Obsidian Fox · Wealth creation · Pretoria
Nothing worth building
gets built alone.
We are a wealth creation company. We put our own capital to work — in listed markets and in private South African businesses — and we help the ones we back grow. Every company we have been part of was built with people better than us at their part of it. No one is an island. That is the whole method.
~/obsidian-fox · Engineering · Pretoria
Every business is a system.
Most are running undocumented.
This is the layer underneath. We build the data, the automation and the AI that everything on the other side of this page quietly depends on — inside the businesses we back, and for companies who need technical leadership without hiring it permanently. Our director has been CTO at two software companies since well before it was a service line.
Capital
Our own money,
in two places.
Obsidian Fox invests what it has earned. No fund, no outside capital, no client money — which is exactly why we can say what we think rather than what a mandate requires.
Stock
We trade and hold listed equities for our own account. It is a discipline as much as a position: nothing teaches you to read a set of financials faster than owning the consequences of misreading one.
Buy-in
We take equity in South African businesses with room to grow, and then we work. Capital alone has never fixed a company. What changes the trajectory is the systems, the people and the attention that arrive with it.
Stack
What we run,
and what we run it on.
Same rule as the capital. We do not recommend a system we have not run ourselves, in production, with our own money on the outcome. It is a slower way to build a service list and a much harder one to argue with.
Products we run
SimplyFleet and Flynt are not case studies. They are live systems with paying customers, and our director is the person the alerts go to. Everything on this page was learned at two in the morning at least once.
Systems we build in
Inside the businesses we back, technology is not a line item — it is how the investment actually earns. Instrumentation first, then the automation, then AI where it genuinely pays. Usually in that order and rarely in the order asked for.
How we help you grow
The things that quietly
hold a business back.
Rarely the product. Usually the plumbing — the tax structure nobody revisited, the payroll that eats three days a month, the brand that stopped matching the company two years ago. We work on those, ourselves and through a network of specialists we used on our own businesses first.
Investment & buy-in
Where it fits both sides, we take a stake and stay. This is first because it changes everything below it: an adviser with money in the business is answering a different question from one billing by the hour. It is also the whole point of the company — we are not a consultancy that occasionally invests, we are an investor that does the work the investment needs.
Our own capital — no fund, no outside moneyAdvisory
A second opinion from someone who has run the thing, not read about it. Strategy, operations, and the conversation nobody on your payroll can afford to start with you.
Direct 03Tax planning
Structure, timing, and the things owners routinely overpay for because nobody ever sat down and modelled the alternative. Handled by a practitioner registered with SARS, not passed to somebody else's firm.
Registered tax practitioner 04Payroll planning
Remuneration structuring, benefit design, and getting PAYE, UIF and SDL running without a monthly fire.
Direct & partner-delivered 05Accounting planning
Management accounts you can steer with, month-end that closes on time, numbers arriving early enough to act on.
Audit & review via registered partners 06Funding readiness
The pack a bank or funder actually wants, and an honest answer on whether we would lend against it ourselves.
Direct 07Marketing
Positioning first, campaigns second. Most marketing problems turn out to be an unresolved question about who the customer actually is.
Partner network 08Look & feel
Identity and the material a business hands people. Companies are judged on this before anyone reads a word about what they do.
Partner network —Not sure which?
Most owners aren't, and the thing actually in the way is rarely the thing that gets named first. Start with a conversation.
Tell us what's going on →What we build
The layer nobody looks at
until it fails.
Most technical debt is not a code problem. It is three systems that don't talk to each other, a process that lives in one person's head, and nobody with the authority to decide which of them goes. We do the deciding as well as the building, which is the part that usually turns out to be missing.
Data & instrumentation
Nothing below this line works on numbers you cannot trust. Before we automate a process or put a model anywhere near it, the data has to be clean, current and arriving on its own. It is the least interesting work we do and it is always first — because an automation built on a bad number is just a faster way to be wrong.
