The materiality assessment came back material. The Scope 3 figure in last year's assured statement was overstated, and the incident runbook has done its job: the error is contained, the assurer is briefed, the corrected number is traceable. Now comes the act that decides whether the market reads this as competence or as a scandal. You have to restate a published figure, in public, and get it re-assured, without losing the credibility the whole disclosure rests on.

Restatement or Prospective Fix: The First Decision

Not every corrected number becomes a restatement, and getting this fork right is the difference between a proportionate response and either negligence or overreaction. A restatement is a formal correction of a previously published figure: you go back to the prior disclosure, change the number, and tell readers you changed it and why. A prospective fix corrects the approach going forward without reaching back to restate the prior period. The decision between them is driven by materiality, the same assessment the incident runbook produced, and it is a judgment for the disclosure and finance owners, informed by the assurer, never an improvisation.

Restate when the error is material to the prior published figure: it moves a total, a trend, an intensity ratio, or a target-tracking number in a way that would influence a reasonable user, or it touches a sensitive claim, a regulated CBAM value, or a figure the assurer specifically tested. Use a prospective fix when the error is genuinely immaterial to the prior period: correcting the method or the data flow going forward, documenting the change internally, and not reopening the prior statement. The trap in both directions is real. Restating an immaterial rounding difference erodes credibility by making the disclosure look unstable; declining to restate a material error is the move that becomes the greenwashing headline and the regulatory file. The protocol forces the decision to be documented, because the choice not to restate is itself something an assurer and a regulator will test.

A restatement handled openly is evidence that the controls work. A material error buried as a quiet prospective fix is evidence that they do not. The market forgives the first and punishes the second.

Why the Decision Belongs to Humans, With a Documented Basis

It is tempting, in a program that has automated so much, to imagine a rule engine that decides restatement automatically: error above X percent triggers a restatement, below it triggers a prospective fix. Resist this. Materiality in sustainability is not a single number; it blends a quantitative size test with qualitative judgment about trends, targets, sensitive impacts, regulated values, and what the assurer relied on. A five percent error in an immaterial category may not warrant restatement, while a two percent error that reverses a publicly committed decarbonization trend almost certainly does. The decision is a judgment made by the disclosure and finance owners, informed by the assurer, and recorded with its reasoning. The documentation is not bureaucracy; it is the defense. When a regulator later asks why the company did or did not restate, the answer is the memo: here is the size, here is the trend and target impact, here are the qualitative factors, here is the threshold we applied, here is the conclusion, here is who decided. A restatement decision without a documented basis is indistinguishable, from the outside, from a decision made to protect the narrative, which is exactly the suspicion a discloser cannot afford.

Re-Assurance: Bringing the External Assurer Back In

Here is the part that makes sustainability restatement different from a simple correction: the original figure was externally assured, so correcting it means the assurer's conclusion no longer holds for that number, and the corrected figure needs assurance too. This is re-assurance, and it is why the external assurer is a partner in the protocol, not a party to be presented with a finished result. With 73% of large global companies now obtaining external assurance, and most engagements at limited assurance and trending toward reasonable, the assurer's involvement is not optional politeness; it is what makes the restated number credible.

The re-assurance conversation, carried by the assurance liaison, covers several things. What exactly is being corrected and why. How the corrected figure was derived and whether it traces to evidence (an assurer will not re-assure a fix that is itself unsupported). Whether the correction affects the assurer's prior conclusion and what re-assurance procedures are now needed on the restated figure. And, crucially, whether the error reveals a control weakness the assurer must consider for the current engagement. The liaison's early, candid engagement, ideally beginning in the notify phase of the incident, is what turns re-assurance into an orderly procedure rather than a defensive scramble. The independence of the assurer is preserved throughout: the company restates and the assurer re-assures; neither does the other's job.

