Why Better Data Can Make Your Carbon Footprint Go Up

There is a tendency to think of a corporate carbon footprint as a known quantity: something that exists and simply needs to be calculated. For much of Scope 3, that isn't really the case.

An organisation may know exactly how much it paid a supplier, but it is unlikely to have directly measured all the greenhouse gas emissions associated with the goods or services it purchased. Instead, those emissions have to be estimated using the best information available. Depending on the category and maturity of the data, that might involve physical activity data, supplier-specific information, industry averages, spend-based emission factors or a combination of different approaches.

The resulting footprint is therefore influenced by two things: what the organisation actually did and how well those activities were measured.

That distinction becomes increasingly important as a carbon inventory matures. The methodology used in the first year of reporting is unlikely to remain unchanged forever. Better supplier information becomes available, classifications improve, new data sources are introduced, emission factors change and mistakes are identified. These are all signs of a maturing carbon-accounting process, but they can also change the reported footprint.

Our experience includes a multi-year rebaselining exercise involving more than a million procurement records across different reporting periods, suppliers, business units and expenditure categories. Rather than simply accepting previously reported results, the work involved going back into the underlying data and testing whether the assumptions and classifications supporting those results were still appropriate.

We reconciled inconsistent supplier identities, reviewed supplier and expenditure classifications, examined category mappings, updated aspects of the methodology and investigated the suppliers and activities responsible for the largest calculated emissions. Rather than attempting to manually review every record, the analysis focused heavily on materiality: identifying where a classification, assumption or data-quality issue was significant enough to meaningfully affect the overall result.

As those improvements were made, the historical footprint changed.

That does not mean the organisation's historical environmental performance suddenly changed. We weren't changing what had happened. We were improving the model used to estimate the emissions associated with what had happened.

It sounds like a subtle distinction, but it becomes extremely important when the resulting numbers are being used to measure progress against a baseline.

Consider a simplified example. An organisation has £50 million of historical spend with a major supplier. The supplier has previously been mapped to a broad industry classification, and the corresponding spend-based emission factor has been used to calculate its emissions.

Further analysis shows that the original classification wasn't particularly representative of what the supplier actually provided. A better classification results in a materially different emission factor.

Nothing about the £50 million of historical expenditure has changed. But applying a more appropriate factor to such a large value can significantly alter the calculated emissions.

This is one of the challenges of working with large corporate datasets. An assumption that looks relatively minor at an individual-record level can become significant when it is applied repeatedly across millions of transactions or large amounts of expenditure. Equally, there may be thousands of imperfect classifications elsewhere in the dataset that have almost no impact on the overall footprint.

This is why materiality is far more useful than pursuing theoretical perfection. The objective should not be to eliminate every imperfect data point. It should be to identify where poor data or weak assumptions could materially change the conclusions being drawn from the inventory.

Classification isn't the only reason an improved footprint might increase. Better coverage can have exactly the same effect.

Suppose an organisation captures 80% of the relevant activity in its first inventory. Improvements to systems and data collection increase that to 95% the following year. Reported emissions could rise substantially even if the organisation's underlying activities remained broadly unchanged.

If Year 1 is reported as 500,000 tCO₂e and Year 2 as 575,000 tCO₂e, the obvious interpretation is that emissions increased by 15%. But some or all of that increase might simply represent activity that existed in Year 1 and was never captured.

The same issue arises when organisations move from broad expenditure assumptions to more representative activity data, obtain better supplier-specific information or improve the allocation of emissions between categories. Better information doesn't inherently produce a lower carbon number. It produces a number that should better represent reality.

Sometimes that number will be lower. Sometimes it will be higher.

This leads to a distinction that needs much more attention in carbon reporting: measurement change is not the same as emissions change.

Real-world emissions change occurs when something actually changes in the organisation or its value chain. Perhaps less material was purchased, employees travelled less, a supplier changed its manufacturing process, renewable energy replaced fossil-fuel generation, or a lower-carbon product replaced a higher-carbon alternative. Those are genuine changes in environmental performance.

Measurement change occurs when the way we calculate the footprint changes. This could result from better supplier classifications, increased data coverage, different emission factors, improved activity data, corrected errors, changes in allocation, changes to the reporting boundary or a more appropriate calculation methodology.

Both can move the reported carbon footprint. Only the first necessarily represents decarbonisation.

