ifrs s2 in south africa: the disclosure is a data problem.

§1 · the question.

three regulatory processes are converging on the same south african number. the first is a disclosure standard: ifrs s2, the global baseline for climate-related disclosures, effective internationally since january 2024 and moving toward adoption here. the second is an assurance standard: issa 5000, already adopted by south africa's audit regulator, effective for periods beginning in december 2026. the third is existing law: mandatory facility-level greenhouse-gas reporting with criminal penalties for false information, a carbon tax assessed on the same numbers, and a carbon-budget regime whose first commitment period begins in 2026.

none of these, on its own, forces a south african company to publish an assured, investor-grade climate disclosure today. together they describe a trajectory with one destination: an emissions figure that is published, challenged, and defended — to an assurance provider, to a regulator, to investors reading the annual report.

the question this paper asks is narrow: when that destination arrives, what will it demand of a company's data estate? the answer, argued from the standards' own text, is that ifrs s2 is a data problem before it is a reporting problem. the reporting layer — the document, the tables, the disclosures themselves — is the easy part. the hard part is that every disclosed figure presupposes a chain of custody from source reading to published number, and most data estates were not built to keep one.

§2 · what ifrs s2 actually requires.

ifrs s2 was issued by the international sustainability standards board in june 2023 and is effective for annual reporting periods beginning on or after 1 january 2024, with earlier application permitted only if ifrs s1 — the general requirements standard that governs how sustainability disclosures are prepared and presented — is applied at the same time. it asks for disclosure across four content areas: the governance processes, controls and procedures used to monitor and oversee climate-related risks and opportunities; the strategy for managing them; the processes used to identify, assess and prioritise them; and performance — metrics and targets, including progress against any climate-related target the entity has set or is required to meet.¹

the metrics pillar is where the data obligation concentrates. an entity discloses its absolute gross greenhouse-gas emissions generated in the reporting period — scope 1, scope 2 and scope 3, subject to materiality — measured in accordance with the ghg protocol corporate standard unless a jurisdictional authority requires a different method. 'gross' means before any netting against carbon credits; 'absolute' means tonnes, not intensity alone. scope 3 is defined by reference to the ghg protocol's corporate value chain standard: an entity must consider all fifteen categories of value-chain emissions, determine which are relevant to its business model, include the relevant ones in its measure, and disclose which categories that measure includes. it cannot confine its measurement to the ghg protocol's minimum category boundaries; the issb standards require the entire value chain, upstream and downstream, to be considered.²

materiality does real work here, but less than preparers sometimes hope. the test is whether omitting, misstating or obscuring the information could reasonably be expected to influence the decisions of the primary users of the report — and it applies category by category, not to scope 3 as a whole. an exclusion is a decision that has to be made, documented, and defensible, which is itself a data exercise: you cannot show a category is immaterial without enough information about it to support the claim.

for scope 3 the standard goes further than naming categories: it prescribes how the measurement itself must be assembled. the scope 3 measurement framework requires an entity to prioritise its inputs and assumptions by four characteristics — data based on direct measurement; data from specific activities within the entity's own value chain; timely data that faithfully represents the jurisdiction of, and the technology used for, the activity; and data that has been verified. and — the requirement most preparers miss — the standard requires the entity to disclose the extent to which its scope 3 figure is measured using inputs from specific activities in its value chain, and the extent to which those inputs are verified. the standard does not just ask for the number; it asks you to publish how good your data is. the counterweight is a proportionality mechanism — an entity uses all reasonable and supportable information available at the reporting date without undue cost or effort — which limits how far you must go to collect, but does not excuse what you chose not to keep.³

targets get the same treatment. s2 does not require an entity to set an emissions target; it requires disclosure about any target the entity has set or is required by law to meet — which gases and which scopes it covers, how it was set and is reviewed, performance against it, and whether it is gross or net. a net target triggers more: the associated gross target must be disclosed separately, along with the entity's planned reliance on carbon credits, the scheme the credits come from, their type, and the factors needed to judge their credibility.⁴ every one of those disclosures is a time series someone must be able to regenerate.

