KASPR Research · Brazil
When Brazil logs on
A close runoff, a campaign fought on phones, and a network that differs from one district to the next. What the timing of a mobilisation post is worth.
On 25 October Brazilians choose their president in a runoff between Luiz Inácio Lula da Silva and Senator Flávio Bolsonaro. The first round on 4 October was close, about 47% to 45% (The Washington Post), and the last polls before the runoff put the two within a point of each other (Reuters).
With a race that tight, turnout decides it: undecided voters and those who might stay home could settle the result (The Washington Post). Much of the work of bringing them out happens on social media, and the platforms have become a campaign issue in their own right: the electoral court has ordered Meta to treat both candidates equally (The Washington Post).
A call to vote, a rally invitation or a get-out-the-vote video has hours, not days, to travel. So when it goes out matters. This story takes no side; the same timing applies to every campaign, party, movement and civic group.
Read the white paper (PDF)Scroll to read ↓
Change in engagement per ten milliseconds of national latency, with 95% intervals; 1.42 billion posts.
Start with the research. In KASPR’s study of 1.42 billion TikTok posts, every extra 10 milliseconds of national delay in the posting hour went with 1.3% fewer views, 0.8% fewer likes and 1.1% fewer comments, comparing posts within the same country, week, weekday and hour. Slower hours shrink how far a post travels; the people who do watch engage just as much.
The clock matters more. Holding the network fixed, posts at 16:00 local time did best, and a 03:00 post typically got 21% fewer views. A 19:00 post would typically have got about 3% more views an hour earlier. You can check what an hour either way would have meant for your last post.
Brazil’s internet is not one network. Each dot is a district. Of 5,406, 2,350 have enough measurement of their own; the rest take their region’s values and are drawn faint. The middle 80% of measured districts sit between about 229 and 267 milliseconds, and a handful exceed 500. Hover a district, or switch the map to see how much the hour matters there.
By state, the northeast responds quickest and the north slowest. Alagoas, Piauí, Rio Grande do Norte and Pernambuco sit near 222 to 233 milliseconds; Rondônia, Acre, Roraima and Amazonas between 285 and 304.
Across the day, most of Brazil holds steady. Nationally the typical week moves by about three milliseconds, lowest before dawn and highest around 20:00 to 22:00. But in 386 measured districts KASPR’s model rates the hour as mattering a little, and in 10 it matters a lot.
In São Paulo, Rio de Janeiro and Rio Branco a weekday barely moves: two milliseconds or less between night and the evening peak. Parintins in Amazonas is slower all day rather than at one hour. The level differs from place to place; the daily shape is shared.
Elsewhere the evening bites. In Florianópolis a weekday runs from about 242 milliseconds at night to 255 in the evening; in Groaíras and Orós in Ceará the rise is around 17 to 18. That is when supporters are most likely to be scrolling, and Monday around 21:00 is the roughest slot of the typical week.
What that could cost. Applying the study’s national estimate to those swings, posting at the evening peak rather than the steadiest hour would mean about 2.3% fewer views in Groaíras or Orós, about 1.7% in Florianópolis, and under half a percent for Brazil as a whole. Indicative only: the estimate is for national hour-to-hour changes. Test your own post with Rewind to see what an hour earlier is typically worth.