Climate risk in private holdings: why sector averages fail, and what they hide across the total portfolio

In brief

• The status quo: Asset owners and OCIOs assessing climate risk across listed equity, public debt, private debt and private equity get company-specific figures for listed holdings and sector averages for most private ones. The data behind a private company's product mix and site locations exists, but it sits scattered across annual reports and websites rather than in a structured disclosure.
• Upright's approach: Upright builds a company-specific climate risk figure for any company, disclosing or not, in four layers: products and services, verified sites, physical hazards per site, and effects on financial statements. Transition risk resolves at product level against five stakeholder-driven forces (regulators, investors, consumers, employees and suppliers); physical risk is computed per site for six hazards using NASA NEX-GDDP-CMIP6, ISIMIP, WRI Aqueduct and STORM across IPCC/NGFS-aligned scenarios.
• What it changes: One chain runs identically across public and private holdings, giving asset owners a comparable figure for every position and revealing correlated exposure that sector labels hide, such as a listed product line and its private upstream supplier sharing the same transition risk driver.
• In this article: Why private holdings end up with sector averages, the four layers that replace them, how a sustainability LLM makes each figure traceable step by step, and why one methodology across asset classes changes the diversification picture.

Why private holdings get a sector-average climate risk figure

There's a clear need for asset owners and OCIOs to get a complete climate risk view across the full portfolio: equity, public debt, private debt, and private equity assessed on the same basis. Historically, this has been difficult to achieve. Standard-setters have been building toward it: NGFS scenarios span both physical and transition risk, and major financial institutions and policy organisations publish guidance on multi-asset climate risk.

In practice, this is harder to achieve for private holdings than for public ones. For a listed utility, the transition risk figure can be based on disclosed emissions and a real product mix. For a private portfolio company, the same figure is usually a sector average. The underlying data just isn't available anywhere in a structured, granular form. It's scattered across annual reports, company websites, and other public sources, rather than filed once in a standard disclosure.

This is a commonly understood issue in the market: data quality and coverage for private equity and private credit lags behind listed markets, and existing engagement benchmarks were largely built for listed companies, without extending cleanly to other asset classes.

We see the next frontier in climate metrics as closing this gap directly: building an equivalent number for companies that don't disclose, rather than relying on a broader estimate.

How to model climate risk for a company that doesn't disclose: four layers

For a public company, much of the relevant information is already available: disclosed emissions, site locations, financial statements. For a private company, it typically isn't, or isn't in a structured, usable form. And for both, translating any of it into an actual effect on the company's financial statements is a separate challenge in itself. Closing the gap therefore starts with what a company actually does, and where it operates, rather than the sector it sits in. Four layers address this.

Products and services: transition risk at product level, not sector code

Transition risk is triggered by five stakeholder-driven forces (regulators, investors, consumers, employees and suppliers), each mapped to the financial line item it affects. This resolves at the product level against a structured taxonomy, rather than a sector code. As a simple example: two companies sitting in the same broad sector classification can carry very different transition exposure depending on what they actually produce, one facing direct carbon costs on its own output, another facing costs passed through its supply chain. A sector average doesn't distinguish between the two; product-level modelling does.

Sites: building a verified location list from public sources

Private companies rarely publish a consolidated list of where their operations sit, and without a location, physical risk can't be assessed at all. An automated identification and verification service builds this list from public sources: sector-specific search strategies surface candidate sites, an independent verification pass checks them, and each site is geocoded and tagged by supply-chain position (internal, upstream or downstream). Site data lands as a draft that a person reviews and publishes. The goal is a defensible, source-traceable site list built from scattered public information, rather than an unverified guess.

Physical hazards, resolved per site

Once a company has real coordinates, each site is run against six physical hazards: heat stress, floods, heavy precipitation, tropical cyclones, wildfires and drought. The inputs are established datasets (NASA NEX-GDDP-CMIP6, ISIMIP, WRI Aqueduct and STORM), run across IPCC/NGFS-aligned scenarios and multiple time horizons. Probability and magnitude are computed per site, not assumed at the country or sector level.

Financial statements: every risk as a P&L, balance sheet and cash flow effect

Every modelled risk and opportunity, physical or transition, is translated into its effect on the company's P&L, balance sheet and cash flow, using the financial statements already available to the investor, whether from public disclosure or from holding the company directly.

From a fixed score to a traceable calculation: what a sustainability LLM adds

A sector-average estimate is difficult to verify, since there's often nothing underneath it to check against.

This is where a purpose-built sustainability LLM changes what's possible. Rather than reporting an aggregate risk figure, Upright's sustainability LLM can walk through the calculation: which driver triggered the risk, what probability and magnitude were assigned and why, which financial line item it hits, over what time horizon, and how that resolves into a euro or dollar figure against the company's actual financial statement. If a portfolio manager asks why a specific holding is flagged as high risk, the answer is the calculation itself, not a fixed score.

That step-by-step visibility isn't only about being able to check the work. It means an investor can assess each step against their own in-house knowledge of the holding, a fact about the site, a detail about the product mix, something only they would know, and adjust the calculation accordingly, rather than accepting a fixed number with no way in.

Climate risk across equity, debt and private assets: why one methodology matters

An asset owner running equity, public debt, private debt and private equity side by side doesn't need five different methodologies stitched together with five different confidence levels. They need one consistent chain: products and services, sites, physical and transition risk across scenarios and horizons, and the effects on financial statements, applied identically whether the holding discloses everything or nothing.

This has a sharper implication than better reporting on its own. Take an illustrative case. A listed company's product line carries a transition risk tied to a specific driver, such as declining demand for a fossil-fuel-based product under tightening policy. A private supplier sitting upstream in that same value chain, making a component sold into that product line, is exposed to the identical driver, even though on paper it looks like an unrelated industrial company. An investor holding the listed company through equity and the supplier through a private fund might reasonably assume the two positions are diversified: different asset class, different manager, different vintage. In practice, both are exposed to the same transition risk driver, and would tend to move together if it worsens faster than expected. Seeing that correlation requires resolving risk at the product level, rather than treating both as unrelated points labelled with the same broad sector.

Closing the private-markets gap does two things at once. It gives investors visibility into holdings that were previously opaque, and metrics they can act on. And because the same methodology runs consistently across public and private assets, across equity and fixed income, it's what makes a cross-asset-class view possible in the first place: one answer for the whole portfolio, rather than five that don't reconcile.