Scoping the Re-Assurance

Re-assurance is not automatically a full re-run of the entire engagement, and scoping it well saves cost and time without cutting the credibility that is the whole point. The scope follows the error. At minimum, the assurer re-assures the corrected figure itself. Beyond that, the scope expands to whatever the error touched or the assurer now reasonably questions: figures derived from the same corrupted input, other outputs from the same AI process, and, if the cause is a systemic control weakness, a broader look at the controls that were supposed to prevent it. Two factors push the scope wider. First, the assurance level: a reasonable-assurance figure carries a higher testing bar than a limited-assurance one, so its re-assurance is deeper. Second, the assurer's own risk assessment: if the error shakes their confidence in a class of numbers, they may widen substantive testing regardless of what you propose, and that is their call to make. The assurance liaison negotiates a proportionate scope, but the discloser does not get to dictate it, because an assurer who cannot scope independently is not providing assurance at all. A scoping note that lays out the affected figure, the cause, the remediation, and the periods and assurance levels involved gives the assurer what they need to plan proportionate procedures quickly.

What the Assurer Will Test in a Restatement

Expect the assurer to probe three layers. The corrected number itself: does it trace to evidence, with the right emission factor from an authorized source, the right primary-versus-secondary labeling, the right method. The cause: what let the wrong number through, and is it a one-off or a systemic control gap. And the process: did the organization detect, assess, notify, and correct with discipline, or did it stumble. A restatement that comes with a clean cause analysis and a documented, well-run process is far easier to re-assure than one that arrives as a bare corrected figure with no story of how the error happened and was caught.

Disclosing the Correction: What Readers Are Told

A restatement is a disclosure in its own right, and how it is written matters as much as the number. The correction disclosure should state, in plain language: what figure was restated, the original and corrected values, the reason for the correction, the period affected, and the fact that the corrected figure has been re-assured. It should not bury the change, spin it, or blame the tool. "The prior figure was overstated due to supplier data later corrected; the restated figure is X and has been re-assured" is a sentence that reads as control. "Refinements to our methodology" attached to a materially different number reads as evasion and invites exactly the scrutiny it hopes to avoid. Legal and the governance body shape the wording so it is accurate, complete, and consistent with any regulatory or market-disclosure obligation the correction triggers.

The Version Trail: Reconstructing the Correction

The credibility of a restatement rests on being able to show, definitively, that the correction was disciplined and the corrected number is right. That proof is the version trail, and it is the artifact that makes the whole protocol assurable. The version trail records the original figure and its evidence, the corrected figure and its evidence, the cause of the error, the materiality assessment and the restate-or-prospective decision with its basis, every approval, the assurer's re-assurance, the correction disclosure, and the dates for all of it. Where AI was involved, it records the model, prompt, source, and factor versions used for both the original and the corrected figure, so the change is fully explainable and the original number can be reconstructed as it stood at the time. A version trail is not paperwork for its own sake; it is the answer to every question a regulator can ask a year later, delivered as a folder instead of a defense.

A Worked Example: A Restatement and Re-Assurance Protocol in Action

Take the manufacturer from the incident lesson, whose Scope 3 Category 1 figure was overstated by a wrong-facility supplier datum ingested through an AI parsing step. The incident is contained and assessed material. Now the restatement and re-assurance protocol runs. The protocol is a short, ordered set of steps the governance body owns.

Step 1: Confirm the fork. The disclosure and finance owners confirm, in a documented memo, that the error is material to the prior published Scope 3 total and trend and touches a figure the limited-assurance engagement tested. Decision: restate, not a prospective fix. The basis is recorded.

Step 2: Derive the corrected figure with provenance. The team obtains the supplier's corrected, provenance-tagged activity data, re-runs the calculation with the authorized emission factor, and produces the restated Category 1 figure and the restated Scope 3 total. Every input traces to evidence; the fix is not itself an unsupported number.

Step 3: Engage the assurer for re-assurance. The assurance liaison, who alerted the assurer during the incident, now formally engages re-assurance of the restated figure. The assurer tests the corrected number, reviews the cause (a parsing step that mislabeled a wrong-facility datum as primary) and the remediation (a labeling verification control and governance-body review of the parsing use case), and confirms whether the control gap affects the current engagement. The assurer re-assures the restated figure.

Step 4: Draft and clear the correction disclosure. Legal and the governance body draft the correction: the figure restated, original and corrected values, the reason (supplier data later corrected), the period, and the fact of re-assurance, in plain language. It is cleared against any market or regulatory obligation.

Step 5: Publish and record the version trail. The restatement is published with the corrected figure and disclosure. The version trail is completed: original and corrected figures with evidence, the AI model, prompt, source, and factor versions for both, the materiality memo, the restate decision, the re-assurance, the disclosure wording, and every date and approval.

