Emergency versus planned: the real cost of a surprise swap
The same part costs far less bought on a schedule than during an outage. Where the difference actually comes from.
The thinking behind the engine — written to survive diligence, same as everything else on this site.
Why PPAs price degradation they cannot see — and what a defensible remaining-life record changes for both sides of the contract.
Request the paper →White paperHow an auditable asset-health record lets underwriters and lenders reassess risk on their own terms.
Request the paper →Technical noteThe three cases where a simpler model beat ours — and why showing them is what makes the rest of the numbers credible.
See the evidence →The same part costs far less bought on a schedule than during an outage. Where the difference actually comes from.
Chasing every scarcity event can degrade a battery faster than the revenue justifies. How a degradation gate keeps dispatch honest.
Why the deployment bottleneck kills most industrial-AI pilots — and how the Alliance stack sidesteps it.
Direct answers with their sources and their limits attached — three on the money, three on the method.
The outage is the small half. With replacement lead times running two to four years, a failure is a capacity event.
Warranty cycle limits cap revenue. The honest answer needs a degradation cost per dispatch, not a cycle counter.
8 to 12% over preventive, 30 to 40% over reactive — government figures, and narrower than the ones you have been shown.
Across four evaluations on public datasets, our physics terms did not improve accuracy once. The scorecard, the datasets, and where physics still earns its place.
It rejected about 70% of degraded rows. The rejection rate is the result, not a caveat on it.
That a method works. Not a failure rate, not a fleet statistic, not a claim about modern cells.
A read-only assessment proves the forecast on your own data before any hardware conversation. You see the record; you decide what it is worth.