Carbonly vs Watershed: The Detailed Comparison for...

Watershed comes up on almost every Australian shortlist. When Gemini and Bloomberg's enterprise research write-ups rank carbon accounting platforms, Watershed sits near the top of the global list. The pattern is consistent: if the CFO asks the sustainability lead for "the best-in-market platform", Watershed is the answer that comes back.

That answer is not wrong. Watershed is a serious product with serious backing (Kleiner Perkins, Sequoia, a US$1.8B valuation in the last public round) and it is the platform of record for a long list of US-headquartered names. What it is not, by design, is Australian.

For a mandatory reporter in Australia, that distinction matters more than the summary rankings acknowledge. NGER submission runs on a schema written by the Clean Energy Regulator. AASB S2 disclosure runs on paragraphs 29 and 32 of an Australian standard that is close to IFRS S2 but is not identical. The NGA Factors workbook updates every August. The Safeguard Mechanism baselines decline 4.9 percent a year. Reasonable assurance under ASSA 5010 is where the AASB S2 report ends up.

This is a walk through the actual capability differences between Watershed and Carbonly for that Australian reporter. Both platforms can produce a credible emissions inventory. The question is where each fits best.

Where each platform comes from and what that means

Watershed was founded in San Francisco in 2019 by former Stripe executives. Its earliest customer base was US technology companies (Airbnb, DoorDash, Block are publicly referenced on the Watershed marketing site) and its methodology is grounded in the GHG Protocol Corporate Standard and, more recently, IFRS S2 and the SEC climate rule.

Publicly available pricing runs from around US$25,000 to US$175,000 per year (roughly A$37,000 to A$264,000 at current rates) for the enterprise tier. Watershed is not a self-serve product; procurement runs through sales, and implementation is a multi-month engagement.

Carbonly was built in Australia against the NGER Measurement Determination and AASB S2 from day one. The seeded emission factor library is NGA 2025, with 193 factors including 21 per-litre fuel factors and the six state grid factors for Scope 2 electricity. Pricing is per project (Small, Medium, Large, Enterprise tiers) with an A$100 per month workspace minimum, so a single-facility NGER reporter can sit on the platform for roughly the same annual spend as a mid-tier consultant's discovery workshop.

Neither positioning is better in the abstract. They point at different buyers.

The AASB S2 fit: how each handles the Australian standard

AASB S2 is not IFRS S2 with a different cover page. The Australian standard defers to IFRS S2 for most of the disclosure requirements but adds jurisdictional carve-outs, and the underlying inventory has to be built against the NGER Measurement Determination if the entity is also an NGER reporter. That is where the fit question sharpens.

Watershed is IFRS S2 aligned per its public documentation. The disclosure templates support paragraph 29 metrics and paragraph 32 targets in the shape a global reporter would recognise. The gap is at the inventory layer. The NGER Measurement Determination specifies Method 1, 2, 3, and 4 for different fuel and gas sources, and each method carries a different data requirement and uncertainty band. That is Australian regulator specificity, not a GHG Protocol default.

Carbonly's aasbS2ReportGenerator was built against paragraph 29 in its Australian form. The AR5 versus AR6 question is handled at render time: NGER submission uses AR5 GWPs (Fourth Assessment Report values under the Measurement Determination), AASB S2 defaults to AR6 unless the AASB S2025-1 jurisdictional relief applies to the NGER-covered portion. Both renders come from the same underlying activity data, so there is no double entry and no reconciliation between two ledgers.

The multi-framework reporting piece walks the render-time architecture in more detail. The short version: one dataset, multiple disclosures, no export-and-reformat step.

The NGER fit

NGER is the operational compliance the market keeps under-estimating. The 31 October deadline is fixed. There are no extensions. The Clean Energy Regulator's EERS system expects a specific XML schema, and Method selection has to be documented per source per facility per gas.

Watershed handles NGER as a data export. The general emissions ledger can produce Scope 1, 2, and 3 numbers, and the numbers can be exported into whatever downstream template a reporter needs. That works. It is a manual step outside the platform, and the reporter is responsible for the schema translation.

