Sustainability and CSRD Reporting Software: Why No One Can Trace the Number in Your Annual Report Back to a Utility Bill

Why the collection problem is the whole problem

It is the second week of the reporting cycle. The group controller has a workbook with a tab per site and forty two tabs. Eleven sites have returned nothing. One returned electricity in kilowatt hours for eleven months because the meter was replaced in August and the new account starts a different billing period. A site in Poland reported district heating in gigajoules, a site in Texas reported natural gas in therms, and someone converted one of them with a factor pasted from a search result. The Spanish site's landlord provides electricity as part of the rent and will not break it out, so last year somebody estimated it from floor area and wrote the method in a cell comment that has since been deleted.

The tools in this space are real and some are excellent. Workiva is strong at the disclosure document and its audit trail. Persefoni and Watershed do carbon accounting properly. Sphera covers operational health, safety and environment depth. Novata serves private markets portfolio collection. What none of them solves, because it is not solvable as a product, is your group's actual data reality: which entity owns which meter, which lease is triple net, which joint venture is consolidated for financial purposes but not operational control, and which of your forty two sites has a controller who will answer emails in a reporting week.

The reason this matters more than it used to is assurance. Once a limited assurance opinion sits on a sustainability statement, the number stops being a marketing figure and starts behaving like a financial one. The assurance provider will select samples and ask to see the source document behind a disclosed figure, the calculation applied to it and the approval that let it into the statement. If the honest answer is a workbook and a memory, the engagement gets longer and more expensive every year, and the restatement risk is entirely yours. Note that the scope and timing of European requirements have moved through the simplification process and continue to be adjusted, so confirm which wave and which standards apply to you with your auditor rather than with any article, this one included.

Problem 1: your data providers are people with day jobs

Sustainability data does not come from a system. It comes from a facilities manager who has to find a waste transfer note, a fleet supervisor who exports fuel card data, an HR (Human Resources) partner who pulls headcount by contract type, and a procurement analyst who cannot get a supplier to answer. None of these people report to sustainability, none of them are measured on this, and all of them are being asked in the same fortnight as the financial close.

Platform vendors solve this with survey modules. Surveys work when the question is simple and the respondent is motivated, and they fail exactly where your data is hardest: the site that has three electricity accounts, one of which was closed mid-year, and a landlord invoice that bundles water with service charge.

What a custom build does: treat collection as a workflow with named owners, due dates, escalation to the site's manager rather than a shared mailbox, and guidance written for your actual estate. Where possible, remove the human: pull utility data from supplier portals or invoice feeds, fuel from the card provider, travel from the corporate booking tool, fleet mileage from telematics. Document extraction reads uploaded invoices for consumption quantity, unit, period and account number, then flags a period that does not join to the previous one, which is how you catch the eleven month year before the auditor does. The measure of success is boring: the share of data that arrives without anyone typing it.

Problem 2: emission factors are versioned data, not constants

A factor library is not a lookup table you paste once. Factors change annually by publisher and by country, grid intensities are restated retroactively, and scope 2 has both a location-based and a market-based treatment that require different inputs including contractual instruments. When you restate a prior year, you must be able to say whether the change came from better activity data or from a factor revision, because those two explanations mean completely different things to a reader.

What a custom build does: every calculation stores the input, the factor identity, the factor version, the publisher, the conversion path and the code version that produced the result. Recalculating a prior period with a new factor set is an explicit, logged operation that produces a comparison rather than an overwrite. Unit conversion is centralised and tested, because the therms to kilowatt hours mistake is not hypothetical, it happens in nearly every workbook we have been shown. When your auditor asks why last year moved, you produce a variance breakdown by cause rather than defending a spreadsheet.

Problem 3: assurance wants a path from the invoice to the sentence

The assurance provider picks a disclosed number, then walks backwards: which sites contributed, which readings, which document, who approved it, what changed since the last version. In most organisations that walk stops at a workbook with no version history and an email thread.

What a custom build does: an append-only record where the uploaded source document is stored, hashed and linked to the data point it supports, every edit carries an author and a reason, and approvals are captured as sign-offs at site, function and group level. The disclosure references the aggregation, so clicking a figure in the draft statement opens the contributing records and their evidence. This is the difference between an assurance engagement that samples ten items and one that expands scope because the first three could not be traced.

Problem 4: acquisitions and divestments break every baseline you publish

You publish a base year and a reduction target against it. Then you acquire a business with eleven sites, divest a division, and change the consolidation approach for a joint venture. Every published number is now on a different footing, and recalculating the base year requires knowing precisely which entities were in scope in which period and under which consolidation approach.

