Measuring the rate of change of impact rather than planetary proxies

For fifteen centuries, astronomers knew something was wrong: the planets sped up, slowed down, and even appeared to go backwards. The maths of the era kept failing.

The fix? Add more circles.

Epicycles were small circular orbits layered on top of the main ones, to patch the gap between theory and what the sky actually did. When they didn’t work, astronomers added more epicycles on top of epicycles. By the time Copernicus looked at it properly, the model had become a very sophisticated mechanism for being very precisely wrong. It produced numbers which looked like answers, but they weren’t.

In 1604, after 40 failed attempts, Kepler solved it by abandoning the assumption that orbits had to be circular at all. The answer was the ellipse. The fact you’re reading this is based on our understanding that planets orbit the sun in elliptical orbits.

We are doing exactly this with carbon accounting, and we’ve been doing it for years.

The problem

Most carbon accounting is built on spend data: take a financial figure from an invoice (e.g. what a company paid for electricity, goods or services), multiply it by an emission factor from an economy-wide table, or another abstracted model, and generate a number in tonnes of CO2e that looks like a measurement. However, it’s not really a measure, it’s a financial proxy in a costume, with a ‘choose your own adventure’ rulebook.

NB: this isn’t a criticism of the standards bodies. They are explicit that spend-based data is the entry point, not the destination (both the GHG Protocol and PCAF codify data quality hierarchies for exactly this reason). The problem is what happens in practice: the market often settles at the bottom of the hierarchy (as a compliance checkbox), and the financial sector treats the resulting numbers with a credibility they were never designed to carry.

Emission factors are national averages, often aggregated across entire sectors, sometimes years out of date, and blind to what was actually bought, when, from whom, made of what, using which energy mix. Two companies buying identical goods from the same supplier at different negotiated prices will report different emissions. Negotiate well and report lower emissions.

This yields an abstracted accounting process with a green label on it, rather than physics.

The intersection between the real economy and the financial economy is part of the challenge: we rarely create exchange rates between them. What is deemed ‘good enough’ for some are epicycles for others. MIT have carried out some excellent research showing, for example, the spreads on ESG score across the market for the same businesses.

Having spent 20+ years in carbon reporting, including aggregating over 40,000 carbon calculation methodologies, even when everyone is “using the GHG Protocol”, footprinting can deliver materially different values (double, half) depending on methodology choices. For regulators, that’s confusing, for banks, insurers and pension funds, it introduces risk: they need assurable, comparable data to allocate capital, and if they don’t have it at scale, sustainable finance will always be its own category, not core businesses.

And so the epicycles begin

Spend-based estimates too crude? … add supplier-specific emission factors. 
Suppliers don’t have them? … add industry-average adjustments. 
Reporting boundaries don’t match? … add scope allocation rules. 
Outputs not comparable? … add assurance frameworks. 
Frameworks can’t assure what isn’t measured?… add materiality thresholds.
Confused and finding all this too complex? … let’s get AI to solve it for us. 
GOTO start.

Each layer of correction is a response to the inadequacy of the layer beneath it and none fix the underlying problem. Instead they make the system more elaborate, more expensive and more ‘defensible’ (compliance-led rather than value-driven), while preserving and compounding errors.

The EU’s proposal to use eInvoicing data for SME sustainability reporting could be seen as the latest epicycle. While it is a genuine attempt to reduce burden, and will produce more granular numbers, it’s still a financial signal being used as a proxy for a physical one.

A “CO2e per unit” figure on an invoice line is only as good as the supplier’s own calculation, which is almost certainly itself spend-based or, at best, aligned with a standard but not harmonised in its actual implementation. Estimates referencing estimates, each with compounding error, processed through enough steps that the original uncertainty becomes invisible. The latest GHG process added provenance, to much aplomb, and while provenance is essential (and should have been mandated from the outset), without sorting out the source, it’s another epicycle that avoids addressing market incentives. Making numbers look more specific doesn’t make them more true.

Read the meter

Kepler’s insight wasn’t that the geocentric model needed better maths, it was that the geocentric model of the universe was the problem.

The equivalent insight in carbon accounting is quite simple: if you want to know how much energy a business has consumed, and at what carbon intensity, measure it, and read the meter. We can apply the actual grid carbon intensity at the time and place of consumption: the result should be as close to a physically grounded, continuously updated and independently verifiable figure as we can get. This can happen at scale for both electricity and gas, and when we create the right rules, can extend to water, transport and other resources that are harder to measure. Remember, though, we are instrumenting the world (there are more sensors connected to the internet than there are people).

This is what the Perseus project (https://ib1.org/perseus) is doing: half-hourly smart meter data, transmitted through the national smart meter programme, matched to grid carbon intensity at the time and location of use, and delivered through a permissioned, provenance-stamped data flow with a complete audit trail from meter to calculation.

It’s as close to the real physics-based data as we can get.

Why this matters

Banks need to underwrite green lending and sustainable finance, pension funds need to report financed emissions, insurers need to model transition risk, treasury needs to model our sustainable future economy.

None of these can be built on proxies without inheriting all the errors, the incomparability and legal risk that comes with them. As was once described to me by a very senior consultant in the sector, “a defensible financial decision needs more than a random number generator”.

The UK sustainable finance market should be around £40 billion, in terms of high-value, auditable outcomes I’d estimate today it’s more likely £1 billion. The end-game should be that sustainable finance calculations are baked into all relevant economic flows, as part of the core of loan books, not a category.

The gap is a failure in connected data flows, and a failure to link these to market incentives, and a failure to make it trivially easy for people to connect their own needs with measurable impact.

So what?

There’s a reason Ptolemy’s model survived fifteen centuries: it was internally consistent and it had the authority of tradition. It did give predictions people could act on (even if they were systematically wrong) and changing it meant abandoning assumptions so fundamental that most people didn’t recognise them as an assumption.

Spend-based carbon accounting has the same properties: it is embedded in reporting frameworks, audit standards and corporate processes. It has the authority of standards bodies and decades of practice. The assumptions this rests on (that financial flows are an adequate proxy for physical ones) is so foundational it’s rarely examined. It wasn’t ‘wrong’ to say “we need something to act upon” or “don’t let the perfect be the enemy of the good” but I argue that we’re past that point now: “we should not let the mediocre be the enemy of the good”

Perseus is the ellipse: not because it’s simpler (it demands real data flows, real permissioning, real provenance infrastructure) but because it starts from what’s actually measured: emissions are physical phenomena and physical phenomena can be measured. A measurement is worth more than any number of corrections applied to a proxy.

We don’t need more epicycles: we need to read the meter.

You can join IB1 as an individual (for free) or as a commercial member: https://ib1.org/join

If you’d like to share this on

• LinkedIn, please repost this.
• Medium, please share this