The Entrepreneurial Efficiency Equation (η)

In April 2021 Zymergen went public at a valuation above three billion dollars. It had spent eight years and more than eight hundred million dollars building a way to manufacture materials biologically rather than petrochemically, and it had a product to show for it: Hyaline, an optical film for flexible displays. The science worked.

Four months later the stock fell 68 per cent in a single evening, the chief executive was gone, and the company had told the market to expect no product revenue that year.

What nobody had established was whether a customer could get the film into a manufacturing line. One question, never asked with enough force, decided everything else.

Every founder has a version of that question live right now, and it never arrives as cleanly as it reads here. Of everything that is wrong this week, which one thing is holding the rest still?

Progress is a ratio

A venture’s rate of progress behaves like a ratio. What the team generates sits on top. What the world charges to deploy it sits underneath.

Dashboards rarely measure either. They count activity, lines of code and sales calls, when the number that decides the outcome is how much uncertainty the week removed.

Schumpeter named creative destruction. Knight separated uncertainty from risk. Barney showed why some resources hold their advantage. Sarasvathy showed how founders reason when the future cannot be forecast. Each mapped the terrain, and none was building a speedometer. They are rigorous, and they are silent on a Tuesday morning when the burn rate is high and growth is flat. That silence is what this piece is for.

The four on top are the ones you control. Problem-solving velocity is how fast unknowns become decisions. Innovation is how non-obvious and defensible the resulting advantage is. Resources are cash, talent, distribution and borrowed credibility. Adaptability is how quickly you can reconfigure without breaking what already works.

The two underneath you can reduce or route around but never own. Uncertainty is the untested assumptions standing between you and a working business. Network complexity is the gatekeepers and approvals standing between you and a customer.

The operators carry the argument. Each one is a claim about how ventures fail, and each one can be checked: why the terms multiply inside each pair, why the pairs add rather than multiply, and why one friction sits under an exponent while the other does not.

Why the numerator multiplies

In 2006 the surgeons at Geisinger Health System took the American Heart Association and American College of Cardiology guidelines for coronary bypass and turned them into forty verifiable steps, from the first clinic visit to discharge. Nothing on the list was new. Every step was already standard practice, and every surgeon in the department already knew all forty. Resources sat at full strength: the people, the evidence, the theatre and the record system were all in place, and nothing remained to be discovered.

Then they measured how many patients received all forty. Fifty-nine per cent.

Not fifty-nine per cent of the steps. Fifty-nine per cent of patients got the complete set, and the rest got most of it. Added up, the department looked close to flawless, because almost every step happened almost every time. Multiplied, it sat at 59 per cent, because a patient who missed one step missed the set. The weak half was adaptability, the department’s capacity to reconfigure itself so the whole list ran by default rather than by memory, and no further surgical skill was going to change that.

So Geisinger changed the machinery rather than the standard. The forty steps were hardwired into the electronic record as templates, order sets and reminders, so the complete list ran whether or not anyone remembered it. Inside four months every patient was getting every step. Compliance slipped once afterwards, to 86 per cent, then returned and held. Length of stay dropped 16 per cent, from 6.3 days to 5.3.

Nothing was added to the numerator. One half of a pair was brought up to the other, and the result moved.

That is what the numerator claims about your team. Its four terms work in two pairs. Resources times adaptability is the deployment pair, the one Geisinger fixed. Problem-solving velocity times innovation is the discovery pair, and it breaks the same way: a genuinely non-obvious insight held by a team that takes two months to decide anything will watch a worse idea reach the market first.

A pair can also break from outside the company. When a tool raises velocity across an entire market at once, everyone’s numerator rises together and no one separates, which is what generative AI did to differentiation.

That decides where the next week goes. Effort spent on your strong half runs into a weak multiplier and mostly disappears. Effort spent on the weak half raises the multiplier itself, and everything already built starts paying.

The uncomfortable part is that the weak half is rarely the one you enjoy, and never the one your recent progress points to.

Why money cannot outrun uncertainty

Everything so far sits above the line, where the terms are yours to strengthen. The two terms in the denominator look alike and are not.

