The Hidden Clocks of Power

You take your power company into a new market. You hire the team, sign partners and start the long process of getting qualified, specified, licensed or contracted.

Twelve months later revenue, or the milestones that stand in for it, is below plan, and the CFO asks the obvious question: how much longer are we funding this?

Underneath it sit three questions. Can we sell it? Can we make money on it? Can we last long enough? The common ways each one breaks are below, plus three cases where nothing in the business broke at all. The list is not complete. The point is that one number sits on top of all of them.

The detail behind each branch:

Can we sell it?

• Wrong market. Customers want it, but the market is too small, too fragmented, too regulated or too costly to reach to be worth having at any realistic size, however well you sell. You pick a different market, or a narrower part of this one.
• Wrong offer. Customers do not want this, or will not buy it yet from you, because you have no track record in their market. Stopping or redesigning is right.
• Product not ready. Customers want it, but it does not yet perform in the field, or you cannot deliver it at volume. You fix the product before you add sellers.
• Wrong people. The offer is right and the team cannot carry it. Capability, personal relationships, leadership. You change who is doing it.
• Wrong execution. The team is right and the machine around them is not. Coverage, channel, process, pricing discipline. You change how it is done.

Can we make money on it?

• Wrong economics. It sells, and the price, the cost to build and serve, or the margin does not work. You reprice or you stop.

Can we last long enough?

• Wrong balance sheet. It works in seven years and the company can fund three. Or one bad week, a price spike or a collateral call, takes the cash before the strategy has had time to work. Or the company can fund seven and will not, because the capital would rather be somewhere else. From inside the first and the last feel the same, but one is a constraint and the other is a decision someone has already made about you. You bring in a partner or outside capital, shrink the entry to what you can fund, or stop while stopping is still cheap.
• No sponsor. For an entry inside a larger company, the core business will not back it. Sales compensation favors the existing products, other units fight it, or leadership attention moves on. This can starve a strategy that still has its budget. You find a sponsor with real authority, or you drop it.

Nothing is broken

• Wrong clock. Everything is working and results take longer than you assumed. If the balance sheet allows it, you fund the longer wait. If it does not, you look for a way to turn the market clock into a contract clock.
• Wrong plan. The forecast was too high. The timing assumption was right and the market is smaller than the plan said, but the business is still worth having at the lower number. That is what separates it from a wrong market. It is also the easiest explanation to accept, because it asks for the least change. You rewrite the plan and judge the business against the new one.
• Market moved. A policy change, a tax credit rule, a rate case or a competitor's price cut changed the ground after you entered, in a way you could not reasonably have priced in. The call was sound on what you knew. You decide again on today's facts.

The list is the same for every kind of seller. What each item looks like on the ground is not. Some sell to utilities, some compete with them for the customer, and some sell to the customer directly.

The balance sheet question differs too. A developer needs tax equity and project debt, an equipment maker needs working capital and factory investment, a software company needs runway until recurring revenue covers its cost, and a battery or demand response business that owns equipment at customer sites needs asset finance against revenue that program rules can change. A retail supplier meets the question fastest, because the grid operator and hedge counterparties can call for more collateral within days, long before the strategy has had time to work.

A market can also move because of the sellers in it. ERCOT battery fleet revenue rose 8 percent in 2025 while revenue per megawatt fell 37 percent, because the fleet's own growth competed away the scarcity it was paid for. The detail is in ERCOT Fleet Revenue Rose 8 Percent. Revenue per Megawatt Fell 37 Percent.

The expensive mistake is fixing the wrong one: replacing a team when the problem was timing, or cutting the budget when the problem was the plan.

You have to decide before you know which is true. The data will not settle it in time, so it is a leadership call.

Two questions tell you what kind of decision you are making

How long until we know, and what do we lose if we stop.

Price changes and campaigns sit bottom left, fast to resolve and cheap to stop. A long-running pilot sits bottom right, slow but cheap to stop. Turbine slot reservations and interconnection positions sit top right, with an operational technology security program just inside the same quadrant. Hiring a sales team and entering a new market appear twice: hollow where they get plotted, solid where they belong.