the standard offers reliefs for the first year of application — an entity need not disclose scope 3 in its first annual reporting period, and may continue using the measurement method it used immediately before adoption — but they are reliefs of timing, not of architecture; the obligations arrive in year two regardless.⁵

the standard is also still moving. in december 2025 the issb issued targeted amendments to s2's greenhouse-gas requirements — permitting entities to limit scope 3 category 15 measurement to financed emissions as defined, permitting industry-classification systems other than gics for disaggregating financed emissions, and clarifying the jurisdictional relief — effective for periods beginning on or after 1 january 2027, with early application permitted.⁶,⁷ for a preparer, the lesson in the amendments is not their detail; it is that the measurement rules will keep being clarified, and a data estate hard-wired to one year's interpretation will be reworked more than once.

§3 · where south africa stands.

south africa has no mandatory ifrs s2 date. as at june 2026, no instrument compels a south african company to publish s2-grade disclosures. what exists is a voluntary layer, a regulatory trajectory, and — less noticed — an existing body of emissions law that already carries penalties.

the voluntary layer belongs to the jse. in june 2022 the exchange published its sustainability disclosure guidance and a companion climate disclosure guidance — voluntary best-practice documents for listed companies, with a basic set of metrics rooted in established global standards: gri, tcfd and the then-emerging issb baseline.⁸,⁹ in september 2024 the jse said it was revising that guidance to align with the finalised ifrs s1 and s2, its chief executive describing the ifrs standards as 'increasingly becoming the dominant international framework'; no mandatory timeline was attached.¹⁰

the trajectory runs through the regulators. the ifrs foundation has been engaging south african leadership — treasury, the jse, the directors' institute, the integrated reporting committee — since at least march 2024.¹¹ the financial sector conduct authority's march 2025 sustainable-finance report records commissioned research into climate disclosure requirements for retirement funds, collective investment schemes and large listed corporates, and an intention to align any requirements with ifrs s1 and s2 as the global baseline.¹² in july 2025 the fsca signed a cooperation agreement with the international finance corporation — described as the first formal step by a south african regulator toward issb adoption — covering a jurisdictional roadmap, regulatory frameworks tailored to local conditions, and capacity-building for supervisors and regulated entities.¹³ commentary around the agreement expects an eventual mandate to land first on jse-listed companies, public-interest entities such as banks and insurers, and companies above size thresholds.¹⁴ south africa has not yet published a jurisdictional profile with the ifrs foundation — seventeen jurisdictions had by june 2025 — which is an honest measure of where the process stands: moving, not arrived.¹⁵

the least-discussed part of the picture is the law that already exists. the national greenhouse-gas emission reporting regulations (2016, as amended) require facility-level reporting to the environment department: annual submission by 31 march for the preceding year, and — the clause that matters most for this paper — retention for at least five years of all underlying data, calculations, algorithms and procedures used to estimate emissions, available to the authority on request. providing false or misleading information is a criminal offence, with fines up to r5 million on first conviction and r10 million thereafter.¹⁶ the carbon tax has been assessed on substantially the same reported numbers since june 2019.¹⁷ phase two of the tax is in design,¹⁸ and draft carbon-budget regulations published in august 2025 under the climate change act contemplate company-level carbon budgets, mitigation plans and annual reporting from a first commitment period beginning 1 january 2026.¹⁹

so the common framing — 'climate disclosure is voluntary in south africa' — is true only of the investor-facing wrapper. the emissions number itself already carries legal weight: it is reported under penalty, taxed, and about to be budgeted against. ifrs s2 does not create the obligation to get the number right. it adds an audience that is allowed to ask how you got it.

there is one pull from outside. the eu's corporate sustainability reporting directive reaches south african groups and exporters through customers and subsidiaries — but the december 2025 'omnibus' agreement narrowed it. a third-country group is now in direct scope only above roughly €450 million of eu turnover with a substantial eu subsidiary or branch, and in-scope companies face limits on the value-chain data they may demand from suppliers under 1,000 employees.²⁰,²¹ the direct pull on mid-size south african suppliers has weakened accordingly; what remains is the indirect version — large south african groups in scope through their eu operations, and large suppliers fielding the data requests the omnibus still permits.

