Skip to content
11 min read

The Platform Shift Nobody Budgeted For

Every platform transition arrives as a line item in somebody else’s budget first. By the time it is in yours, the terms are set.

In conversation with Matthew Ball

Editorial cover: the words The Platform Shift Nobody Budgeted For on warm paper

Working draft

This is a working draft, not yet the published essay. The final text has not been pasted in.

Every platform shift I've lived through arrived the same way: as a line item in somebody else's budget, months before it showed up in mine. A hyperscaler changes its pricing tier. A vendor deprecates the API your integration depends on. A regulator issues guidance that makes the architecture you shipped last year the wrong one to have shipped. By the time the shift is visible on your roadmap, the terms of it were already set — by a product decision made in Seattle or Redmond or Brussels, not in Johannesburg, and not by anyone who had South African financial services in mind when they made it.

Buying the terms instead of setting them

This is the position most South African financial services technology is in, and it's worth being honest about why: we are, almost without exception, price-takers on the platforms the industry runs on. The cloud regions, the AI model providers, the compliance tooling — the overwhelming majority of it is built to a roadmap decided somewhere else, for a market where South Africa is a rounding error in the total addressable market calculation.

That's not a complaint about the vendors, who are behaving exactly as you'd expect. It's a description of a structural position, and structural positions have consequences. When a platform shift lands — a pricing change, a deprecation, a new compliance requirement the vendor didn't design for — the institutions running on it don't get a seat at the table where the shift was decided. They get an email with a migration deadline.

You can be reactive to a platform shift you didn't see coming, or you can be reactive to one you did. Those are the only two options if you never set the terms yourself.

I've watched this play out concretely: a compliance tool three institutions I know of were all quietly depending on changed its data-handling terms with the kind of notice period that assumes the customer is a startup that can re-platform in a sprint, not a bank with a change advisory board and a regulator to answer to. Nobody in South Africa was in the room when that decision got made, because nobody in South Africa was material enough to that vendor's business to be in the room.

The line item shows up as somebody else's roadmap

The reason this is a budgeting problem and not just a technical one is that the cost of a platform shift rarely shows up where you'd plan for it. It doesn't arrive as a capex line called "platform migration." It arrives as engineering time quietly absorbed into whatever sprint the deprecation deadline lands in, as a compliance review nobody scoped, as a delay to something else that was actually on the roadmap. The true cost is real and it's usually larger than the number anyone put in a spreadsheet, because nobody put a number in a spreadsheet — the shift wasn't a planned project, it was an emergency wearing the shape of routine maintenance.

What buying the terms would look like instead

I don't think the answer is "build everything yourself" — that's not a serious option for most institutions, and it trades one set of risks for a worse one. The answer is being deliberate about which layer of the stack you're willing to be a price-taker on and which layer you insist on controlling the terms of, rather than defaulting to price-taker on all of it because that's the path of least resistance today.

For most South African financial institutions, the layer worth controlling is the one regulators will eventually ask hard questions about: where does the data actually go, who can access it, and can you prove it, independent of whatever the vendor's roadmap decides next quarter. That's the layer where being a price-taker isn't just inconvenient — it's the layer where a platform shift decided somewhere else can turn into a compliance finding here, with your name on it and not the vendor's.

Everything else — the parts of the stack that are genuinely commodity, where switching cost is real but survivable — is a reasonable place to keep buying the terms. The mistake isn't buying rather than building. The mistake is not deciding, in advance, which layer you can afford to buy terms on and which one you can't — and finding out the difference only after the migration deadline email has already landed.

StrategyPlatformsFinancial services