Many of the most consequential decisions in power sit in the top right, where a quarterly review cannot settle the question.

Your market entry sits higher on this chart than your budget review suggests. The saving shows up on a budget line. The qualifications, the relationships and the market knowledge that scatter when the funding stops show up on no line at all, and that is often the bigger number.

Why this is hard to see from inside is well studied. When a long gap separates a decision from its result, learning slows sharply, because the lesson arrives after the decision that needed it.¹ I studied this at MIT Sloan and it is still hard to spot in a live business. Boards meet quarterly, while utility sales cycles run years and equipment delivery and transmission builds run longer still. In that room the fastest clock wins, because it has the freshest numbers. The same gap between a fast commercial clock and a slow one stalls data center power deals, which is the subject of One Coordination Failure, on Two Clocks.

Four questions govern the decision

Know the clock. How long before you can reasonably know whether it is working? Then ask whose clock it is. A sales cycle can sometimes be shortened. A merchant power price cycle cannot: you can land the site, the equipment and the interconnection on schedule and still wait years for the signal that justifies the asset. What changes that is converting a market clock into a contract clock, which is the real case for a long dated offtake. What you are buying is a clock you can plan against. Who commits first in that contract, and how much, is set out in The Reciprocal Commitment Ladder. A regulated counterparty runs on a different clock again, because capital in rate base earns a return for decades while an operating saving is reset at the next rate case.² Before trying to accelerate a result, establish whether the delay is yours to manage.

Define the proof. In power, revenue arrives too late to be the only signal, so the proof has to come earlier: qualifications won, specifications written, pilots converted, offtakes signed. What must accumulate, to what level and by when, what must convert, and at what margin? A growing pipeline with flat conversion is not progress, and where it stalls narrows the diagnosis.

• Little qualification suggests a market, offer, targeting or access problem.
• Qualification without specification suggests product fit or field performance, or the people working the account.
• Specification without orders suggests price, terms, competition, procurement, delivery capacity or timing.
• Healthy conversion at every stage but slower than expected suggests the clock.
• Healthy conversion at the expected pace from a smaller pool suggests the plan was too high, if that smaller pool still pays back. If it does not, the market is wrong.
• Work accumulating behind one stage suggests a capacity constraint, and adding opportunities may make it worse.

Those stages are an equipment maker's. A developer's run from site control to interconnection to signed offtake. A software vendor's run from pilot to security review to paid rollout. A retail supplier's run from quote to signed contract to renewal. A battery or demand response business runs from enrollment to tested performance to payment. The reading works the same way.

You do not need to track all of these. Pick the one or two conversions that test the assumption you are least sure of, and agree on them before you start. For most sellers that is the last column of the table above.

The tell that separates people from process is dispersion. If one seller converts and others working similar accounts do not, look at the people. If everyone stalls at the same stage, look at the process first. Where there is no sales team, compare channels or customer groups the same way.

The funnel tells you where to look rather than what you will find, and some of these never appear in it at all. Economics shows up in margin, not in conversion: you can win every deal and still be wrong. Affordability shows up on the financial clock, which is why funding the wait has to be its own question. A missing sponsor shows up in budget and headcount decisions, and a market that moved usually shows up in a ruling or a competitor's price before it shows up in the pipeline.

Fund the wait. How much will we spend before the evidence has to arrive, and who with real authority has signed for it? A number and a name, agreed at the start.

Price the exit. If we stop, what is hard to recover?

Answer them before the investment starts, not in the meeting where people are already defending it or trying to kill it. The rest of this is why each one earns its place.

Stage totals are not interchangeable

GE Vernova's second quarter 2026 gas headline was 116 GW, which combines 53 GW of equipment backlog and 63 GW of slot reservations. Net the quarter and backlog grew 44 to 53 while reservations grew 56 to 63, so the committed stage grew faster.³ That is a healthy funnel. What reservations and signed orders each say about demand is in The Data Center Buildout Has a Toilet Paper Problem. ERCOT shows the same shape at system scale.⁴

Source: ERCOT Monthly Operational Overview, April 2026.