§4 · the assurance trajectory.

disclosure standards say what to publish. assurance standards decide what survives. in november 2024 the iaasb issued issa 5000, the first comprehensive global standard for sustainability assurance engagements — usable across topics and reporting frameworks, by accountant and non-accountant assurance practitioners, effective for engagements on information covering periods beginning on or after 15 december 2026.²²

south africa did not wait. the independent regulatory board for auditors has adopted issa 5000 for south african engagements, with early adoption encouraged,²³,²⁴ and has begun issuing practice guidance on sustainability assurance engagements.²⁵ read that sequence carefully: in south africa the assurance standard is in force before the disclosure mandate exists. whenever mandatory s2-aligned disclosure arrives, the profession that will test it is already standing on its standard.

early adoption cuts both ways. a company that volunteers for assurance now buys itself a rehearsal — findings arrive while they are still private, and the estate gets fixed before the disclosure is mandatory and the findings are not. a company that waits inherits the opposite sequence: its first assured disclosure is also its first test of whether the trail exists, in public, on a deadline. the first cohort of voluntarily assured south african disclosures will also quietly set the local bar for what 'sufficient evidence' looks like — a bar later entrants will be measured against without having helped to set it.

what does an assurer test? not the report. an assurance engagement on an emissions figure works backwards along the figure's construction: the boundary — does the organisational boundary match the consolidation approach the methodology claims? the method — which emission factors, which version, applied when, approved by whom? the source data — which meter, which invoice, which supplier return, and can the entity produce it? completeness — were all fifteen scope-3 categories considered, and is every exclusion documented? and consistency — does the emissions boundary reconcile to the financial statements the disclosure sits beside? limited assurance demands less evidence than reasonable assurance, and the global direction of travel is from the first toward the second; jurisdictions that mandate disclosure tend to phase assurance upward behind it. but both levels demand that the trail exist. a figure that was assembled once, in a spreadsheet, by a person who has since left, does not have a trail. it has a story.

§5 · the data each disclosure presupposes.

walk the s2 metrics backwards into a south african operation and each one lands on specific systems.

scope 1 is fuel and process. the figure presupposes procurement records for every litre of diesel and every ton of stationary fuel, fleet and generator logs, process measurements where the chemistry itself emits, and stock reconciliations that close the gap between fuel bought and fuel burned. for facilities already reporting under the national regulations this is not a new duty — the five-year record-keeping obligation attaches to exactly these records — but s2 raises the audience from one regulator to every reader of the annual report.

scope 2 is electricity. the figure presupposes utility and municipal invoices, meter readings that reconcile to those invoices, and the contractual layer — wheeling agreements, renewable procurement, embedded generation — that determines what story the company may credibly tell about its purchased power. south africa adds a wrinkle of its own: the load-shedding years moved real consumption off the grid and onto diesel generators, and by the scope definitions that movement reclassifies the emissions — purchased electricity is scope 2, fuel burned in your own generator is scope 1. a company whose generators carried it through 2023 has a structurally different emissions profile from its 2019 baseline, and the evidence for exactly how different lives in fuel procurement records and generator run-logs. if those were treated as operational exhaust rather than data, the reclassification cannot be reconstructed.

under everything sits the factor layer. an emissions figure is activity data multiplied by an emission factor, and factors have vintages: the grid factor changes as the generation mix changes, supplier factors are revised, the published sets are updated. which factor, which vintage, applied to which period is part of the method — and when a factor is restated, every figure built on it needs to be re-derivable. an estate that stores only the multiplied result, and not the factor and activity data separately, cannot answer the simplest assurance question there is: show me the calculation.

scope 3 is everyone else. fifteen categories considered means procurement ledgers mapped to supplier data or emission factors, inbound and outbound logistics, business travel, the use and end-of-life of sold products, investments. the data arrives from counterparties, in their formats, on their reporting calendars — the standard explicitly permits using value-chain data from a different reporting period, provided it is the most recent available without undue cost or effort, covers a period of the same length, and significant intervening events are disclosed. that is an honest concession that this data lives on other people's calendars; it is also one more thing the estate has to track. and because the measurement framework requires prioritising direct, primary, verified data — and requires disclosing how much of the figure rests on each — the quality of the pipeline is not an internal matter. it is published.

