Adapted from my new book Quick Innovation MBA – available on Amazon, major bookstores, and the Tiisetso Maloma online store (RSA only).
What if you could see what is workable before the measurements exist?
Whether you are an investor evaluating the next big bet, an entrepreneur building something that has never existed, an innovator inside a large organization, a product manager shaping what users will love, or a student trying to understand how change actually happens — this piece offers a different lens. I have spent over a decade studying how innovation actually happens — not in spreadsheets, but in the real world where outcomes are uncertain, sequence matters, and the future refuses to repeat the past. What I have learned might surprise you.
Seeing what is workable does not mean predicting the future or calculating probabilities. It means seeing clearly enough to know where to place your attention, your capital, and your limited time.
This is the real challenge of innovation.
Data is extraordinarily useful for many things: measuring performance, optimizing systems, and understanding what already exists. But when the task is creating something new, data alone rarely tells you what to build next.
At best, data reveals the constraint.
It shows what is not working, what is inefficient, or what appears impossible within the current system.
But the solution—the innovation itself—is rarely visible in the measurement.
It emerges when someone decides to work on the constraint directly, even when doing so appears expensive, slow, or uncertain.
Why Data Struggles With Innovation
Data-driven methods work beautifully in what former trader and statistician Nassim Nicholas Taleb calls Mediocristan.
Mediocristan is the world of stable, bounded systems: heights, weights, manufacturing output, logistics efficiency. In these domains, averages are meaningful. The past resembles the near future. Variance is limited.
If you are managing a factory, optimizing a supply chain, or running A/B tests on an existing product, data-driven methods are not merely helpful—they are essential.
But innovation rarely happens in Mediocristan.
It happens in environments defined by three characteristics.
Extremistan
In Extremistan, outcomes follow fat-tailed distributions. A small number of products dominate entire markets.
One smartphone platform captures most global profits. One AI architecture reshapes an industry.
One search engine becomes the infrastructure of information.
In such systems, averages are misleading. The exceptional—not the typical—determines the future.
Path Dependence
Innovation unfolds through sequence: some breakthroughs are only possible once the right components exist and enough people are ready to participate. You could not have launched YouTube in 1995 and expected it to succeed as it did in 2005 — broadband was slow, video compression technology was primitive, consumer cameras were rare, and the Internet itself was still comparatively small.
By the mid?2000s, however, the Internet as a platform had reached approximately one?billion users worldwide, creating a scale of participation that made large?scale video sharing viable. When YouTube launched in 2005, early viral content took off quickly: within months some videos reached one million views, and by 2006 the site was serving over 100?million video views per day as uploads and engagement exploded.
No dataset from 1995 could have predicted this; the idea might have been imaginable, but it was not yet workable — the path simply didn’t exist yet.
Non-Ergodicity
Innovation systems are also non-ergodic: outcomes do not average out across time or across players.
What happened to Facebook in 2004 tells you very little about what will happen to a social media startup founded today. The conditions that enabled one success rarely repeat.
In such environments, metrics that assume stability and repeatability begin to break down.
They describe what exists.
They struggle to illuminate what might emerge next.
The Innovator’s Job: Enter the Constraint
This is where innovators behave differently.
They do not avoid constraints.
They enter them.
Data often points to a problem that appears unsolvable within existing parameters. Innovators treat that constraint as the starting point of exploration.
Consider three examples.
NVIDIA and the CUDA Bet
In the mid-2000s, NVIDIA’s GPUs were designed primarily for gaming graphics. Market data suggested this was their core business.
But internally, engineers recognized a constraint and opportunity: GPUs possessed massive parallel processing power that software developers could not easily access.
Solving this required building an entirely new computing platform: CUDA.
The investment was enormous. It required years of engineering work, new developer tools, and changes to NVIDIA’s business model. For a long time, the commercial payoff was unclear.
Metrics would have highlighted the cost and uncertainty.
Yet solving that constraint ultimately transformed the GPU into the central engine of modern AI (we are in 2026).
Tesla and the Battery Constraint in 2008
Early electric vehicles faced a fundamental challenge: battery safety and energy density.
Lithium-ion cells could overheat under heavy load, and large battery packs were difficult to manage safely.
Tesla approached the problem differently. Instead of waiting for a revolutionary battery chemistry, they re-architected the battery pack itself.
The innovation? Thousands of small commodity cells were combined with sophisticated cooling systems and battery management software. Liquid cooling allowed individual cells to operate safely within tight thermal limits.
This was not simply a measurement problem. It was a design problem—a new configuration of existing components.
They invented (innovated) a proprietary solution.
BYD and Charging Friction (2020-2025)
Electric vehicle adoption has long been limited by charging time and range anxiety.
BYD approached this constraint by aggressively developing battery technologies and fast-charging systems that reduce waiting time and increase usability.