Every figure is correct and only the last is load.

Stocks tell you what has accumulated. Conversion tells you whether it is moving. A headline gives you neither, which is why defining the proof means naming a conversion and not just a total.

The exit that happens without a decision

Nobody stops a security program. The budget stays flat while coverage grows, an upgrade slips a year and then again, a role stays open because the business case is hard to write. Each is defensible. Together they raise exposure, and no meeting ever decided to accept it.

Sales looks unsuccessful because revenue has not arrived. Security looks unnecessary because the incident has not arrived. Both can be timing errors, and the second shows that an exit can be paid for in installments by people who never thought they were exiting. A market entry with no sponsor dies the same way.

A test you can actually fail

Any weak team can say the market needs another year. If every accumulating activity counts as progress, the strategy can never fail. That is rationalization.

We built a screen at Tafel Power to predict where generation interconnection requests would reach completion, and registered the test before building the model. It failed. Completion was roughly the same in every tier, the top tier did slightly worse than the bottom, and the factors we expected to matter did not. We could have moved the thresholds until it looked right, which would have fixed the model and destroyed the test. We dropped the screen and kept the failed run on record. That is testimony rather than a citation, offered because the argument is worth less if I only recommend the discipline to other people.

Define the evidence before you know whether you will like the answer.

Patience needs a budget

Two clocks run at once. The commercial clock builds relationships, qualifications and orders. The financial clock burns cash. If the financial clock hits zero first, it does not matter that the strategy would eventually have worked.

Patience is waiting. Strategic patience is funding a defined period of learning. It needs a budget, the evidence that should appear before revenue does, and a point where you will say the evidence did not arrive. Otherwise patience becomes drift.

Stopping does not put you back where you started either. A sales team can be rehired, but some things do not come back on the same terms: a surrendered manufacturing slot where the replacement is years out, a withdrawn queue position, a lapsed qualification, a specialist who leaves with the account history. Capital left in a weak strategy has its own cost, because it cannot be used somewhere better. The trade is between the cost of continuing when you are wrong and the cost of stopping when you are right, and those are rarely equal.

Back to the decision

None of this argues for waiting longer. Sometimes low revenue means the strategy is bad, sometimes a large pipeline is mostly noise, and sometimes the right answer is to stop early.

At month twelve you still cannot tell which of these you are in, and by the time the data settles it the decision will have been made for you. What you could have done is fix the terms before the first hire: how long the cycle that matters actually runs (sales, qualification, interconnection, renewal or program enrollment) and whether that clock is yours, what should have accumulated and converted by month twelve and at what margin, how much the company would spend before the evidence had to arrive and who would defend that budget when it got hard, and what you would lose that you could not buy back.

Then share the answers with the people doing the work. A team that knows the clock and the proof aligns on the same evidence and reads a slow quarter as a checkpoint. A team that does not spends it assigning blame.

That is the whole discipline, and it is only available in advance.

Know the clock. Define the proof. Fund the wait. Price the exit.

---

Footnotes

1. Hazhir Rahmandad, Nelson Repenning and John Sterman, "Effects of feedback delay on learning," System Dynamics Review 25, no. 4 (2009): 309-338. doi:10.1002/sdr.427 ↩

2. The nineteen-utility dataset behind this is at Tafel Power, "How a Utility Gets Paid," 16 September 2026. https://tafelpower.com/insights/how-a-utility-gets-paid ↩

3. GE Vernova Inc., second quarter 2026 earnings release, filed as an exhibit to Form 8-K on 22 July 2026. https://www.sec.gov/Archives/edgar/data/1996810/000199681026000147/gevpressrelease2q26.htm ↩

4. ERCOT Monthly Operational Overview, April 2026. https://www.ercot.com/files/docs/2026/05/13/ERCOT-Monthly-April-2026-FINAL.pdf ↩