Carbon accounting companies: who does what in the 2026 landscape

Carbon accounting companies fall into four recognizable groups: enterprise software platforms, consultant-led hybrids, professional-services practices, and a newer wave of ledger-first tools. This post maps who is in each group and, more usefully, who each group is actually built for. Two disclosures up front: we are one of the companies on this map (Carbonaccounting.ai, early access, ledger-first group), and no dollar pricing appears below, because none of the vendors named publishes a price list and we will not fabricate one. Any comparison article that prints their prices is guessing; treat it accordingly.

Group 1: the enterprise platforms

Persefoni positions as climate management and accounting for the enterprise, with strong financial-audit framing and investor-grade language; it reference-checks well with large, regulated companies. Watershed pairs measurement with decarbonization programs and is arguably the strongest brand in the segment. Sweep is EU-centered, CSRD-first, strong on group structures and supply-chain data collection. Normative brings science-led, spend-based methodology depth from the EU side.

What unites them: quote-based pricing, a sales-led motion (demo after discovery call), implementation projects, and a product genuinely designed for companies with sustainability teams and data resources. If you are a multinational facing reasonable assurance with budget to match, this group exists for you. The structural criticism is the same for all four: you cannot see the product on your own data before the procurement cycle, and for a mid-market company in its first reporting year the implementation weight is the product's main feature and main bug at once. Our fuller comparison, including when these platforms are simply the right answer, is on the Persefoni alternatives page.

Group 2: the consultant-led hybrids

Greenly is the clearest example: software plus bundled advisory hours, aimed at SMBs and the lower mid-market, with humans doing a meaningful share of the classification and interpretation. Several regional players run the same shape. The honest appeal: a small company gets a finished inventory without building any capability. The honest limit: the human hours are the product, so quality tracks the assigned analyst, the marginal cost never really falls, and the methodology can be thinner than the interface suggests, particularly on Scope 3, where sector factors quietly do most of the work.

Group 3: the professional-services practices

Every Big-4 firm and a long tail of boutique climate consultancies build inventories as engagements. This group is covered in depth in services vs software, so the summary here is short: unmatched for judgment calls, regulatory interpretation and assurance preparation; structurally unable to leave a living system behind. The deliverable is a document and a spreadsheet model, priced to be rebuilt annually. Increasingly this group licenses software from the other groups and sells judgment on top, which is the sane equilibrium.

Group 4: ledger-first software

The newest group treats the problem as accounting infrastructure first: ingest the AP export and meter data, classify every line under the GHG Protocol with stated methods and confidence, keep the evidence link from every figure to its source, version the factors and methods, and generate disclosures as views over the ledger. Pricing published, product accessible before sales contact, AI doing the classification pass that used to be the consulting invoice's biggest line.

Carbonaccounting.ai is in this group, and here is our status stated plainly, because this is our own blog and inflating it would be silly: the classification engine is live today as an ungated demo (sample companies or your own pasted lines); the platform around it, imports, review workflow, report packs, supplier requests, is planned and in development; planned pricing is public on the pricing page; nothing is charged during early access; and we have no customers to name yet, which is why no customer logos appear anywhere on this site.

How to shortlist across the groups

• Reasonable assurance this year, multinational, dedicated team: shortlist group 1, and take references from companies your size in your sector.
• Under 200 people, no internal owner, first inventory ever: group 2's bundled hours or a boutique from group 3 will get you a respectable v1; insist on keeping the model and data.
• Mid-market, first or second mandatory year, someone in finance or ESG who can own a monthly review: group 4's economics and auditability fit best, with group 3 bought by the day for boundary and materiality judgment.
• A consultancy running many client inventories: license group 4 tooling, automate the classification, sell judgment. Your margin problem is the data wrangling, not the advice.

Five questions that expose any vendor quickly

• Can I see it work on my data before a sales conversation?
• Is pricing published anywhere?
• When a factor set updates, is last year's number still reproducible, and who approves the change?
• Which categories of my Scope 3 does your methodology actually cover beyond a spend factor?
• What exactly is shipped today versus on the roadmap? (Every vendor has a roadmap; the tell is whether they label it.)

How this market got here

The category is younger than it looks, and its history explains the shape of the four groups. The GHG Protocol Corporate Standard, published in 2001, gave companies a common language for Scope 1, 2 and 3 emissions, but for its first decade carbon accounting was almost entirely voluntary. The companies doing it were large, reputationally exposed, and happy to pay a consultancy once a year for a report nobody audited. That world produced group 3: inventories as bespoke engagements, priced like legal opinions, delivered as documents.

The Scope 3 Standard in 2011 changed the technical center of gravity. Once value-chain emissions were formally in scope, and for most companies they represent 70 to 90 percent of the footprint, the work stopped being a facilities exercise and became a data exercise: thousands of purchase lines, supplier records and logistics movements to classify and match to factors. Consultancies absorbed this by adding junior hours; the first software platforms absorbed it by building spend-based estimation engines, and group 1 was born, funded on the expectation that disclosure would eventually become mandatory.

