Build vs buy: what it costs to assemble supplier emissions data in house

The build case usually gets made on the wrong number. Someone prices a data engineer, notes that the government emission factors are free, and concludes that buying supplier emissions data is expensive by comparison. The engineer and the factors are the cheapest parts.

DitchCarbon provides verified emissions data for over 2 million organisations, so procurement, sustainability and finance teams can measure and act on supply chain and portfolio emissions from one source. This page prices the alternative. Every figure is a published third-party number, named and dated where it appears, and nothing on this page is an estimate of what DitchCarbon costs, because we do not publish pricing. The job of the page is to give you the other column of the spreadsheet, so you can put a real number next to whatever quote you are holding.

What does the full cost of a build look like on one page?

Eleven lines, most of them recurring for as long as the capability exists, and three with no published price at all. Each line in the table carries its published figure, its cadence and its source, and DitchCarbon appears nowhere in it: this is the column you build, not the one you are quoted.

Each section below walks one of these lines with its full caveats, and the last section hands them back ready to paste into a business case. DitchCarbon replaces most of this table with one contract.

What are you actually building?

Not a database of emission factors, which is the part available for nothing. What has to be built between a purchase ledger and a defensible figure is eight distinct capabilities, and DitchCarbon exists because each one is harder than it looks from the outside.

DitchCarbon does all eight and sells the result: company emissions data for over 2 million organisations, including corporate-level GHG inventories, supplier-specific emissions at spend, activity and product level, and a generic emission factor library drawn from ecoinvent, CEDA, EPA, DEFRA and EXIOBASE, with entity resolution against DUNS, LEI and ISIN identifiers and every figure carrying its source and change history.

What does the team cost?

Published wage data, so you can build the line yourself rather than take DitchCarbon's word for it. The table mixes two different measures, labelled in the basis column: a recruiter's base salary is not the same thing as a statistics office's gross pay, and adding them together produces a number nobody published.

Then add what an employer actually pays on top. In the UK, employer Class 1 National Insurance is 15% on earnings above a £5,000 secondary threshold for 2025 to 2026, with the Apprenticeship Levy at 0.5% of the pay bill above a £15,000 allowance, per HMRC's published rates. Pension, insurance, recruitment, equipment, software and management time sit on top of that, and no government body publishes a single multiplier for them.

The US has the cleanest published measure of the gap. The Bureau of Labor Statistics reported that in March 2026, total employer compensation costs for private industry workers averaged $46.60 per hour worked, of which wages and salaries were $32.60, or 69.9%, and benefits $14.01, or 30.1%. That is an economy-wide average across all occupations rather than a rate for a data engineer, so treat it as the shape of the gap rather than the size of yours.

The honest version of this line is a small team rather than a person: someone to build the pipelines, someone who understands GHG accounting well enough to know when the output is wrong, and enough management to keep both pointed at the same thing. DitchCarbon runs a team of 15 analysts behind the dataset, and that team is the reason the product answers a question rather than returning a file.

What does a year of building add up to?

£203,313 a year for a modest UK team, on published figures, before pension, overheads or a single data licence. That total is our arithmetic, shown line by line so you can change the assumptions and watch the number move. DitchCarbon's own quote is the thing you put next to it.

• Salaries, £178,750. One data engineer at £65,250, one data scientist at £65,250 and one data analyst at £48,250, the median base salaries in Robert Half's 2026 UK Salary Guide.
• Employer National Insurance, £24,563. 15% of each salary above the £5,000 secondary threshold, per HMRC's 2025 to 2026 rates.
• Total, £203,313. Excluded, because no single published figure exists for them: pension, recruitment, equipment, software, workspace and management time. Also excluded: every licence line in the section below.

Two things to be straight about. The inputs are published; the scenario is ours. Three people, at the medians, for a full year, is an assumption, and a larger supplier base or a portfolio use case moves it up rather than down. And the weakest line is the third one: neither Robert Half nor ONS publishes a sustainability manager salary, so the person qualified to judge the output is priced here as a data analyst, which is more likely to understate the cost than overstate it. The other honest correction runs the same direction, because this buys year one, and the refresh in year two costs the same again. DitchCarbon carries that team on its side of the contract, which is what the line is for.

