Hi Reader,

I was looking at some of the data we've compiled in the JSE ESG dataset, which reveals an interesting dynamic of how the same data point can tell two different stories.

But first, in case you missed it, a couple of weeks ago we launched an African-centric full-coverage ESG dataset with close to 25,000 disclosures, three years of history and a benchmark layer with the JSE, available now on the JSE Market Data MarketPlace. And we have a webinar coming up next week to showcase the dataset (details below if you're interested).

Now onto the data, well almost, because I think outlining why we've built this is key: the dataset has coverage of local context that didn't really exist for South African issuers before this... not because nobody wanted it, but because most of what passes for ESG data in this market was never built here but in the Global North.

What a company has to do to operate sustainably here ends up producing data that most frameworks don't look for. We believe this is important context that asset owners and managers should take into account in their search for a pipeline of investable businesses in Africa.

For example, B-BBEE procurement spend isn't a line item in a global taxonomy. Local ownership structures aren't really either. So the data exists, technically, but it's been squeezed into categories that weren't drawn with this market in mind. That's the whole reason this dataset exists: not more ESG data, but ESG data with local denominators, built for the comparisons that actually matter here.

Here's what that looks like in practice.

Same issuer. Same metric. B-BBEE procurement spend against total revenue, sitting at 0.4187. Run it against the seven other issuers in its own industry, Consumer Staples, and it comes in at the 31st percentile. Below the peer average. Below the median. The kind of number that gets flagged in a board pack as underperformance.

Run the exact same 0.4187 against the twenty four issuers in its broader JSE super-sector and it jumps to the 52nd percentile. Above the median. Comfortably mid-pack, nothing to explain.

Take a look, this is the actual screen of the data in our Dataset Viewer, but anonymised:

Nobody touched the data. Nobody adjusted the company. The only thing that moved was the group of companies it got compared to.

It's such a clean example of how a percentile isn't a fact about a company, it's a fact about a company and a peer group, stapled together and handed to you as if it were one number. Change the peer group, change the story. Nobody tells you which peer group you're getting until you ask, and most tools don't let you ask.

So the real question underneath "is this company doing well" is never as simple as it looks. Rather, the question is compared to what... compared to who... and who decided that comparison was the fair one?

We're walking through more of this, live, on Tuesday 29th September, 3pm SAST with the JSE team. Register here. Join us if having this kind data is decision-useful for you as an asset owner, manager, or issuer.

I don't have a clean answer to which percentile is the "right" one in the example above for a company like this. I'm not sure there is one.

Which comparison would you trust if it were your own number on the screen?