Direct — always first, never optionalAutomation & AI
Where a person is the integration between two systems, that person is a defect. We find them, remove them, and put AI only where it pays for itself — which is considerably fewer places than the market is currently claiming.
Direct — into businesses we back 03Fractional CTO
Technical leadership without the permanent hire. Architecture calls, vendor arguments, hiring the first engineers, and telling a board the truth about a delivery date. This is already the day job at two companies.
Direct 04Technical due diligence
Before you buy it or buy into it: what the stack actually is, what it will cost to keep alive, and which of the risks are real. Written for the person signing, not the person coding.
Direct — pre-transaction 05Systems built to order
Internal tooling and platforms that don't exist off the shelf because your process is genuinely yours. Built to be handed over with the documentation to run it, not to create a dependency on us.
Direct & partner-delivered 06Integrations & plumbing
The joins between the CRM, the accounting package, the spreadsheet, and the person re-keying between all three. Boring, high-return, and almost always the first thing worth doing.
Direct 07Legacy & migration
Moving a business off something that works but can no longer be changed. The risk is never the technology. It is the six undocumented behaviours nobody knew were load-bearing.
Direct 08AI evaluation
Deciding where AI actually pays before anyone spends on it. A good deal of this engagement is talking clients out of things. It is the cheapest work we do and routinely the most valuable.
Direct — frequently a no —Not sure what you've got?
Most owners aren't, and the diagram on the wall stopped matching reality some years ago. Start with an honest look at what is actually running.
$ tell us what's running →Instrumentation
Most owners fly on feel.
We fit instruments.
Ask an owner how last month went and you get a story. Ask the system and you get a number. The first thing we build in a company we back is the panel that makes the second answer available before the first one has been decided — and we choose what goes on it from the evidence, not from habit.
Why these four. Cash buffer days — cash divided by average
daily outflow — is the best-evidenced survival metric there is: the JPMorgan Chase Institute
measured 597 000 small businesses and found a median buffer of 27 days, and by its 2020
follow-up half held fewer than 15. It is a stock measure, so it moves weeks before a P&L
shows anything. Debtor days earns its place here specifically: National
Treasury data shows 95 399 invoices worth R12.4bn sitting past 30 days at the end of
Q2 2025, up 17% on the quarter. Revenue concentration is the risk none of
the others can see — it stays invisible until the client leaves. And
books current to is not a performance figure at all; it tells you how far you
can trust the other three.
Illustrative figures — an example panel, not investment performance and not any real
client's numbers.
Instrumentation
Same discipline,
different instruments.
The panel on the other side of this page measures whether a business survives. This one measures whether it can change. They are the same question asked at different layers — and this one has the better research behind it, because software delivery is one of the few things in business that has been measured properly for a decade.
Why these four. They are the measures from DORA — the DevOps
Research and Assessment programme, running since 2014 and published annually by Google Cloud,
which is the largest body of evidence on software delivery performance that exists. Two describe
speed: how often you can ship, and how long a change takes to get from decided to running. Two
describe stability: how often a change breaks something, and how fast you recover. The finding
that made the programme famous is that these do not trade off the way engineering folklore
assumes — the organisations shipping fastest also tend to break things least, because both
come from the same underlying discipline. Time to restore sits on the panel
because it is the number that decides how all the others feel from outside the building.
Toil is there because it is the honest measure of how much of your engineering
budget is being spent doing something a machine should have done.
Illustrative figures — an example panel, not any real client's numbers.
And on ourselves
We run the company
the way we build them.
The instrumentation is not only something we sell. It is how Obsidian Fox decides where its own money goes and whether it should stay there.
Built together
Neither of these
was a solo effort.
Two of the businesses we have been part of creating. Both still trading, both still serving the customers they were built for, and both built with people better than us at their part of it. That is the whole of our track record and we would rather show it than describe it.
