Using financial materiality in sustainability is about Time and Context
This article is the seventh in a series exploring the role of trustworthy AI in mainstreaming sustainable investing.
For the past three months, we have been working on proof-of-concepts with corporate reporters, assurance providers and institutional investors. We are assessing whether a data-driven approach to financial materiality can solve the real pain points experienced every day by users of sustainability reports.
We have found two hard problems. First, users of financial materiality assessments connect those reported topics to specific balance-sheet items and model how their individual performance could contribute to the company’s value. Because audited balance sheet is published annually, this analysis is inherently out-of-date by the time it is produced. We call this the Time element.
Second, users understand a company's performance in terms of alpha (how an investment performs relative to a benchmark) and beta (how much its price moves compared with the broader market). So we need to consider the gravitational effects of industry and benchmark systematically so we know what's uniquely material to the company. We call this the Context element.
We have also found that both elements need to be solved together. Solving for the Time element allows our data-driven financial materiality assessment to become more Predictive but only of the company in isolation. Without considering industry peers and index companies, we could end up being precisely wrong. And solving for the Context element gives our data-driven financial materiality assessment much greater Explanatory power but only that of 20/20 hindsight. Without considering the trends over time, we could end up being broadly wrong.
So solving for both Time and Context together allows our financial materiality assessment to be genuinely Insightful. That is, both precisely and broadly correct. With apologies for the driving metaphor (my daughter has just passed her driving test so somewhat top-of-mind), we examine each element below.
Looking in the rear view mirror (Time)
Financial materiality is fundamental to an investor's decision-making in allocating capital. It allows the investor to focus on those topics that impact a company's financial condition, cash flows, or enterprise value. Typically, this is a three-step analysis:
Identify material issues. Taking what the company has stated in their annual report as financially material, compare with last year and perhaps a small peer group.
Quantify financial impacts. Connect that material issues with revenue items reported in the annual, audited balance sheet. Calculate potential costs associated with those items. Adjust asset lives/write down value of tangible assets that might be affected. Assess if material risks increase the cost of borrowing money.
Adjust valuation models. Add/subtract expected costs and revenues in your discounted cash flow (DCF) model. Raise/lower the discount rate according to assessed risks. Compare valuation multiples against peers with better/worse risk profiles.
The problem is that this analysis is always historic so the user is forced to make projections based on out-of-date data. In the first step, the material issues are identified by the company's stakeholders using last year's performance, their own previous experience, and expertise gained in the past. In the second step, the revenue items reported are 6 - 9 months out-of-date due to the data-gathering, review, audit and publication process. In the third step, even the best DCF models compound the problem by importing comparative valuation multiples that have been calculated using data even further in the past.
Of course, a DCF model is deployed to overcome the historic nature of the data. But its core components can be improved with a more robust material topics determination methodology, a more dynamic mapping of material topics to the balance-sheet items and other items, and more up-to-date data allowing for a near-real time risk profiling of peers and benchmarks.
Using your wing mirrors (Context)
To accurately assess which topics are financially material to a company, it is vital to know whether those same topics are material to the company's peers, industry, benchmark. Taking the British oil and gas company BP as an example, if the GHG Emission topic is material to the company, European oil and gas majors, and the FTSE100 benchmark alike, then there is little useful information here. But, if the Employee Engagement topic is material to BP but not its peer group, nor to its benchmark, then this is useful information to the investor. Rising sentiment on this topic indicates real improving performance on associated balance-sheet items. But declining sentiment on this topics indicates the reverse.
Gathering comparable data from other companies in the peer group and benchmark is a laborious, complicated, and awkward task. While some improvements are being seen in terms of comparability, for example the work of Sustainability Reporting Navigator, the source material is still the published annual report, and so falls foul of the Time element discussed above. To stretch the driving analogy, using your wing mirrors is certainly additive but you remain vulnerable to other vehicles emerging out of your blind spots.
