Key takeaways
- On very small budgets, targeting and copy have to compensate for reach you cannot afford to buy. There is no room to waste impressions on the wrong audience.
- South Africa's consumer market splits into two audiences with opposite requirements, and they occupy the same cities, so location targeting mostly fails as a shortcut.
- Device type, language settings and cultural interests were more reliable segmentation signals than geography.
- Writing for second-language English speakers is harder, not easier. Idiom, wordplay and cultural reference all fail, and only precision survives.
- Splitting a small budget in half to test two versions, then putting everything behind the winner, is crude but it works when you have no historical data.
Two markets, one postcode
Most segmentation has a geographic shortcut somewhere in it. Postcodes, suburbs, catchment areas. Marketers reach for it instinctively because in most countries income and location correlate strongly enough to be useful.
In South Africa that shortcut breaks.
The country is, by the World Bank's measure, the most unequal in the world, ranking first among 164 countries with a Gini coefficient of around 63. The relevant consequence for anyone writing advertising is not the number itself but its shape: a very large working-class and low-income majority, and a smaller but substantial affluent cohort, living in the same metros. Cape Town and Johannesburg contain both, often within a few kilometres of each other.
So you have two audiences who need almost opposite things from a piece of copy, and you cannot separate them by drawing a circle on a map.
It's worth noting the ranking is contested. The Helen Suzman Foundation has pointed out that only 149 of the 218 countries and territories the World Bank lists have reported Gini figures in recent years, so "most unequal in the world" is a comparison against incomplete data. But the inequality is not in doubt. The league table is shakier than people quoting it usually admit.
What you target when you cannot target place
If geography will not separate the audiences, something else has to.
Device type was the bluntest signal and, for a long time, the most reliable. As recently as 2017 a significant portion of the lower-income market was still on feature phones and BlackBerry devices. That tells you something about spending power, and it also tells you something about the creative: heavy image assets and video were wasted on people whose handsets and data bundles would not load them.
Language settings worked well. Someone whose phone is set to isiXhosa or Zulu is a different reader from someone whose phone is set to English, and the platforms let you target on it.
Cultural interests did the rest. Local musicians were particularly useful, because music taste in South Africa segments along lines that income data does not capture. Sport helped too. Football and rugby reach meaningfully different audiences, and interest targeting on either put you in front of a reasonably predictable crowd.
None of that is sophisticated. It is a set of proxies assembled because the obvious approach was unavailable.
Every impression has to be pointed at someone who might actually care, and the words have to do work that money would otherwise do for you.
- Kyle Smith, Marketing Manager at Miraculous Equation
Simpler is not easier
The copy for the lower-income, largely second-language English audience had to be simpler. It does not follow that it was easier to write.
Idiom fails. Wordplay fails. Anything that depends on a shared cultural reference fails. Sarcasm, in particular, does not survive the journey. What you are left with is precision: say exactly what the thing is, exactly what it costs, exactly what happens next.
That is harder than it sounds, because most copywriting instinct pushes the other way. The clever line, the double meaning, the callback in the last sentence. All of it has to go, and what remains has to still be worth reading.
The affluent audience was the reverse problem. More room for register and nuance, less tolerance for anything that felt like it was shouting. But the targeting had to be tighter, because the audience was smaller and the cost per result higher.
One campaign, two audiences
The clearest example was an events company running live, in-person dating events, trying to take a bite out of the dating app market.
Their problem was supply: not enough men signing up. So the strategy was a discounted rate for men, and the targeting went after interests that skewed male. Beer, sport, cars, among others. Blunt, but the budget did not allow for subtlety and it worked.
The same client ran LGBTQ+ specific events, and that required an entirely different approach. Higher income brackets, and interests around theatre and opera, fashion, pop culture, lifestyle and aesthetics, events and activism.
Same company. Same city. Same month. Two campaigns that shared almost nothing.
Testing when you have no data
With a new client there is no historical performance to work from, so you have to generate some.
The method was crude and it held up. Split the budget in half, run two versions of the creative or the copy, let them run, then put everything behind whichever one got more bites.
At R200 (about €10) a campaign and rarely more than R1,000 (about €50) a month, that is a small sample by any reasonable standard. Which is the honest limitation of everything above: I was making decisions on sample sizes that a proper analyst would laugh at. Some of what looked like insight was probably noise. You cannot run multivariate tests at that spend, and you cannot separate a good week from a good campaign.
What the constraint did teach, though, was discipline about the variable. When you can only afford to test one thing, you think much harder about which thing.
What transferred
I have since spent years writing about cybersecurity frameworks, blockchain infrastructure and AI adoption, for audiences that look nothing like a Johannesburg events crowd.
The habit that came with me was not a tactic. It was the assumption that the reader's context determines the copy, and that the context is rarely what you would guess from your own.
A technical audience and a mainstream audience can be sold the same product and need completely different writing. So can two audiences in the same city. The work is the same either way: find out who is actually reading, then write for them rather than for the version of them you find convenient.
About the author
Kyle Smith has spent a decade explaining blockchain infrastructure, cybersecurity frameworks and AI adoption to people who don't build them. He is Marketing Manager at Miraculous Equation, founded the digital agency Essential Millennial, and started out in newsrooms. Based in Braga, Portugal.
.png)
.png)