The Wright brothers
The Wright brothers had plenty of data. They studied the glider performance records of Otto Lilienthal and other early aviation researchers. They built their own wind tunnel and generated new lift and drag tables after discovering that much of the existing aerodynamic data was inaccurate.
They also studied available engines. What they found was discouraging: no engine existed that was both light enough and powerful enough for sustained powered flight. The data pointed to the wall — weight.
But the data did not show the solution. The Wright brothers decided to build their own engine. With their mechanic Charlie Taylor, they designed a lightweight internal-combustion engine with a cast aluminum block — a material that had only recently become practical thanks to new industrial production methods. The result powered the Wright Flyer in December 1903.
The measurements revealed the constraint. The solution came from working on it.- – – –
Solving that constraint does not simply improve one product. It reduces friction for an entire category.
In each of these cases, the innovators did not wait for metrics to confirm the solution. They invested in solving the constraint itself.
Frameworks for Seeing What Might Work
If metrics cannot reveal the solution, how do innovators orient themselves?
Over the years, I have borrowed, adapted, and created frameworks to make the invisible visible — to help see what might be possible before the data exists. A framework is not a prediction tool; it is a perception tool. It structures complexity, highlights opportunities, and helps you ask better questions about what could work.
From a decade of studying innovation systems, building ventures, and observing how breakthroughs emerge, I have found four frameworks particularly useful for seeing beyond the limitations of metrics.
Framework 1: The Adjacent Possible
The Adjacent Possible, originally described by biologist Stuart Kauffman, refers to the set of innovations that become feasible given the components available at a specific moment.
Innovation moves step by step. You cannot leap arbitrarily far into the future. You can only build what becomes possible when the right components exist.
YouTube succeeded also because broadband networks, video compression, and digital cameras matured simultaneously.
The Adjacent Possible helps innovators ask:
- What new components exist today that did not exist five years ago?
- What combinations have suddenly become feasible?
Even NVIDIA’s CUDA platform emerged from recognizing that GPUs already possessed the necessary computational architecture.
Framework 2: The Human Greed Pyramid
Technological feasibility alone does not create successful innovations. Products succeed when they serve deep human motivations and inclinations.
The Human Greed Pyramid maps such motivations: survival, security, belonging, status, curiosity, pleasure, aesthetic appreciation, etc.
Innovation occurs when technology satisfies these inclinations with less friction or greater power.
Tesla’s vehicles appeal not only to environmental concerns but also to status, performance, and technological fascination (it is a luxury brand after all).
BYD’s fast-charging systems reduce the anxiety associated with electric mobility.
The framework asks a simple question: Which human motivations are being served—and which are currently underserved?
Framework 3: Stacking for Agility
Innovation is rarely about inventing entirely new components. More often, it involves recombining existing ones.
I call this Stacking for Agility.
The basic pattern is simple:
Established Utility + New Component = New Utility
Tesla’s battery architecture illustrates this principle clearly. Lithium-ion batteries already existed. Cooling systems already existed. Advanced software control systems already existed.
The innovation came from stacking these elements together into a new configuration that solved the battery safety constraint.
The same logic appears in NVIDIA’s GPU strategy. The GPU already possessed extraordinary parallel processing capability. CUDA stacked software tools and developer access onto that hardware capability, unlocking entirely new uses.
Stacking reveals how existing technologies can suddenly produce new forms of value.
Framework 4: Spaza Metricals
The final framework comes from observing spaza shops—the small neighborhood convenience stores common across South African townships.
These businesses succeed because they optimize three characteristics:
- Frequency — customers visit often
- Friction — transactions are simple and fast
- Essentiality — the goods meet everyday needs
I call this framework Spaza Metricals.
It evaluates whether an innovation might eventually become infrastructure.
Platforms like NVIDIA’s AI computing stack, Tesla’s charging ecosystem, and BYD’s battery systems increasingly exhibit these characteristics. They are used frequently, operate with low friction for their users, and support essential modern capabilities.
When a technology begins to satisfy these three conditions, it often transitions from product to infrastructure.
Working on the Constraint
None of these frameworks guarantee success.
What they do is help innovators identify constraints worth solving.
Data can reveal the constraint. Frameworks help you see how it might be solved.
But solving it often requires committing resources long before the metrics become favorable.
NVIDIA invested heavily in CUDA before the AI boom made its value obvious.
Tesla redesigned the architecture of electric vehicle batteries before electric vehicles were widely accepted.
BYD continues investing in battery and charging infrastructure to remove adoption barriers.
In each case, the innovators treated the constraint itself as the opportunity.
– – – –
Metrics describe systems that already exist.
Innovation creates systems that do not yet exist.
This does not mean data is useless. Data is indispensable for measuring progress once the path becomes clearer.
But the path itself often emerges only after someone begins walking it.
The future is rarely built by those waiting for confirmation from the spreadsheet.
It is built by those willing to work on the constraint—until the measurements finally catch up.