Institutional investors, banks and insurers rely on physical climate risk data to understand how changing climate conditions will affect the assets underlying their portfolios. The problem is that physical climate risk data vendors, when evaluating the exact same asset, frequently reach different conclusions about which hazards pose the greatest risk, how severe those risks are, and sometimes whether meaningful risk exists at all.
Nik Steinberg
The 2025 benchmarking by the Global Association of Risk Professionals (GARP) demonstrated significant dispersion across thirteen vendors and highlighted several modelling differences contributing to that spread. Building on that work, the Investor Leadership Network (ILN) commissioned research to trace how certain methodological choices flow through the modelling chain, and translate into differences in physical and financial risk estimates.
The ILN methodological study asked seven vendors to assign climate risk metrics for the same dummy portfolio, containing three real assets, five listed equities, and two linear assets. They were also asked to respond to a 24-question methodological questionnaire, aimed at capturing the choices that drive their decisions, including how assets are identified, and how estimates of physical hazard exposure are translated into financial loss estimates.
Responses and scoring details were assessed across three dimensions: rigor, completeness, and transparency; and eight distinct methodological areas: Model Skill and Agreement, Spatial Resolution, Exposure and Materiality, Scenarios and Probabilities, Hazard Indicators, Vulnerability Indicators, Financial Metrics, and Physical Risk Reduction.
Across three real estate assets, vendors largely disagreed on which hazard ranked as the primary hazard. For listed equities, vendors identified vastly different corporate assets which they scored to arrive at equity-level climate risk scores. For an infrastructure asset, vendors diverged on whether a material flood risk was present at all. This can be seen in the figure below which shows how many vendors identified any inundation at 11 locations on the A4 toll road near Paris.
Count of vendors that indicate flood risk (blue = no flood, red = any inundation) across 11 sites of the A4 toll road near Paris.
A key takeaway from this exercise is that, although vendors differ widely in methods and findings, each brings a distinct view of how future climates will impact investable assets, and so long as methods are robust and well-justified, vendor differences may add richness and diversity to our view of an uncertain future. Conversely, methods lacking rigor and completeness could potentially lead to blind spots in screenings or even maladaptation if acted upon.
The study also surfaced several recurring blind spots. Most vendors focus narrowly on risks to individual assets, with limited attention to the compounding and cascading risks affecting communities, supply chains, and the operating environments surrounding those assets. Physical risk reduction guidance is similarly narrow, concentrated on engineering solutions at the asset level, with little attention to corporate governance, insurance strategy, or the potential benefits of closer collaboration with municipalities and local utilities.
The methodologies examined are generally suitable for screening and regulatory disclosure. For site-specific capital allocation, high-stakes decisions at the asset level require additional rigor: a deeper understanding of an asset's unique vulnerabilities, stress-testing against specific hazard thresholds, and supplementary site-level inspections. Even vendors that scored at the top of this evaluation should be assessed further before being relied upon for asset-level capital decisions.
The core contribution of this study is a questionnaire and criteria for investors to match their specific needs to the right vendor, or combination of vendors. It can be used directly by investors in vendor conversations to help surface the reasoning behind vendor choices, the assumptions built into their models, and the limitations they may not volunteer without being asked.
Rather than asking which vendor is best, investors should ask which vendor is best suited to their use case. A vendor quantifying climate impacts on local GDP may not be appropriate for asset-level risk quantification. A vendor with strong financial modeling may have gaps in hazard science. In the absence of standards, buyer education and vendor transparency are essential.
The study calls for voluntary vendor transparency as a near-term step and points toward voluntary participation in vendor ensembles, modeled on the Coupled Model Intercomparison Project (CMIP) in climate science. Under shared protocols, multiple vendor methodologies could be run side by side, allowing investors to see where outputs converge, where and why they diverge, and where deeper inquiry is warranted. For vendors, participation in such an ensemble would expand reach and strengthen credibility without compromising proprietary methods. Formal standards are unlikely to emerge fast enough to keep pace with how quickly this market is expanding. In the meantime, buyer education and vendor transparency are desperately needed.
A note on limitations: The seven vendors assessed in this study represent a subset of the broader marketplace, selected from those with prior engagement with ILN members. Scores were assigned on a relative basis and should be read as comparative indicators rather than absolute ratings. Evaluations reflect capabilities at the time of the study and are not endorsements of any vendor.
Nik Steinberg is a climate risk and adaptation specialist. He was Managing Director of Climate Research at Four Twenty Seven (now Moody's), Senior Advisor of Climate Risk for a USAID implementing partner, and now leads Vector Climate Group, supporting organizations as they navigate complex climate and nature-related risks and opportunities. He undertook the study for the Investor Leadership Network.