The result: a materially wrong assured figure was corrected in public, re-assured by the same external assurer, disclosed plainly, and left fully reconstructable. When an investor or a regulator asks about the restatement, the company hands over a folder that reads as competence. Contrast the credibility-losing path: the company reprints the number quietly as a methodology refinement, does not re-engage the assurer, and hopes. The assurer discovers the unassured change, the disclosure looks evasive, and a clean correction becomes the greenwashing story. Same error, same corrected number; the protocol is what preserved the credibility.

Cross-Framework Propagation and Reducing Repeat Risk

Two consequences of a restatement deserve their own attention because they are where a correction either ends cleanly or spreads. The first is cross-framework propagation. Most mature disclosers run one fact base into several disclosures: the same emissions figure feeds ESRS, ISSB, and, where relevant, a CBAM declaration. A material error in that shared figure is therefore not one error in one place; it is the same error waiting in every disclosure the figure fed. The disciplined move is to restate the underlying figure once at the fact base, then cascade the corrected value deliberately into each framework's disclosure, re-assuring each affected assured figure and recording the propagation in the version trail. The failure move is to fix it in the headline statement and forget that the intensity ratio in the ISSB filing and the embedded-emissions line in the CBAM declaration carried the same wrong number. A regulator who finds a correction applied in one framework but not another sees not a clean fix but an incomplete one, which reopens the credibility question the restatement was meant to close.

The second consequence is repeat risk. A restatement that corrects the number but not the cause is a restatement you will make again. This is why the assurer probes the cause, and why the protocol should not close until the root cause has been remediated through the governance body. If the error came from an AI parsing step that mislabeled data, the fix is a labeling verification control and a governance-body review of that use case, not merely a corrected figure. If it came from an ungoverned prompt or a silent model update, the fix is bringing that use case under change control. The restatement is the visible correction; the remediation is the invisible one that prevents the next incident. A function that treats every restatement as a lesson fed back into its controls restates less over time, and a declining restatement rate, paired with clean handling of the restatements that do occur, is exactly the credibility signal a board and an assurer want to see.

The Credibility Arithmetic

It helps to be explicit about why the open path wins, because under pressure the quiet path always looks cheaper. In the short term, a buried fix is less work: no re-assurance cost, no public disclosure, no difficult conversation with the board. But the arithmetic is asymmetric. If the buried fix is never discovered, the company saves a modest cost. If it is discovered, and assured figures tend to be discovered because an assurer is literally paid to look, the company faces the original error plus the concealment, which is now the story: not "a number was wrong" but "an assured number was wrong and the company hid the correction." The downside of concealment is unbounded and reputational; the downside of openness is bounded and procedural. A rational discloser, doing the expected-value math on a figure that a paid external party will re-examine, chooses the open path every time. The protocol simply makes the open path the default, so the decision is not left to a frightened team on a bad day.

Key Takeaways

• The first decision is the fork: restate a material error in a prior published figure, or use a prospective fix for a genuinely immaterial one, driven by the documented materiality assessment and informed by the assurer.
• Both directions have traps: restating an immaterial difference erodes credibility, and declining to restate a material error becomes the greenwashing headline, so the choice not to restate must itself be documented and defensible.
• Sustainability restatement is not a simple correction: the original figure was externally assured, so the corrected figure needs re-assurance, and the assurer is a partner in the protocol, not a party presented with a finished result.
• The assurer will not re-assure a fix that is itself unsupported: the corrected figure must trace to evidence, with the right factor from an authorized source and correct primary-versus-secondary labeling.
• Expect the assurer to test three layers: the corrected number, the cause of the error, and the discipline of the detect-assess-notify-correct process.
• The correction disclosure is a disclosure in its own right: state plainly the figure restated, original and corrected values, the reason, the period, and that it was re-assured, and never dress a material change as a mere methodology refinement.
• The version trail is what makes the restatement credible: original and corrected figures with evidence, the AI model, prompt, source, and factor versions for both, the materiality and restate decisions, the re-assurance, the disclosure, and every date and approval.
• A restatement handled openly is evidence the controls work; a material error buried as a quiet fix is evidence they do not, and the market forgives the first and punishes the second.