This means that when an organisation says its Scope 3 footprint has fallen by 20% since its baseline year, the percentage alone doesn't tell us enough. We need to understand how much of that movement came from genuine changes in activity or carbon intensity and how much came from changes in the underlying measurement.

The same scrutiny should apply when emissions increase. An organisation should not necessarily be considered to have gone backwards simply because better information resulted in a higher estimate.

This is where improving carbon data creates a practical reporting problem. If a significant improvement changes the way current-year emissions are calculated, should that improved methodology also be applied to the historical baseline?

In some cases, yes.

Suppose an organisation introduces a significantly better classification methodology in Year 5. If Years 1 to 4 continue to use the old methodology while Year 5 uses the new one, the resulting trend is no longer entirely comparable. The movement between the years now reflects both changes in actual emissions and changes in how those emissions were calculated.

Where the effect is significant, historical emissions may need to be recalculated using the improved methodology. This is the purpose of rebaselining: not to rewrite history, but to maintain a meaningful comparison with it.

However, better data should not automatically mean rebaselining. Carbon datasets are continually evolving. Supplier classifications will be corrected, emission factors will be updated, new information will become available and minor errors will be identified. Recalculating years of historical emissions every time this happens would be impractical and would create a baseline that never stops moving.

This is why organisations need a clearly defined materiality or significance threshold within their base-year recalculation policy.

A materiality threshold provides a practical test for deciding whether an improvement is significant enough to justify changing historical results.

If correcting a supplier classification changes the total footprint by 0.01%, rebuilding several years of historical data is unlikely to improve anyone's understanding of performance. The correction can be documented and incorporated appropriately without necessarily reopening the entire baseline.

If improved classifications, additional coverage or a methodological correction materially changes the baseline, however, maintaining the old number may actually make the reported trend less credible. At that point, rebaselining becomes important because the organisation is effectively comparing two different measurement systems.

The threshold also needs to be applied consistently. It should not matter whether the methodological improvement increases or decreases reported emissions. If a change is considered material when it increases the footprint, the same principle should apply when it reduces it.

Otherwise, there is an obvious risk of selectively accepting methodological changes that improve reported performance while resisting those that make the trajectory look worse.

There is also a cumulative issue to consider. Individual corrections may each fall below the threshold but collectively become significant. A sensible recalculation policy therefore needs to consider the combined effect of methodological and data-quality changes rather than assessing every correction entirely in isolation.

Materiality gives organisations a way to improve their data without creating an expectation that every improvement requires the historical inventory to be rebuilt. The question becomes not simply 'Has the data changed?' but 'Has it changed enough to affect our understanding of performance?'

Ultimately, the purpose of a baseline is to provide a consistent reference point against which progress can be measured. If the methodology used to calculate current emissions becomes materially different from the methodology underpinning that baseline, the comparison starts to lose meaning.

This is why rebaselining should not be viewed as an administrative inconvenience or an admission that the original footprint was 'wrong'. Carbon accounting relies on estimates, and those estimates should be expected to improve as organisations obtain better information.

What matters is having the governance to manage those improvements properly: a documented methodology, a clear recalculation policy, an established materiality threshold and enough data lineage to understand what changed and why.

There can be an uncomfortable incentive here. Once a baseline has been published and reduction targets have been announced, organisations naturally want the trajectory to move downwards. But the purpose of carbon accounting cannot be to protect the desired trajectory.

If better evidence shows that historical emissions were higher than previously estimated, that should be reflected where it is material. Equally, if improved methodology reduces the historical footprint, that reduction shouldn't automatically be presented as decarbonisation.

As corporate carbon accounting matures, footprints will continue to change. Supplier information will improve, data coverage will expand, classifications will become more granular, methodologies will develop and new technology will make previously inaccessible information usable.

Some of those improvements will make reported emissions fall. Others will make them rise. Neither outcome, by itself, tells us whether the organisation's environmental performance has improved.

The important question is whether the change came from the real world or from the way we measured it.

A credible carbon-accounting system needs to distinguish between the two, improve the underlying data where it matters, and rebaseline when those improvements cross an established materiality threshold.

The purpose of carbon accounting isn't to produce a number that moves in the right direction. It is to produce information reliable enough to tell us whether the organisation actually is.

Sometimes a higher carbon footprint is evidence of better carbon accounting, not worse environmental performance.