Carbon footprint for a software company: travel, commuting and cloud

The laptop on the desk emitted most of what it will ever emit before the box was opened. So did the monitor beside it and the phone in the drawer.

That single fact is the best way into a software company's carbon footprint, because almost everything about it works the same way: it happened elsewhere, earlier, on someone else's site. An office lease, an electricity bill and a year of hardware invoices are the whole of scopes 1 and 2, and for a services business they are a rounding error. What matters is scope 3, which is not an accounting abstraction here but simply the business: people moving, people commuting, services purchased, hardware replaced, and workloads running in a data center you will never visit.

Travel is the best data you own

Business travel is usually among the largest lines and is also the easiest to document properly, because it already flows through a booking tool or expense claims. Pull the year's flights with origin and destination, rail journeys, hotel nights and reimbursed car kilometers.

Flights dominate the category and long-haul dominates the flights: a handful of intercontinental round trips can outweigh the entire rail total for the year. This is also the line where policy connects to the number without waiting for anything. A travel policy that moves short-haul to rail, and asks one question before intercontinental trips, changes the footprint in the quarter it is signed.

One question about commuting

Nothing in your systems knows how employees get to work, so this is the one place a questionnaire earns its keep: usual mode, one-way distance, days on site per week. Keep it under five questions and extrapolate non-responses transparently.

That third question quietly does double duty. A remote day removes a commute and adds home-office energy, heating in winter above all, so remote work shrinks the line without zeroing it. Count both sides with standard factors rather than declaring victory on the commute alone.

Cloud: real emissions, unfinished measurement

Your product runs on servers you will never see, and their emissions are genuinely yours in the scope 3 sense. The difficulty is that measurement here is younger than the rest of carbon accounting. The major providers all publish customer carbon figures, but their tools differ on method and boundary: market-based or location-based electricity accounting, and whether the embodied emissions of the servers themselves are included. Two providers can report figures for identical workloads that differ by a large factor, and both are defensible under their own method.

So take your providers' reported figures, record which method each one uses, present the line as a range rather than a point, and do not compare providers on these numbers. What you control is method-independent and real: region choice, since grid carbon intensity varies enormously between countries, rightsizing, and deleting the workloads nobody remembers deploying.

Good to know: for the hardware you own, the lever is lifespan rather than purchasing policy. A laptop kept four years instead of three cuts that line by a quarter, and no procurement negotiation will do as much.

The invisible stack

The remaining lines come off the accounting export with monetary factors: the SaaS subscriptions the company runs on, agencies and contractors, recruitment, marketing, insurance, the office fit-out. Individually small, together they comfortably beat the office energy. This is the long tail where monetary estimation is the honest method, per the physical-versus-monetary rule: physical data where quantities exist, spend-based factors for the rest.

What a good software footprint looks like

It fits on a page. Travel, commuting plus home office, cloud as a range with the method named, hardware on a lifespan basis, purchased services from accounting, and the office at the bottom where it belongs. Each line names its data source and its factor database.

So rank the scope 3 lines and take the top two to the people who own them, which usually means a travel policy and a hardware lifespan decision. Leave the office lighting for the year the bigger lines are already moving; it is not unimportant, it is just small, and saying so out loud is what keeps the effort where the tonnes are.

FAQ

What are the main emission sources for a software company?

Almost everything sits in scope 3: business travel (flights above all), employee commuting and home-office energy, cloud infrastructure, purchased hardware whose emissions are mostly from manufacturing, and purchased services. Office energy in scopes 1 and 2 is usually a minor line.

How do you count cloud emissions in a carbon footprint?

Use the carbon reporting your cloud providers publish, but record which method each uses, since market-based versus location-based electricity accounting and the inclusion of embodied server emissions differ between providers. Present the line as a range, and act on what is method-independent: region choice, rightsizing, and removing idle workloads.

Does remote work reduce a company's carbon footprint?

Partly. A remote day removes a commute but adds home energy consumption, especially heating in winter. The honest approach counts both effects, using a questionnaire that asks for days on site per week, transport mode, and distance. The net effect is usually a reduction, but smaller than the commute alone suggests.