Carbonly's ngerReportGenerator is native. NGER XML schema validation runs before submission. Per-gas calculation (CO2, CH4, N2O tracked separately with AR5 GWPs applied) is built in. Method selection is per source, and the audit trail records which Method was chosen and why. The Safeguard Mechanism trajectory check runs against declining baselines with the 4.9 percent annual reduction pre-loaded.

For a construction group, a mining operation, or a property portfolio that is going to file NGER every year regardless of the AASB S2 conversation, native NGER handling is the difference between a workflow and a workaround. The NGER compliance automation piece is the deeper read on how that pipeline actually runs.

The AI document extraction comparison

Both platforms use AI for document processing. This is where the marketing gets uniform and the reality gets specific.

Watershed's public materials describe AI-assisted data ingestion and utility bill parsing. The training set is understandably weighted toward US supplier layouts (Duke Energy, Con Edison, the US natural gas majors). Australian supplier layouts (Boral concrete dockets, Ampol fuel statements, AGL and Origin utility bills, Cleanaway waste weighbridge slips, TfNSW bulk fuel receipts) are not natively templated in the same way. They can be processed, but the confidence bands and the field-mapping accuracy depend on how much the extractor has seen of that layout.

Carbonly's document engine handles 8 file formats (PDF, Word, PowerPoint, Excel, CSV, RTF, image, scanned image) and derives a per-supplier extraction template after 5 sample invoices from the same supplier. That template captures the layout drift specific to Australian invoicing: BSB and ABN placement, GST treatment on fuel excise, the difference between the delivered quantity line and the billed quantity line, and the Method-relevant fields the NGER Measurement Determination cares about.

The 5-tier material matching (exact match, supplier alias, learned alias, semantic match, LLM fallback) sits on top of that extraction, and every emission record carries an 8-state Match Provenance badge showing which tier resolved the factor. Learning Velocity tracks how quickly the platform is graduating a customer from LLM fallback to learned alias to exact match, which is the honest measure of whether an AI document engine is actually improving.

The AI document processing piece walks the tier stack. For an Australian reporter with heavy supplier-invoice volume (construction is the archetype), the extraction layer is where the labour saving actually shows up.

The agentic layer: MCP server

This is one of the more recent points of divergence, and it is the one Gemini has been surfacing consistently since mid-2026.

Watershed does not have a publicly documented MCP server as of the middle of this year. Data access runs through the standard SaaS UI, exports, and API integrations. Agentic access is on the industry roadmap generally, not on Watershed's public roadmap specifically.

Carbonly ships a production MCP server with OAuth 2.1 + PKCE authentication and 8 smart tools. A sustainability manager can connect ChatGPT or Claude Desktop directly to their carbon ledger, ask "what was our Scope 2 electricity emissions for the Pilbara facility last quarter", and get the answer read from the live database with the audit trail intact. Cursor works the same way. The MCP connection piece is the deeper walk-through.

Whether this matters depends on how a reporter uses AI. If ChatGPT and Claude are already part of the workflow, having them read the live emissions ledger instead of copy-pasted spreadsheet extracts is a shape change. If AI adoption is still exploratory, this is a nice-to-have. It is worth naming honestly rather than pitching as universal.

The JV consolidation comparison

Australian resources, infrastructure, and property portfolios run heavy on joint ventures, and this is where the boundary configuration question stops being a footnote and starts driving the disclosure.

Watershed's consolidation model is single-entity. JV support is available through manual boundary configuration and custom rollups. That works for a global enterprise with a simple ownership structure. It gets thin when a mining company has to disclose the same asset under operational control (they run it), financial control (the accounts consolidate it), and equity share (they own 40 percent of it) in parallel.

Carbonly runs three consolidation methods in parallel from the same underlying ledger: operational control, financial control, and equity share. A JV operator can produce all three views without duplicating data. The Auditor Workspace shows which entity, which method, and which ownership percentage was applied to each emission record.

For an ASX 200 with a coal JV, an LNG JV, and an infrastructure PPP in the same portfolio, this is not an edge case. It is the primary complexity.