A second-order problem arises if we are able to accurately and continuously collect which topics are reported as financially material in annual reports. How are we to calculate what is material for a peer group of, say, 30 companies in the company's industry? And to calculate what is material for the hundreds of companies in a standard market benchmark? There may be a weak signal in the industry grouping if the same topic reappears but at what frequency? The benchmark has the same problem but at a larger scale. And then, how do we compare the material topics between company, peer group and benchmark in a robust way?
Making progress with a data-driven approach
To make progress here, we are running a number of proof-of-concept projects with corporate reporters, assurance providers and institutional investors. We are taking a data-driven approach to financial materiality to see if we can solve for the Time element and Context element at the same time.
In a previous article, we looked at using closing day share price as the least worst option for a Dependent Variable against which we can correlate all the publicly-available sustainability conversation in the market about individual companies. We looked at the relationship between share price and the three measures of financial performance highlighted by IFRS S1: an company's cash flows, access to finance, and cost of capital. We also looked at what drive short- and long-term share price movements.
With this data pipeline in place, we have been able to make progress on the Time element by significantly improving the inputs into DCF models. In particular:
Users no longer have to simply take what the company has stated in their annual report as financially material. Using an algorithmic approach, the materiality determination can be repeated on a weekly basis to generate a dataset that captures the dynamic nature of materiality and so continuously improve the assumptions in the DCF model.
With up-to-date materiality determination, combined with more frequent metrics like monthly retail sales or production/delivery data instead of last year's balance sheet, the predictive capability of a DCF model is improved considerably.
The analyst can run multiple peers with better/worse risk profiles to more accurately gauge the target company's discount rate. This analysis can be performed monthly instead of waiting for an annual report to be published.
In another previous article, we looked solving materiality's "three-body problem" in order to compare the company in question to a peer group and a benchmark to give context. This is the mathematical study of how three celestial objects (like stars or planets) interact purely through their gravitational pull. While a "two-body" system (like the Earth and moon) has predictable orbits, adding a third body (like the Sun) creates a chaotic, unpredictable system with no general mathematical formula to solve it. In our terms, the reporting company is the Earth, the peer group is Moon and the benchmark is the Sun.
With this data pipeline in place, we have been able to make progress on the Context element by significantly improving the breadth of the DCF models. In particular:
The model isolates those ESG topics that are uniquely material to the company from topics that are shared with the peer group and/or benchmark. This improves the quality of the materiality determination by avoiding false positives and suspicions of greenwashing/greenhushing.
Isolating company-unique ESG drivers also enables the DCF model to distinguish between alpha and beta effects, significantly improving the accuracy of the output.
Importantly for audit purposes, this approach isn't a black box. The entire framework is a convex optimization, variational calculus, and linear algebra with transparency at each step. We will publish more on the methodology next month.
BP example
Returning to our example of the British oil and gas company BP, we ran a 52-week window to 24 July 2026 using the STOXX European 600 Oil & Gas UCITS ETF as the peer group and the FTSE 100 index as the benchmark.
We found that the sector dominates nearly three-quarters of BP's log-returns variance. This result was significant with p-value < 0.05.
However, the index exerts no significant pull on BP's weekly stock price.
Only a quarter of BP's weekly stock price movement is attributed to sustainability drivers unique to BP.
That allocation comprises 5 drivers: Employee Engagement, Human Rights & Community Relations, Energy Management, Employee Health & Safety, and Critical Incident Risk Management.
When there are no major company-specific events, for example between 2016 and 2017, movement in BP's share price is explained by the sector and index and the model returns fewer unique ESG drivers. But when there are frequent major company-specific events, for example during the energy crisis in 2023 and 2024, the model immediately returns a number of specific active drivers.
This information is useful for a reporter preparing/updating a company sustainability report, an analyst gauging the target company's discount rate, and an assurance provider auditing the results of a financial materiality assessment.
An industry standard?
Maxwell Data and the Department of Mathematics at Brunel University of London will present the findings of the proof-of-concepts at an industry workshop at UCL Centre for Sustainable Business on 18th September. If there is consensus, we will work with stakeholders to roll out a data-driven approach to financial materiality for sustainability matters as an industry standard ahead of reporting season in Q1 2027.