last reviewed.
Analyses are dated records. Each analysis uses the figures that were available on the day it was published, and its date is shown at the top. Figures in published analyses are not revised afterwards, so newer data appears on The Numbers and the homepage instead.
- Latest data refresh: 6 October 2026
- Labour market, second quarter of 2026 (released by Statistics South Africa on 11 August 2026). The official unemployment rate is 33.6%, up from 32.7%. Youth unemployment for ages 15 to 34 is 47.4%. The composite underutilisation rate is 46.3%, and 8.5 million people are unemployed. The next release is expected in November 2026.
This page is updated whenever the method changes. If you think a figure is out of date, or a source has released newer numbers, I would like to hear about it.
Get in touch