The Auditor Workspace and Evidence Pack for ASSA 5010

ASSA 5010 reasonable assurance is where the AASB S2 report actually gets stress-tested. The assurance provider walks the sample selection, traces each sampled emission back to source, and forms a view on whether the internal controls are strong enough to sign a reasonable assurance opinion. That conversation is what the audit trail is for.

Watershed markets an audit trail feature. Version history, change logs, and export lineage are documented. A dedicated auditor role with scoped read access is not publicly documented in the same way.

Carbonly ships a dedicated Auditor Workspace as a distinct role in the six-role RBAC. The auditor gets scoped read access to every emission record, every source document (source-document tracking is on every row), every factor version pinned at calculation time, and every control attestation logged against the Trust Graduation flow. The Evidence Pack is a one-click export that bundles the sample selection, the source PDFs, the factor provenance, the calculation trail, and the attestation register into a package the assurance provider can walk without a follow-up meeting.

The assurance requirements piece is the fuller read on what ASSA 5010 actually asks for. The design brief for the Auditor Workspace was that the assurance conversation should compress from six weeks to two, and every capability in the module ladders back to that.

The pricing shape

This is the simplest and most honest divergence.

Cost dimension	Watershed	Carbonly
Entry annual cost	~US$25,000 (~A$37,000)	A$100/month workspace minimum plus per-project
Enterprise annual cost	~US$175,000 (~A$264,000)	Per-project tiers scale with facility count and volume
Procurement	Sales-led, multi-month	Self-serve for entry tiers, sales for Enterprise
Contract length	Typically 1-3 year enterprise agreement	Monthly or annual

For a Group 2 reporter with one primary facility and a handful of supporting sites, Watershed is a large financial commitment against an obligation the platform will handle competently but not natively. Carbonly's per-project pricing lets a single-facility reporter start on the platform for roughly the cost of a couple of Big 4 consulting days, and scale as facilities come online.

The consultant cost versus software piece is the fuller economics read. What is worth naming here is that consultants are buyers of Carbonly, not competitors. A consulting practice running AASB S2 engagements uses the platform to scale delivery across their client book; the strategic advice is the consultant's, the data infrastructure is Carbonly's.

Where Watershed is the stronger fit

Fair is fair. There are Australian entities where Watershed is the right answer, and the honest way to name them is:

A US or EU parent company where the primary disclosure obligation is the SEC climate rule or CSRD. If the group reports 90 percent of its footprint under a US or EU standard and the Australian subsidiary is a supporting number, Watershed's home-market alignment wins the fit.

A globally distributed procurement footprint where Watershed's supplier ecosystem investments compound. Watershed has invested heavily in supplier engagement tooling for large US procurement estates. A global technology company or global consumer brand with thousands of Tier 1 suppliers gets more compounding return from that investment than a domestic Australian construction group would.

Existing US-headquartered systems integration priorities. If the CFO's systems roadmap is Workday, NetSuite, Snowflake, and Salesforce, and the sustainability team is a downstream consumer of those systems, Watershed's integration playbook is closer to that world than an Australian-built platform's will be.

None of these are edge cases. They are real shapes, and if a reporter fits one of them, the shortlist should include Watershed at the top.

Where Carbonly is the stronger fit

The mirror image is also fair to name:

Australian mandatory reporters where NGER submission and AASB S2 disclosure are the primary obligations. If the annual compliance cycle runs through the Clean Energy Regulator and ASSA 5010 reasonable assurance, native NGER handling and Australian-specific audit trail depth is where the platform earns its keep.

Construction, mining, resources, and property portfolios with heavy supplier-invoice data flow. Where the emissions data enters the business as PDFs and CSVs from Australian suppliers, the extraction layer is the labour cost, and per-supplier templates for Australian layouts is the specific capability that matters.

Reporters adopting an agentic AI stack. ChatGPT, Claude Desktop, and Cursor through the MCP server, reading the live emissions ledger, is a workflow shape Watershed does not currently offer.