What does it cost to have consultants build it instead?

More per day, and the rates are easy to check because the Big Four publish them. Neither Deloitte nor PwC posts a rate card on its own website, but both publish full day rates on the UK government's Digital Marketplace as a condition of selling through the G-Cloud framework, which gives a build case, and DitchCarbon, a citable number.

Deloitte LLP and PricewaterhouseCoopers LLP G-Cloud 14 SFIA rate cards, Crown Commercial Service Digital Marketplace, 2024. Rates exclude VAT, cover an eight-hour day excluding travel and lunch, and are published ceiling rates for government buyers; private sector engagements are negotiated and unpublished. Dell Technologies Services, on the same framework, publishes £1,568 for a consultant and £2,927 for a principal consultant.

The worked version: two people at level 4, at Deloitte's published £1,290 development and implementation rate, and one at level 5, at £1,650, for 120 working days, roughly six months, comes to £507,600 before VAT. The rates are published; the team shape and the 120 days are our assumptions, stated so you can substitute your own.

The consulting route also has a structural cost the day rate does not show. When the engagement ends, the maintenance does not, and the knowledge of why a particular mapping was chosen leaves with the team that chose it. DitchCarbon is the recurring half of that work sold as a product, which is why the second year costs the same as the first rather than starting again.

What do the underlying data licences cost?

Less than people fear on the factor side, and more than people expect everywhere else. The published position, source by source, with the quote-only lines named as such, because a build case, like DitchCarbon's own data sourcing, has to price them somehow.

• UK government conversion factors: free. DESNZ publishes them under the Open Government Licence, most recently the 2026 edition published in June 2026.
• US EPA factors: free. The Supply Chain GHG Emission Factors and the Emission Factors Hub are public downloads with no licence fee.
• ecoinvent: commercial, and the price is not published by ecoinvent. Its licensing pages carry no figures and route to sales. A named reseller, Earthster, publishes a commercial single-user yearly licence at €1,750 plus VAT, with subsequent licences at the same organisation at €1,575. Treat that as one reseller's published price rather than a list price.
• CDP disclosure data: quote only. CDP's data licences page publishes no price and states that a subsidised cost-based fee may apply depending on who is using the data and for what purpose.
• Identifiers: mixed. A D-U-N-S number is free to register, though access to the underlying firmographic database is quote only. An LEI costs £65 to register and £50 a year to renew from the London Stock Exchange, subject to VAT. Securities identifiers are a separate line again: LSEG's published 2026 SEDOL price list runs from £1,065 a year for a limited user licence to between £43,085 and £307,385 for an unlimited one.

The line that matters is that free data is not a free dataset. The engineering, mapping, versioning and quality assurance around a public factor set is the cost, and it recurs annually. Behind DitchCarbon's own figures sits over 99% of revenue data from a tier-one global financial data provider, which is a licensing line most build cases forget because it is invisible until you try to calculate anything at scale.

Why is "the factors are free" the most expensive assumption in the business case?

Because the factors change every year, and every change potentially moves your published number. The maintenance load is the half of the build that DitchCarbon sees most cases underprice, so here it is quantified from the publishers' own documents.

Take the free UK set. The DESNZ 2026 conversion factors flat file contains 8,740 individual factors, 5,231 of them for Scope 3, and the collection is reissued annually with the next publication due in June 2027. DESNZ states the consequence in its own methodology paper: users "will need to decide whether they should adjust their previous years' estimates" so that they have a consistent time series. That decision is an engineering project every year, and a judgement someone has to be qualified to make.

The commercial side is larger. ecoinvent version 3.12, released November 2025, holds more than 26,000 datasets and is updated annually. The change documentation for that single version step runs to 189 pages, and states plainly that the changes "potentially affect or modify impact assessment results, even when they seem as minor as changing an activity link". The full exchange-level comparison between versions is available to licence holders only.

