Method

Every technology domain has a measurable rate of improvement.

We call it the K-value: the annual rate at which a domain's performance per cost improves. We estimate it from the innovation record and check it against domains where the performance history is known.

01 · DOMAINS

Define the technology precisely

We measure at any resolution: quantum computing as a whole, or a single architecture such as superconducting transmon qubits. A rate is only as meaningful as the definition under it.

02 · K-VALUE

Measure the improvement rate

Long-run performance data rarely exists, so we estimate the rate from where a domain's patents sit in the citation network, then check it against domains whose performance history is known. How many patents a domain has carries almost no signal. Where they sit does.

03 · CROSSOVER

Forecast the moment that matters

Rate plus today's performance gives you a trajectory. The trajectory tells you when a challenger is likely to cross the threshold your decision turns on, with the uncertainty stated and the conditions that would reopen the answer.

What K forecasts

The slope of technical performance over time, benefit per cost.

Which of two alternatives improves faster.

Which part of a system sets its pace.

When a performance threshold is near.

What K does not claim

Adoption timing, markets, regulation, and politics drive uptake.

Which firm or product wins. We identify the technology; execution is yours.

Short-term demand or price cycles.

Breakthroughs outside the patent record.

In one sentence: we forecast the slope of technical performance. Not the invention, the adoption date, the market share or the winning company.

Provenance

The method is public. Check it.

Peer-reviewed research. Published in Research Policy (2021) across 1,757 technology domains, with an out-of-sample correlation near 0.72. That figure validates the study's own predictions; it is not an accuracy rate for any deliverable.

Issued patent. U.S. Patent 12,099,572 B2, cited by number. We do not restate its claims here.

MIT lineage. Built on over a decade of empirical technology-performance research at MIT, including the Institute for Data, Systems, and Society.

Forecast notebooks

Three calls, logged before the outcome.

YES · SOONER

xMEMS micro-speakers: the crossover consensus missed

We measured an improvement rate far faster than the incumbent’s and recorded the cost crossover well ahead of consensus. The later observation landed inside the range we had logged in advance.

NO · LATER

Fusion: later than the hype

Against widely publicized net-energy targets, our estimated rates for tokamaks and HTS magnets put self-sustaining fusion further out. Announced timelines have since slipped toward our window.

YES · YEARS EARLY

Solar vs. onshore wind, from an early model

From a year-2000 model, the method projected the trajectory of solar becoming cheaper than onshore wind per MWh, matching empirical reality roughly two decades ahead of time (Benson & Magee, 2018).

  > TRY IT ON A TECHNOLOGY YOU KNOW
  >
  > ## Tell us the technology. We'll send you the rate.
  >
  > **Compare notes on a rate**
  >
  > You know a technology better than we do. Name it, add your own estimate of its annual improvement rate if you have one, and we'll send you ours alongside it.