Every greenhouse gas inventory contains estimates. Every single one, including the assured ones, including the ones from companies with sustainability departments larger than your company.
- Every inventory contains estimates. The failure is not estimating, it is not saying which figures were estimated.
- Declare data quality on every record: primary, secondary, estimated or proxy.
- Weight the aggregate by emissions, not by record count. The share of tonnes, not the share of rows.
- An estimate needs a stated method, a rationale and an owner, not an apology.
The problem was never estimating. It is that a metered reading and a rough guess arrive in the same column, formatted identically, and from that point on nothing can tell them apart.
Where the Estimates Actually Are
A typical first inventory for a mid-sized manufacturer:
- Electricity: metered, invoiced monthly. Genuinely primary data.
- Diesel: delivery notes. Primary, though what was delivered and what was burned in the period are not the same thing.
- Refrigerant: a contractor's service sheet. Primary if you have it, absent if you do not.
- Waste: a weighbridge ticket gives tonnage. The split between landfill and recovery is the contractor's standard ratio, not a measurement of your waste. That is an estimate wearing a measurement's clothes.
- Business travel: a travel agency statement. Distances are great-circle between airports, with no allowance for routing or holding. Reasonable, and not what was flown.
- Employee commuting: a survey at best, headcount times an assumed distance at worst.
Six categories, and only the first two are unambiguously measured. This is normal. It is what a real inventory looks like.
Why the Label Does the Work
The GHG Protocol data quality tiers are unglamorous, and they are the most useful thing in a first inventory:
Declared per record. Aggregated into the report. It costs one dropdown at entry and it changes what the inventory can be used for.
Without the label, a reader cannot tell which they are holding, and neither can you, twelve months later, when someone asks why the figure moved.
The Question That Separates Useful From Decorative
The tiers only matter if the aggregate is weighted by emissions rather than by record count.
"Sixty percent of our records are primary data" tells you almost nothing. If the forty percent that are estimated happen to include the plant's electricity, the inventory is mostly guesswork with a good-looking ratio.
The number worth reporting is the share of tonnes, not the share of rows. Our demonstration inventory is 94.1% primary and 4.5% estimated by emissions, and it is the emissions weighting that makes that sentence mean anything.
This also tells you what to fix. Sort your estimated lines by tonnage and work down. Most teams instinctively attack the line that is most estimated; the useful move is the estimated line that is largest.
An Estimate Needs a Method, Not an Apology
The failure is not "we estimated this". It is "we estimated this and did not say how".
Compare the same figure reported two ways:
The number is identical. The second one is defensible, tells a reader exactly how much weight to put on it, and demonstrates that somebody understood the limitation rather than hoping nobody would ask.
This is what an assumption register is for, and why we require the method and the rationale rather than just an impact rating. An assumption without a method is a note to self.
Estimating Is Not a Failure State
There is a tendency, especially in a first year, to treat every estimate as a deficiency to be apologised for. It leads to two bad outcomes.
Teams delay reporting until they can measure everything, which means they report nothing, for years, while the data they could have been collecting expires.
Or they quietly upgrade estimates to look better. A guessed figure entered as primary data is not a small presentational choice. It is the one action in this entire process that makes an inventory actively misleading, because it removes the reader's ability to discount it appropriately.
An estimate is perfectly acceptable. It just has to be labelled as one. That sentence sits in our supplier portal, where an external company is being asked to declare how they arrived at a figure, and it is the sentence that gets the most responses, because it gives people permission to answer honestly instead of not answering at all.
What Good Looks Like in Year One
Not a low estimate percentage. A known one.
An inventory that is 40% estimated, with each estimate carrying a method, a rationale and an owner, is in better shape than one claiming 95% primary data that nobody can substantiate. The first can be improved deliberately, one line at a time, with the largest tonnage first. The second cannot be improved at all, because nobody knows where the soft parts are.
Your second inventory will be better than your first. That improvement only happens if the first one was honest about where it was weak.