What a custom build does: an effective-dated entity and site register that carries ownership percentage, consolidation approach and operational control flag over time, so any period can be reported as published or restated on the current perimeter. Base year recalculation becomes a rule you apply with a threshold rather than a project you dread. Group controllers already have this discipline for financial consolidation. Sustainability numbers lack it because a workbook has no concept of time-varying structure, and generic platforms fill that gap weakly because they cannot know your group's history.

Problem 5: the value chain is estimates today and demands tomorrow

Most of a manufacturer's or retailer's footprint sits outside its own operations, and the first pass is always spend-based estimation because that is the data you have. The pressure then goes two ways: your customers send you questionnaires, and you have to send them to your own suppliers. Both directions run on spreadsheets and neither reconciles.

What a custom build does: hold both the estimated and the supplier-reported figure for the same category, with an explicit hierarchy that prefers primary data when it meets your quality criteria, and never silently double counts. Supplier requests go out with a portal or a structured form that returns data you can actually load, and responses are scored on quality so improvement is visible year over year. When a customer sends you a questionnaire, the answers come from the same store as your statutory disclosure, so you stop publishing two versions of your own footprint.

What this costs and how long it takes

Across the 2,000-plus projects Digital Heroes has delivered, here is the shape for this category. A focused first release covering the site and entity register, data collection workflow with document extraction, the calculation engine with versioned factors, and the source-to-disclosure evidence trail runs $70,000 to $150,000 and ships in 12 to 18 weeks. That is a system your sites use in the next cycle. A full platform adding supplier engagement, target and scenario tracking, multi-framework mapping so one data point serves several disclosure regimes, and disclosure drafting with review workflow runs $180,000 to $450,000 phased over 7 to 14 months.

What drives cost up: the number of countries, because units, utility invoice formats and local reporting obligations all vary. The number of frameworks you report under, since a group filing in Europe, answering a California regime and responding to investor questionnaires needs one data model mapped several ways rather than three collections. Integrations to utility portals, fuel cards, travel management and telematics, each of which is a small project. And the entity structure work, which is where the hidden effort lives, because nobody has ever written down the full list of legal entities, sites, meters and leases in one place. Expect that to be a genuine workstream with your finance team, not a data entry task.

What keeps cost down: starting with scope 1 and 2 across your largest twenty sites, which is usually most of your operational footprint, and leaving value chain categories to the second phase once the collection habit exists.

Build versus buy, and when buying is right

Buy if you are a single entity in one country with a handful of utility accounts. A consultant and a well-built workbook will get you through and a build would be theatre. Buy also if your primary need is the disclosure document itself with strong controls over drafting and sign-off, because Workiva does that job well and rebuilding it is not a good use of money. If carbon accounting methodology is your main gap and your estate is simple, Persefoni or Watershed will get you further faster than a bespoke engine.

Build when two or more of these are true. You collect from more than about twenty five sites or entities and your response rate is the bottleneck. Your group structure changes often enough that restatement is a recurring event. You report under several frameworks and are currently collecting the same data more than once. Your assurance provider has already raised traceability as a finding. Or you have bought a platform and find your team still runs the real process in Excel and uses the platform as a place to put the answer, which is the most common story in this sector by some distance.

Our position is that the calculation engine is the commodity and the collection layer is the differentiator. Anyone can multiply an activity figure by a factor. Getting a facilities manager in Ohio to upload the right invoice on time, with evidence attached, is the part that decides whether your reporting is defensible, and that part is shaped entirely by your organisation.

How to choose a developer for sustainability reporting software

Ask them how they store a calculation. If the answer does not include the factor version, the publisher, the conversion path and the code version alongside the result, they will build something that cannot survive a restatement, and restatement is certain.

Ask how they would model a joint venture that is consolidated financially but not under operational control, across a year in which your ownership percentage changed. If they shrug at effective dating, walk. That single modelling decision determines whether your base year survives your next acquisition.

Ask what they have built that has been audited. Financial systems experience matters more here than sustainability marketing experience, because the requirement is an evidence chain and segregation of approval, which is a familiar problem to anyone who has built for a controller.

Ask who owns the code, the cloud accounts and the underlying data, in writing, before kickoff. At Digital Heroes the client owns everything from the first commit. Your evidence archive has to outlive your software vendor, your consultant and probably your current framework, because the assurance question about a figure you published this year can arrive several years from now.