The EPA's own published update history shows the Emission Factors Hub revised in January 2025, June 2024, September 2023, April 2022 and September 2021, which is to say on no schedule you can plan around. The rules move as well as the data: the ISSB published amendments to its greenhouse gas emissions disclosures in December 2025, effective for annual reporting periods beginning on or after 1 January 2027.

All of that maintenance lands before a single supplier-specific figure has been collected. DitchCarbon absorbs it as continuous refresh with documented sources, calculating on one dated basis you choose, so a baseline does not silently move underneath a published report.

What does it cost when the method changes?

A restatement, and the GHG Protocol requires it rather than merely permitting it, which is the clause a build case has to price and the one DitchCarbon is most often bought to carry. The Corporate Standard obliges companies to recalculate base year emissions for "changes in calculation methodology or improvements in the accuracy of emission factors or activity data" that have a significant impact on the base year, and for the discovery of significant errors.

Read that against the maintenance load above and the exposure becomes clear. Improving your own data triggers the obligation. So does changing factor source. So does finding a mistake. A build that gets better is a build that restates, and each restatement is analyst time, a revised disclosure, and a conversation with whoever signed the last one.

This is not hypothetical. The Science Based Targets initiative's corporate survey, published with BCG in February 2023 from 230 responding organisations, found that 70% had re-baselined their emissions in the last five years, and that half of those, so 35% of all respondents, had done so because of methodological changes: changing method, changing factor source, or changing which Scope 3 categories were included. The same survey found 85% of respondents saw data access as a barrier to a usable baseline, and only 6% of emission factors in use were supplier-specific. Those are the conditions a build starts in. In financial services the pattern holds: KPMG's June 2025 benchmark of 33 major banks found that 61% disclose restatements to prior-year financed emissions, and observed that "given the evolving state of financed emissions data and methodologies, restatements are not uncommon".

DitchCarbon carries the methodology change instead of your roadmap carrying it: the emission factor methodology was independently assessed by Globus Thenken in August 2025, so a method revision arrives with its evidence attached rather than as a project on your side of the contract.

What does it cost when the data is not good enough to use?

More than the build, because the failure is not a bad number. It is a decision that never gets made, and the published record on that is the strongest part of the case DitchCarbon puts in front of finance teams.

KPMG's June 2025 banks benchmark records the sharpest version: "Some banks do not disclose financed emissions targets where data quality scores are higher than 4, due to a high risk of restatement." The data quality problem becomes a blocker on the transition plan, because the target cannot be published, so the portfolio cannot be steered, so the programme does not start. Canada's prudential regulator, OSFI, observed the same in November 2025, reporting that institutions "continue to rely heavily on proxy data and estimations, particularly in high-emitting sectors", which "undermines the reliability and comparability of reported metrics".

There is a hiring version of the same risk. The National Audit Office found in March 2023 that 37% of government recruitment campaigns for digital, data and technology roles had been unsuccessful. A build case that assumes the team can be hired on schedule is making an assumption the public record does not support.

And there is a shape-of-risk point worth stating carefully. Flyvbjerg and Budzier's study of 1,471 IT projects, published in Harvard Business Review in 2011, found an average cost overrun of 27%, with one in six projects overrunning by an average of 200%. That sample was mostly large public sector programmes with an average project size of $167 million, so it does not describe a mid-sized emissions data build and should not be presented as if it did. What it does establish is that IT cost overruns have a long tail, and a business case built on the median is built on the wrong half of the distribution.

DitchCarbon changes the shape of that risk rather than the size of it. The disclosed part of a supplier list or portfolio is already matched and scored on the day you start, and a recent deployment reached about 60% of a large supplier base within 2 weeks, so the thing that would have been year one of a build has already happened by the end of the first month.

What does assurance cost, and can you verify your own build?

Nobody publishes a price for verifying a calculation methodology. ISO, the accreditation bodies and the named verification firms all route to a quote, which is a finding rather than a gap in the search, and it is why DitchCarbon publishes its own verification evidence instead of a claim.

