In short
- Ask what a vendor does not do before you ask what it does. The answer separates the field fast.
- Specific answers name a database, a version or a number. Evasive ones name a roadmap.
- Six things should be shown live on your own data. A follow-up session usually means no.
You are not in the call to watch a demo. You are there to make the vendor do things with your own data and answer in specifics. Ask in a fixed order: what must this produce for us, how is the method held between years, what does year two cost, and what happens if we leave. A good answer names a database, a version, a number or a limit. An evasive one names a roadmap. Send your requirement and one messy file before the call, keep the scripted demo for the end, and make six things happen live rather than in a deck.
This page is about what to say in the room. If you are still working out which criteria should decide it, our evaluation framework for carbon accounting software covers that part.
How should you set the call up?
Four things, done before anyone joins.
Send your requirement in one line. Not "carbon reporting" but the named artefact you owe someone, and the date it is owed. Vendors who cannot produce it should say so at that point, which saves everybody an hour.
Send one real file. Your messiest one. A quarter of energy invoices with two months missing, or a travel export with inconsistent airport codes. Ask them to use it instead of their sample data.
Bring two people. One asks, one writes the answers down verbatim. You will be comparing wording across three vendors a week later, and paraphrase loses the difference.
Ask for pricing and contract answers in writing. Nobody will hold a salesperson to a sentence in a call. Everybody holds them to an email.
What do you ask before anyone opens a screen?
Four questions, and they are deliberately uncomfortable. Ask them first, while the answers are still unrehearsed.
What do you not do? The strongest signal in the whole call. A vendor with a clear answer knows where its edges are. One who cannot name a single edge is either not thinking about it or hoping you will not notice until year two.
Who is this a bad fit for? A different angle on the same thing. It gets you the profile of their unhappy customers, to compare against your own.
Why do customers leave? Everyone loses some. The vendors worth buying from can tell you why.
Which of our exact deliverables have you produced in the last twelve months, and can we see a redacted one? Not a feature claim, an artefact. This is where a checklist answer meets a real document.
What does a good answer sound like next to an evasive one?
The difference is almost always specificity. Good answers contain nouns you could go and look up.
| The question | A good answer sounds like | An evasive answer sounds like |
|---|---|---|
| Where does this electricity factor come from? | Names the database and the version, and shows it against an entry | "We use internationally recognised sources" |
| What happens to last year's figure when factors update? | Says which of the two behaviours it does, and shows a version stamped on a past entry | "Your data is always kept up to date" |
| How does purchased goods and services work? | Explains the split between spend-based and supplier-specific and how you move between them | "We have full scope 3 coverage" |
| What does year two cost for this same shape? | A number, with any first-year discount named as a discount | "We are always flexible at renewal" |
| Who else like us uses this? | Offers a named reference in at least their second reporting cycle | "We have many customers in your sector" |
| What happens to our data if we leave? | Describes the export, the format and when you get it | "You can always download your reports" |
| What is on the roadmap? | Names what is not built yet and does not price it | Answers three earlier questions with roadmap items |
That last row is the pattern to watch for across the whole call. A roadmap answer to a present-tense question is a no with a nicer tone, and it should be written down as a no.
What should you make them show live?
Six things. Do not accept a description of any of them.
Load your worst file. Watch what happens to the missing months and the inconsistent units. How a tool behaves on incomplete data is the most useful thing you will see, because incomplete data is what you have.
Trace one entry back to its source. Pick a figure on screen and ask to see the activity data, the factor and the document behind it, in that order, without leaving the product.
Show two years side by side. Then ask what method each one used, and whether the older one changed since it was published.
Add an entity mid-year. Groups acquire things. Watch whether the consolidation and the comparatives survive it.
Invite an external auditor with read-only access. Ask whether that seat is billable while you are there.
Export everything. Raw activity data, factors applied and results, in a format that is not a PDF.
If any of the six turns into "we can show you in a follow-up session", note which one. Sometimes that is scheduling. More often it is the answer.
Which questions waste the call?
Four, and they are the ones most often on a buyer's list.
How many emission factors do you have? A number with no meaning until you know which databases, which versions and whether yours are in there.
Do you support scope 3? Every vendor says yes. Ask about one category you actually have instead.
Do you have AI? Ask what it changes about a number, who checks it, and what happens when it is wrong.
Is it easy to use? Unanswerable by the person selling it. Ask them to make you complete a task on the shared screen and judge for yourself.
One more worth adding late: is this a carbon platform or a broader ESG suite? Those are different products with overlapping demos, and the distinction is worth settling before you buy. We pull it apart in ESG reporting software versus carbon accounting.
What are our own answers?
Fair to give ours, since we are telling you what to ask. Including where we lose.
What do we not do? Product footprints. The platform does organisational footprints, and LCA, EPD, ECI (MKI in Dutch) and PCF work is delivered as consultancy. If a customer wants a number per article, a licence is the wrong purchase.
Where do we lose? Four places, all from paying customers on a public page. A small business reviewer said on G2 in August 2026 that loading data takes a lot of manual labour, and that once it is in, it works perfectly. A mid-market reviewer said on G2 in June 2026 that they want more integrations with other software. A different mid-market reviewer rated us 3.5 out of 5 in June 2026 and said we are less complete as a broad ESG and CSRD platform, with no data source management feature and no decarbonisation target monitoring. A mid-market reviewer in July 2026 wanted to see the conversion factor and the distance calculator applied behind an entry, which is exactly the traceability test above, and we did not pass it for them.
What does it cost? A free account with no sales call, and Pro from EUR 1,200 a year, published rather than quoted.
How many customers do you have? We will not give you a number. Our own sources disagree with each other and we have not reconciled them, so anything we said would be a figure we could not stand behind. That is the shape of answer you should want from any vendor when the honest answer is that they do not know.
What should you do first?
Write your deliverable on one line, pick your worst data file, and send both to every vendor before the first call. That one email changes what the call is about, from what the product can do in general to what it does with your evidence.
Then run the six live tests yourself before you sit in any call, so you know what a good answer looks like from the inside. You can open a free account with no sales call, and book a meeting if you want someone to talk through your shortlist.
Sources: Hedgehog platform, Hedgehog on G2. Verified 27 August 2026.
Facts on this page were last verified on 2026-09-17.