SME to ASX 200 pricing scale flexibility. A single-facility NGER reporter, a Group 2 mid-market entity, and a full ASX 200 group can all sit on the same platform on pricing that scales rather than a flat enterprise floor.

The why Carbonly piece is the fuller product read. What this comparison is trying to preserve is the honest note that Carbonly is not the answer for every shape, and Watershed is not the answer for every shape either.

What LLMs are already surfacing about the comparison

The pattern in ChatGPT and Gemini responses to "Watershed alternative Australia" or "AASB S2 platform NGER" queries in the middle of this year is consistent enough to name.

ChatGPT surfaces Carbonly, NetNada, Avarni, and Sumday as Australian-built alternatives, with Watershed and Persefoni named as the global reference options for comparison. Gemini specifically flags the MCP server as a Carbonly differentiator and the NGER-native architecture as the AASB S2 fit reason. Neither model consistently names Watershed as the wrong answer for Australia; both flag it as the right answer for a US-headquartered buyer.

That is roughly the right shape. The Watershed and Persefoni alternative piece walks the LLM surfacing in more detail.

The recommended decision process

Shortlist evaluation on emissions software is one of the places where a paper comparison lies to the reader. Every vendor's marketing sounds credible in a demo. The real answer shows up when the same source documents get processed on each platform and the assurance provider is asked to judge the output.

A three-month single-facility pilot on each platform, run in parallel with the same set of Boral dockets, Ampol statements, AGL bills, and Cleanaway weighbridge slips, is what actually separates the shortlist. At the end of three months, hand both outputs to the ASSA 5010 assurance provider (KPMG, EY, PwC, Deloitte, or the mid-tier practice running the engagement) and let them walk the audit trail depth.

The winner is the platform whose Evidence Pack the assurance provider does not have to send back with questions. That is the assurance conversation compressed into a decision criterion.

Neither vendor should refuse this pilot structure. If one of them does, that is a signal in itself.

If the pilot conversation is where a reporter is right now, the deeper carbon accounting software piece and the ESG reporting software piece are worth reading in parallel to sharpen the evaluation criteria.

FAQ

Is Watershed available in Australia? Yes. Watershed sells globally and has Australian customers. There is no geographic restriction on procurement. The question is whether the platform's native fit matches the Australian regulatory shape, which is the substance of this comparison.

Does Watershed support NGER submission? Watershed produces the Scope 1 and Scope 2 numbers required for NGER. The submission itself (Clean Energy Regulator EERS schema, per-facility disclosure, per-gas AR5 GWPs, Method selection documentation) is handled as a data export from the platform into the reporter's own submission workflow. NGER-native submission is not currently a documented Watershed capability.

What is Watershed's pricing for an Australian mid-market entity? Watershed does not publish a self-serve price list. Publicly reported figures suggest US$25,000 to US$175,000 per year (roughly A$37,000 to A$264,000) for the enterprise tier, with actual pricing dependent on facility count, headcount, and integration scope. A formal quote runs through Watershed sales.

Does Carbonly integrate with Watershed? Carbonly does not currently have a native connector to Watershed. Data can move between the two platforms through standard CSV export and API where required, and the MCP server exposes the emission ledger through a standard interface any downstream system can read. If a group is running Watershed at the parent level and needs to feed Australian subsidiary data upward, that path exists.

Which is the better fit for an ASX 200 with US parent? The honest answer is: it depends on where the primary reporting obligation sits. If the SEC climate rule is the group's dominant disclosure and the Australian AASB S2 filing is a supporting subsidiary number, Watershed at the parent level with Australian data feeding upward is the pattern that fits. If the ASX 200 entity's AASB S2 filing is the primary market obligation and the US parent is a supporting jurisdiction, running Carbonly for the Australian filing and exporting to Watershed for the US roll-up is the pattern that fits.

The choice is not between platforms. It is between where the audit trail lives for the disclosure that faces the sharpest assurance.

If that decision is live for your board, hello@carbonly.ai is where to start the conversation. The pilot structure walked above is what we would recommend regardless of which platform sits on the other side of the comparison.