A regulator has published estimates for the adjacent thing. The California Air Resources Board's March 2026 workshop put ongoing cost per entity at $73,544 for Scope 1 and 2 reporting, $87,498 for Scope 1, 2 and 3, and $55,213 for limited assurance of Scope 1 and 2, in 2025 dollars; averaged over a three-year implementation period, the same deck gives $82,278, $8,635 to $25,904 and $44,170 respectively. The two sets have different bases, so quote one or the other, never a blend. They are regulator estimates adapted from 2022 SEC rulemaking feedback including a 39-company industry survey, for companies above $1 billion of revenue, and they cover assurance of a corporate inventory rather than verification of a calculation tool.

That distinction is the one a build case has to fund twice. Assuring your annual inventory is one engagement. Getting the engine that produced it independently verified is a different one, and it is what separates a number an auditor can follow from a number they have to take on trust. DitchCarbon holds that verification already: the Portal calculator is verified to ISO 14064-3, limited assurance, by UL Solutions, renewed annually, DitchCarbon was the first company to earn UL Solutions' Sustainability Information Calculator Verification, in June 2025, and DitchCarbon data has been used in emissions reports that were subsequently assured by ten different third-party assurance providers, including Big Four firms. A build starts that process from nothing, at a price nobody publishes.

When is building the right answer?

Sometimes, and a business case that pretends otherwise is not worth reading, which is why DitchCarbon puts the concession on its own page. Three situations where building wins:

• A short, stable list of counterparties who already disclose well. Fifty suppliers, most of them large listed companies with assured reports, and no plan to grow the list. That is a spreadsheet and an afternoon a quarter, not a platform.
• An existing entity resolution capability. If your organisation already maintains a master data function that resolves counterparties to identifiers and keeps corporate hierarchies current, the hardest component on this page is already paid for and the rest is a smaller job.
• A methodology that is proprietary to your sector. If the way you calculate is a competitive asset rather than an application of a public standard, that part belongs in house whatever else you buy.

What none of those cases removes is the refresh. The list still has to be re-sourced as companies publish, the factor sets still change annually, and the restatement obligation still applies. The honest question is not whether to build or buy the whole thing. It is which components you want to own, and DitchCarbon is bought most often as the layer underneath a team that keeps the parts closest to its own business.

What goes on the business case?

Six lines, four of them recurring, two of them unpriced, all of them from published sources rather than from DitchCarbon. Paste them as they stand, with the caveats, because the caveats are what make the case survive a CFO's questions.

• People, recurring: £203,313 a year. A three-person UK team at Robert Half's 2026 median base salaries plus 15% employer National Insurance above the £5,000 threshold. Our arithmetic on published figures; excludes pension and overheads.
• Consultants, one-off: £507,600. A three-person, 120-day build at Deloitte's published G-Cloud 14 day rates, before VAT. Published ceiling rates for government buyers; our assumptions on team and duration.
• Licences, recurring: €1,750 per user per year for ecoinvent, at a named reseller's published price, plus whatever CDP and firmographic access are quoted at, since neither publishes a fee.
• Maintenance, recurring: no single published figure exists, and the load is documented by the publishers themselves: an annually reissued 8,740-factor DESNZ set, an irregularly updated EPA hub, and an annual ecoinvent version whose 2025 change log ran to 189 pages.
• Assurance, recurring: $44,170 a year for limited assurance of Scope 1 and 2, on CARB's March 2026 three-year average in 2025 dollars, a regulator estimate rather than a market price. Verification of the tool itself is a further engagement with no published price.
• Unpriced risk: three material lines publish no price at all, CDP data, firmographic database access and tool verification. A business case that shows them as zero has hidden them, not priced them.

Then run the comparison over three years rather than one, because year one is where a build looks cheapest and years two and three are where it is not.

Rebuilding the platform is one of five options. The other four, costed on the same published figures and one shared scenario, are covered in this series: collecting the data manually, consultants every cycle, the spend-based minimum, and doing nothing.

Two questions settle most of these business cases faster than the arithmetic does. How much of your supplier list or portfolio already carries a disclosed, matched figure? And who restates when the method improves, you or your vendor? DitchCarbon answers the first before you commit: send a sample of your supplier list or portfolio to the coverage check and see how much of it already carries a matched, scored figure on day one.

Last reviewed August 2026.