Network complexity accumulates. It enters the denominator as plain N. Each additional gatekeeper is one more conversation, one more approval, one more integration, so ten stakeholders cost roughly ten times what one stakeholder costs. You can route around them, absorb them, or buy your way past them, and the bill scales with the count.

Uncertainty compounds, because unknowns interact. It enters as eᵁ, not U. Two unvalidated assumptions do not double your exposure. Each can resolve either way, so two assumptions describe four possible worlds and only one of them is the world you planned for. Six assumptions describe sixty-four, and sixty-three of them contain an error you have already built on.

One of these frictions can be outrun. The other cannot.

The venture capital industry has been acting on that for thirty years. Paul Gompers, studying a random sample of 794 venture-backed firms, found that investors do not release capital in proportion to need. They release it in tranches, concentrating short rounds and close monitoring exactly where information is thinnest: early stage, high technology, few tangible assets. Every tranche buys the right to stop.

Staging is the asymmetry written into a contract. It exists because money released ahead of knowledge funds a larger version of the same mistake.

The exponent is a claim about direction, not a fitted parameter. Nobody has measured its base. What it asserts is that one friction outruns linear effort and the other does not.

That sets the order of the work. Raise resources while uncertainty keeps growing and you have bought a bigger team to build on a foundation nobody has tested. Prove the hardest thing first is not folk wisdom. It falls out of the shape.

Reading Zymergen

Reading Zymergen through the equation takes about a minute.

Three terms decided this company. Two of them were strong.

The numerator was strong. Innovation was real: the company had spent since 2013 building a genuinely non-obvious platform, and Hyaline was a product of it. Resources were extraordinary by any standard, above eight hundred million privately and roughly five hundred and thirty million more at the April 2021 listing.

The denominator was untouched. Two assumptions carried the entire business and neither had been tested. The first was whether customers could integrate the film into manufacturing lines they already ran. The second was how large the market actually was. In August 2021 the company disclosed that several key target customers had hit technical problems doing exactly that integration, and that the market looked smaller than expected. Product revenue for the year went to zero.

The Securities and Exchange Commission later found that Zymergen’s own sales team had put the 2021 display market at roughly forty-two to one hundred million dollars, under a tenth of the billion-dollar figure presented to investors. The company settled those charges with a thirty million dollar penalty, having filed for bankruptcy the year before. The disclosure question belongs to regulators. The structural one is more useful to a founder: the number that decided the company’s fate was knowable, and the work of knowing it had not been done.

Capital made this worse rather than better. Money at that scale removes the pressure that would otherwise have forced the integration test in year two, when it was cheap. A smaller balance sheet is a cruder instrument and it asks the binding question sooner.

Change the units and this is the most ordinary seed-stage failure there is. A real insight, a funded team, and one untested assumption about whether anyone can actually use the thing, carried for eighteen months because testing it is uncomfortable and building is not.

The structural read

η is read rather than scored. Its value is structural: which terms multiply, which add, and which one compounds.

The arithmetic forces that reading. Rate the six terms one to five and compute η, and it looks like measurement. It is not. Ordinal judgements are not quantities. Across every possible combination of scores, reducing uncertainty by one point beats reducing network complexity by one point. Not usually: in all of them.

That is the exponent doing exactly what it is for, and any scoring scheme built on it returns the same answer whatever you feed it. Read as a shape rather than a score, any of the six can bind.

Anyone using it as a forecast is misusing it. It forces the question most founders skip: of everything wrong right now, which term is holding the rest still?

Reading your own equation

Reading starts with the shape you are in. Ventures do not spread evenly across the six terms, and most sit in one of three.

Escape velocity. Uncertainty binds. Feasibility itself is unproven, and no amount of commercial motion answers that. The tell is that every confident projection sits downstream of one untested assumption. Zymergen was here and spent as though it were somewhere else. A company in this shape is not running a speed problem at all. It is missing a prerequisite, which is a different question with a different method: not how fast you are moving, but what makes the thing buildable at all. The move is evidence rather than progress: the smallest test that could kill the assumption, run before anything else is built on top of it.

Stamina. Resources bind. The path is known, the innovation is real, and there is not enough fuel to walk it. The tell is that you can describe exactly what three more engineers would do. The move is substitution rather than acquisition: leverage, partnership, borrowed distribution, and scope narrowed to the work that kills the constraint.