SimplyFleet
Sold to companies running vehicles — the ones whose maintenance history lives in a lever-arch file and whose fuel spend lives in somebody's inbox. Trading in two countries, against established competitors, since well before it was comfortable.
Flynt
Built for businesses whose process is spread across three places and one person's memory. Onboarding, people admin, and the daily running of a company that has outgrown the way it was first put together.
How we think — a worked example
Everyone knows 70% of South African businesses fail.
Nobody knows who counted.
You have heard the figure. It is in every pitch deck, every incubator brochure and most government policy papers. We went looking for the study behind it.
The trail leads to a 2019 paper in the Africa's Public Service Delivery & Performance Review, which is the source almost everyone cites. Read it and you find a literature review — a survey of what other people have written — which states the 70% figure and does not say where it came from. A 2017 UCT thesis claims five in seven small businesses fail in their first year, roughly the same percentage describing an entirely different thing. That one does cite a source. The source is a magazine article.
It is not the only one. The line that 82% of small businesses fail because of cash flow appears on hundreds of finance websites, always credited to "a U.S. Bank study" that nobody can produce. Meanwhile the two counts that are properly sourced cannot agree on how many businesses there are to begin with:
We are not saying South African businesses don't fail. We are saying an entire advice industry is built on numbers nobody can source.
This is roughly how we look at a business, too. Not scepticism for its own sake — there is enough of that about — but the ordinary discipline of asking where a number came from before deciding what it means. It is also why the panel above carries its sources.
How we think — a worked example
Everyone knows 70% of digital transformations fail.
Nobody has measured it since 1993.
It is the most-quoted number in enterprise technology. It opens vendor decks, justifies budgets, and sells the very consulting engagements it warns you about. We went looking for the study.
The usual citation is McKinsey, whose 2015 piece on change management states that 70% of change programmes fail to reach their goals — without methodology, sample or source. BCG later reported that 70% of digital transformations fall short of their objectives, though its own breakdown shows most of those created value and simply missed a target, which is a materially different claim. Follow either back far enough and you arrive at Hammer and Champy's Reengineering the Corporation, 1993, which put the figure at somewhere between 50 and 70 per cent — and described it, in the same book, as an unscientific estimate.
In 2011 Mark Hughes, then at Brighton, took the published sources behind the 70% claim and examined them one at a time in the Journal of Change Management. His conclusion was that there was no valid and reliable empirical evidence behind any of them. That paper is fifteen years old. The number has been quoted continuously ever since, including by people who cite Hughes.
We are not saying transformations don't fail. We are saying the number telling you so was a guess about a different practice in 1993, and nobody has re-measured it.
This matters more on this side of the page than the other, because in technology the frightening number is almost always attached to something being sold. It is the same question we ask of a vendor benchmark, a stack we are reviewing, or a model's stated accuracy: where did this come from, and who counted? It is also why the panel above carries its sources.
Start a conversation
Tell us what's actually going on. Not the version for the bank.
There is no team between you and us, no qualification call, and no obligation on either side after the first conversation.
Start a conversation
Tell us what's actually running. Including the parts nobody documented.
No qualification call, no discovery deck, and no obligation on either side after the first conversation. If the honest answer is that you don't need us, that is what you will get.
Before you engage us
One thing said plainly.
Obsidian Fox both advises businesses and, in some cases, invests in them. That is the model, and it means our interests and yours can point in different directions during a transaction. So: we disclose it before any advisory work begins, we keep any equity negotiation a separate conversation from the advisory engagement, and we will tell you to take independent advice on the deal terms rather than rely on ours. If that ever stops being true, walk away.
Before you engage us
One thing said plainly.
We advise on technology and, in some cases, build it and then keep it running. That means when we recommend a system, there is a version of that sentence where we are recommending ourselves. So: we will tell you when something off the shelf does the job, and we say so often. Anything we build is handed over with the documentation and access to run it without us. And we price the assessment separately from the build, so the recommendation is never paying for itself. If any of that ever stops being true, walk away.