Can Sustainability Reporting Be Trusted If the Data Isn't?

In 2025, 69% of FTSE 100 companies adjusted previously reported climate or sustainability figures. Deloitte UK's analysis, published in July 2026, found that the proportion of companies making these adjustments had risen for the third consecutive year. Adjustments due to errors accounted for 20% of restatements, down from 26% the previous year. Greenhouse gas emissions accounted for 65% of the restatements, and 44 of the 59 companies that adjusted greenhouse gas figures made changes to their Scope 3 emissions. Deloitte's 2026 analysis also found that 51 companies made multiple adjustments.

These figures show why the reasons behind an adjustment matter. Some changes reflect better data, improved methods or changes in a company's business structure. Others correct errors. A revised figure doesn't automatically mean the original was wrong, but readers need to understand what changed and why.

Can someone trace a reported figure back to its source, explain how it was calculated and show what changed? That is what makes sustainability data credible.

The challenge is particularly visible in ESG reporting because the figures often draw on many different sources: invoices, fuel cards, GPS records, payroll and supplier data.

Follow one figure from its first record to the published report, and the challenge becomes easier to understand.

Where does a sustainability figure start?

It starts with a record created for another purpose.

An electricity invoice tells the finance team what to pay. A fuel-card statement helps the fleet manager reconcile monthly expenses. A GPS log records a vehicle's movements. Payroll records employee compensation.

Much of the information used in sustainability reporting already exists in operational systems. It was collected to support everyday business activities, not necessarily to answer questions about environmental impact or sustainability performance.

That is an advantage. Companies often have relevant data before they begin preparing a sustainability report.

The challenge starts when the information moves between systems and people.

What happens to the data on its way to the report?

Often, it passes through spreadsheets.

In PwC's 2024 Global CSRD Survey, 59% of 547 executives and senior professionals identified data availability and quality as a major obstacle to sustainability reporting. The survey also found that 74% used spreadsheets to process the data.

By the time a figure reaches the sustainability manager, it may have passed through several departments, each with its own files, processes and assumptions.

Consider what can happen along the way:

• An estimate is used because the March invoice arrived after the reporting deadline.
• A value recorded in kilowatt-hours is entered as megawatt-hours.
• A site is excluded because its meter reading arrived too late.

These are illustrative examples, but they show why the steps between the original record and the final figure matter. Someone may have a reasonable explanation for an estimate or adjustment, yet that explanation can disappear when the number moves into another spreadsheet.

The person reviewing the report a year later may see the figure without seeing the decision behind it.

Why can two people arrive at different totals?

Because the calculation depends on definitions as well as data.

Which vehicles belong in the reporting boundary? Should leased cars be included? If a warehouse is acquired in June, should its emissions be counted for the full year or only for the period after acquisition? Does “employee” mean headcount or full-time equivalent?

Two people can work carefully with the same records and still produce different totals if they apply different definitions.

A related problem appears in external ESG ratings. A study by Florian Berg, Julian Kölbel and Roberto Rigobon examined ratings from six ESG agencies and found that differences in measurement explained 56% of the divergence between their ratings. Differences in scope accounted for 38%, while differences in weighting accounted for 6%. The Aggregate Confusion research project at MIT Sloan explains how different measurement choices can lead to different assessments of the same organisation.

The study examined disagreements between rating agencies, not errors in companies' sustainability reports. Its findings nevertheless illustrate how different measurement choices can lead to different assessments of the same organisation.

When information is missing, rating agencies may also use proxy data - an alternative indicator intended to represent the missing information.

For companies reporting greenhouse gas emissions, the GHG Protocol Corporate Standard emphasises transparency and consistency. It calls for a clear audit trail and documentation of changes to data, inventory boundaries and calculation methods.

The principle is straightforward: when a figure changes, the reason and the effect of that change should remain visible.

What does a traceable figure look like?

Three things should accompany a reported figure: an owner, a source and a history.

An owner. Someone is responsible for the data set and can explain what it includes, which definitions apply and how exceptions are handled.

A source. The figure can be traced to the system or record from which it originated. Where possible, data is collected automatically on a defined schedule rather than copied manually between spreadsheets.

A history. The team can see how the figure was calculated, which assumptions were applied and what changed between reporting periods. Ideally, a reviewer can move from a dashboard total to the underlying records and identify the reason for an adjustment.

This is where automation and business intelligence (BI) have a practical role - in that order.

Automation can collect invoices, fuel-card data and GPS records on a schedule, reducing repetitive transfers between people and files. The rules used to map, combine or estimate values still need to be documented. Automation makes a process more repeatable; it doesn't make every input or assumption correct.

A BI dashboard can then present the resulting figures and, when designed to support it, let users examine the underlying records. Without a reliable data path, a polished dashboard may simply present spreadsheet figures in a more attractive format.

The final check is to compare independent records of the same business activity.

In one of our projects, we compared sales visit reports with ERP records and company-vehicle GPS data. Discrepancies were flagged automatically for the manager to review. The comparison gave the manager another source of evidence alongside the visits reported by the sales team.

The same principle could be applied to sustainability reporting. Fuel-card consumption can be compared with GPS mileage. If the relationship between the two changes unexpectedly, the team has a reason to investigate. The difference could reflect a genuine change in vehicle efficiency, incomplete records or an error in one of the data sources.

The comparison doesn't automatically identify the cause. It helps the team find the discrepancy while there is still time to investigate it, rather than discovering it during year-end reporting.

So, can sustainability reporting be trusted if the data isn't?

A sustainability figure is only as credible as the evidence behind it and the team's ability to explain how it was produced.

When the path from the original record to the reported figure is visible, an adjustment can be explained and assessed. When that path is unclear, readers have less evidence with which to judge the result.

The underlying records may already exist in the ERP, fuel-card files, GPS logs and other operational systems. The missing piece is often the documented connection between those records and the final report.

Start with one figure from the latest sustainability report. Trace it back to its original record. Write down the systems it passed through, the definitions and assumptions applied, and the changes made along the way.

Then ask:

Which figure in your last sustainability report would be hardest to trace back to its original source?