Trust moat. Network complexity binds. The product works and cannot get in. The tell is a pilot that succeeds and does not convert, blocked by someone who was never in the room. This is the shape most often misdiagnosed as a product problem, and the one where the friction, once crossed, becomes the defence: a coordination cost you have paid and a competitor has not. The move is access rather than persuasion: find the person who can say yes and design the pilot so they are in the room from the start.

Resources are also the most legible of the six, countable from the outside in a way that velocity and adaptability are not. That is why a stamina-constrained company is usually diagnosed correctly and an escape-velocity company usually is not.

Naming the shape tells you which conversation to stop having. An escape-velocity company running an enterprise sales motion is optimising a term that is not binding.

Each term has things you can actually look at, none of which is a score.

What you strengthen

1. Problem-solving velocity

   Decision-cycle time, experiment throughput, and the gap between asking a question and having an answer you would act on.

2. Innovation

   Whether the advantage can be stated in a sentence, whether anyone has copied it, and how long a competent team would need to.

3. Resources

   Runway, access to senior talent, and borrowed credibility that opens a door you could not open yourself.

4. Adaptability

   The speed and quality of your last reconfiguration, the tightness of the feedback loop, and how modular the thing you have built actually is.

What you reduce or route around

1. Uncertainty

   The count of critical assumptions still unvalidated, where critical means the venture does not work if the assumption is false.

2. Network complexity

   The number of gatekeepers, approval steps and coordination paths standing between the product and a paying customer.

WORKING TOOL

Which term is binding, and how would you know if you were wrong?

1. Which pair is broken, and which half of it?

   Take P_s × I and R × A in turn and ask whether either half sits near zero. A near-zero half explains stalled progress better than four mediocre ratings, because the other half cannot rescue it.

2. Is your hardest assumption upstream or downstream of this week’s work?

   Name the one thing that, if false, makes the rest irrelevant. If the week’s work sits on top of it rather than testing it, uncertainty is binding whatever the calendar says.

3. When a deal stalls, was the reason inside the room?

   If the person who blocked you was never in the conversation, the constraint is network complexity. If they were in the room and unconvinced, it is innovation.

4. What would falsify your answer?

   Name the observation over the next fortnight that would mean you picked the wrong term. A diagnosis you cannot be wrong about is a preference.

The reading is meant to be repeated. If the same term is still binding a month later, either the intervention missed it or you named the wrong term. Either way you have learned something, and faster than another month of spreading effort evenly would have taught you.

What to do once you have named the term is a larger question, and each of the six behaves differently enough to deserve its own treatment. Those pieces follow this one, and each will do for its term what the forty steps and the staging contracts did here: take one structural claim and put it against evidence. Adaptability is the first of them.

Two things keep this a reading rather than a verdict. The first is that it is a snapshot, and the terms move each other: raising resources can lower adaptability through the bureaucracy it funds, and lowering uncertainty often raises network complexity by making you interesting to stakeholders who ignored you before. The second is that weighting is contextual, so what binds a materials company at Series A is not what binds a vertical SaaS company at seed. Neither has room here.

That is also how the equation grows. Adding a term will mean adding a claim about how it combines with the others and carrying the burden of showing it, rather than adding a symbol for its own sake. Where evidence contradicts the shape, the shape changes.

Conclusion

Under runway pressure the instinct is to work on everything, because everything is visibly wrong. The evidence says otherwise. Geisinger had every resource in place and every surgeon already knew all forty steps, and the number that mattered was still 59 per cent, because a pair is only ever as strong as its weaker half. The staging contracts say that money released ahead of knowledge funds a larger version of the same mistake. Zymergen says both at once, at a cost of more than a billion dollars.

None of that is an argument for working less. It is an argument about order. The same week of effort returns almost nothing against a weak multiplier and compounds against a strong one. Which of those you are living in is decided before the week starts, by what you chose to work on.

Reading η is a discipline of subtraction. Find the term holding the others still, and accept that most of this week’s obvious work is not this week’s work.

The founders who compound know, this month, which single term is deciding the ratio, and they leave the other five alone until it moves.