group structure multiplies all of it. the reporting entity for sustainability disclosures is the same entity as the financial statements — a parent disclosing on its consolidated group discloses climate information for that group — but the emissions boundary is then drawn by the ghg protocol's measurement approaches, equity share or control, and the approach chosen must itself be disclosed with reasons. scope 1 and scope 2 must additionally be disaggregated between the consolidated accounting group and other investees — which presupposes the estate can split emissions along the same lines the cfo consolidates. subsidiary figures must arrive on a common method, a common period and a common format, or someone reconciles them by hand.

targets add a time dimension: a baseline, a method that must stay comparable year over year, and performance disclosed against each target with an analysis of trends. an acquisition or a disposal, perhaps surprisingly, does not restate prior-year comparatives — comparative emissions remain those of last year's reporting entity, as constituted in last year's financial statements — so the current figure and the comparative can describe two different organisms, and the divergence has to be explained rather than papered over.

now name the failure mode, because every reader who owns this number will recognise it. the figure is assembled once a year, near the deadline, in a spreadsheet, from csv exports out of a dozen systems and emailed returns from subsidiaries that each report in their own format. corrections overwrite the cells they correct. the person who built last year's workbook has moved on. nothing records which meter reading, which invoice, which factor produced the published tonne. under a voluntary regime this is survivable, because nobody is entitled to look. under the reporting regulations it is already a latent offence if the submission proves false. under an assurance engagement it is a finding. under an investor's challenge it is a credibility event.

the pattern is general: the disclosure is annual, but the data is continuous. an estate that only materialises the figure at year-end has no way to defend it, because the defence — lineage — has to be captured at the moment each number is born, not reconstructed at the moment it is questioned.

§6 · what assurance-ready means.

strip the standards to their operational core and an assurance-ready emissions estate does five things.

first, lineage at figure level: every published tonne traceable, by reference, to the source reading, invoice or supplier return it came from, through every transformation in between. second, methodology before measurement: factors, boundaries and category-relevance decisions documented and versioned, so the method that produced a figure is the method on file, not the method as remembered. third, versioned data: corrections appended rather than overwritten, so a restatement is an explainable event instead of an archaeology project. fourth, reconciliation to the financial estate: the same reporting entity, the same period, the same consolidation logic as the financial statements the disclosure accompanies — because that consistency is among the first things an assurer tests. fifth, named ownership: someone accountable for each figure, with the sign-off recorded.

there is a sixth property the scope 3 framework makes unavoidable: graded quality. because the standard requires disclosing the extent to which the figure rests on primary and verified inputs, the estate must carry, for each input, a record of where it came from and whether it was verified. data quality stops being a private engineering concern and becomes a published attribute of the disclosure itself.

none of this requires any particular software, and this paper is not a procurement argument. it is an observation about architecture: these properties cannot be retrofitted onto a year-end spreadsheet, because they describe how the data must be kept, not how the report must be written. a company that starts keeping its emissions data this way now will meet the mandate, whenever it lands, with a figure it can defend. a company that waits for the mandate will be rebuilding its estate and answering its assurer in the same year.

§7 · method & limits.

this is a desk-research paper. every factual claim is footnoted to a public primary or authoritative source, fetched and read during june 2026; the full source log, including conflicting claims found and rejected, is retained alongside this paper. it makes no assertion about the readiness of any specific company, and it is not legal, tax or assurance advice.

its main limit is motion. the fsca's roadmap work, the jse's guidance revision and the issb's amendment programme were all in flight as this was written — the december 2025 amendments to s2 are themselves evidence that the measurement rules keep moving. dates and scopes stated here are as at june 2026 and should be re-verified before being relied on. the analysis in §5 and §6 — what the disclosures presuppose of a data estate — depends on the standards' structure rather than their dates, and should outlive the specifics.