That expectation arrived in a cluster. The EU's CSRD entered into force in 2023 and pulls tens of thousands of companies, including non-EU groups with significant EU activity, into assured sustainability reporting (our CSRD reporting page covers the mechanics). California's SB 253 put large companies doing business in the state on a path to mandatory Scope 1, 2 and, later, Scope 3 disclosure. CBAM attached emissions data to imports. The common thread: a bespoke annual document does not survive contact with recurring, assured, regulator-facing filings. Mandatory plus repeatable plus auditable is precisely the combination that turns a consulting service into a software category, which is why group 4 exists now and did not in 2015.

How the groups handle Scope 3 differently

Scope 3 is where the groups genuinely diverge, and since it is usually the bulk of the number, it is the right place to test a vendor. The GHG Protocol allows several calculation methods per category, from spend-based estimates at the bottom of the accuracy ladder to supplier-specific data at the top, and every company on this map sits somewhere on that ladder by design.

• Group 1 invests heavily in the top of the ladder: supplier engagement portals, product carbon footprint exchange, and hybrid methods that blend supplier data with modelled estimates. This is the right ambition for a multinational with procurement leverage, and it is also why implementations run long; collecting primary data from a supply chain is an organizational project, not a software feature.
• Group 2 mostly lives at the bottom of the ladder: spend mapped to sector-average factors, with advisory hours interpreting the output. Defensible for a first voluntary inventory, thin for anything an assurance provider will sample, and the thinness is rarely disclosed with the enthusiasm the dashboard is.
• Group 3 is method-agnostic and judgment-rich: a good consultant will screen all fifteen Scope 3 categories, argue about which are material, and document why the immaterial ones were excluded. The weakness is repeatability; the screening logic lives in a memo and a spreadsheet, not a system.
• Group 4 starts from the transaction ledger: classify every AP line to a scope, category and method, state the confidence, and make the upgrade path from spend-based to activity-based to supplier-specific an explicit, per-category decision recorded in the ledger rather than an implicit property of whoever built the model. Our own approach to this is described on the Scope 3 reporting page.

The practical question for a buyer is not "who has the best Scope 3 methodology" in the abstract; it is which rungs of the ladder you need this year, per category, and whether the vendor can show you where each of your categories currently sits and what moving up a rung would require.

What changes with AI classification, and what does not

Every group now says "AI" somewhere on its website, so it is worth being precise about what the technology actually changes in this market. The mechanical middle of carbon accounting, reading a supplier name and a line description and deciding "this is purchased goods, category 1, spend-based method, this factor," is text classification. Language models are good at it, and it is the single largest block of hours in a traditional engagement. That block moving from billed human time to software is the real economic event, and it is the reason group 4 can publish prices while group 3 quotes.

What AI does well here is drafting, not deciding. The distinction matters for anyone signing a disclosure. A classification pass over ten thousand lines will be confidently right about most of them, and the useful system behavior is to say so line by line: high-confidence classifications flow through, low-confidence ones are flagged for a human, and the flag itself is recorded. A vendor whose AI produces only totals, with no confidence signal and no review queue, has automated the work without automating the accountability, which is the part an auditor cares about. This is the design we demonstrate in our live classification demo; the review workflow around it is part of the planned platform, and we label it as such.

What AI does not change: boundary decisions, materiality judgment, regulatory interpretation, and the organizational work of getting data out of procurement systems. Those stay human, which is why the sensible long-run division of labor keeps group 3 in the picture, buying its judgment by the day on top of a ledger a machine keeps current.

Red flags when evaluating carbon accounting companies

Whatever group you shortlist from, the same warning signs apply, and most of them are visible before the first sales call.

• No methodology page. If a vendor will not say in public which standards, factor databases and calculation methods it uses, the answer is either "whatever the analyst chose" or "spend times sector factor for everything." A primer like our what is carbon accounting page should be table stakes for the vendor too.
• A single blended number. Any output that does not break down by scope and Scope 3 category is a marketing artefact, not an inventory. You cannot review, assure or reduce a blended total.
• "Audit-ready" without lineage. Audit readiness is a property of the evidence trail, every figure traceable to source documents, factors and method versions, not a property of the PDF template. Ask to see a figure traced backwards in the product.
• Guaranteed compliance outcomes. No vendor controls what your assurance provider accepts or how a regulator reads your entity structure. Confidence about the process is fine; guarantees about the outcome are a tell.
• Competitor prices in comparison content. None of the enterprise vendors publish pricing, so any article quoting their prices, including articles comparing them to us, is fabricating data. Judge what that implies about the rest of the article.
• Roadmap sold as product. Every vendor in a young category has unshipped features, including us. The red flag is not the roadmap; it is the absence of a label distinguishing it from the product.

We built this site to pass our own test: the demo runs before any signup, the pricing is on the page, the planned parts are labeled planned, and the market map above names our competitors' genuine strengths. If that is the kind of vendor behavior you want more of, the carbon accounting software homepage is where our part of it lives.