[loganbyers]

Hey, this is an interesting and emerging market, what will maybe be called 'carbon analytics' or 'GHG analytics'. There are definitely lots of opportunities in this space.

Without letting perfect be the enemy of good here... I have some questions.

It seems nice to have a backstop for what GHGP calls "Corporate-level data" which is the lowest level and least specific for purchased goods and services in GHGP's Scope 3 calculation guidance.

Do you have any concerns that your users might shortcut the work to produce high quality estimates of their Scope 3 emissions when you have made it so easy to get lower quality estimates?

As a second angle on this, do you have any accuracy or uncertainty estimates for the CO2e values a user receives? Let's say my company goes to a supermarket one month and buys $1000 worth of beef, and the next month buys $1000 worth of lentils from the same store. The GHG impacts of these purchases are (in reality) entirely different, but your API would tell me the emissions are the same. I know this line-item accounting is a massive challenge, but it seems there is an equal risk of green-washing as brown-washing here. This might be acceptable at the aggregate level but potentially harmful and very inaccurate at the individual level. Is there sufficient information returned from the API for your users to communicate data quality in a way aligned with the GHGP reporting standard?

Second line of questioning - GHGP guidance says companies must re-account historic emissions when data becomes available that significantly change estimated emissions. Would your API be compatible with this requirement? The Docs are locked off and the example on the homepage doesn't show any time component (presumably Uber's data would be different for 2022 than 2020).

Some disclosure: I work at the World Resources Institute with many colleagues who co-authored the GHGP guidance, though I have very low association with that project. I am acting on my own here.

[tedpower]

Hey, good questions! You know your stuff.

Re: do you have any concerns that your users might shortcut the work to produce high quality estimates of their Scope 3 emissions when you have made it so easy to get lower quality estimates?

> I think the high order bit here is that 99%+ of companies don't measure their emissions at all. This is for a good reason — measuring your emissions historically has been quite labor-intensive. Even for large companies, there is always a 'long tail' of 'Scope 3 goods and services' transactions that are hard to measure. Our goal is to create a scalable solution so that a much larger share of companies are able to participate.

Re: your grocery store example —

> This is a fair point. Our main belief is that realtime, actionable data trumps perfectly attributed data, if perfectly attributed data requires a bottoms-up manual model. The advantage of the spend-based approach is that (1) it's realtime, and (2) it aligns incentives at the company level. The holy grail, however, would be itemized spend data (level 3 data), where you could factor in the emissions of your specific line-items. Unfortunately, that data is nearly impossible to get (yet). Maybe that's Bend 2.0 :)

Re: re-accounting historical emissions —

> Yes! We use the emissions 'factor' that most closely matches the transaction date. So for example, if you bought a Starbucks coffee in 2020, we would use the 2020 Starbucks factor, and if you bought a Starbucks in 2021, we would use the 2021 Starbucks factor. If Starbucks is late to publish their 2022 report, we would recalculate the emissions when the info is updated. For our category fallbacks, we also take currency / region into consideration.

Happy to chat more, either with you or the WRI folks! Thanks for the questions.

[loganbyers]

Thanks for the responses - I think I typically agree with your approach that there are substantial benefits to getting people to crawl or walk rather than making them run.

On your belief that actionable data _trumps_ perfectly attributed data, I'm not entirely convinced. I think actionable data _complements_ attributable data. But you need the attributable data to accurately measure the impact and learn what specific actions caused that impact.

I think all companies want to make actions that are well-informed, 'the right choice', and have the potential to demonstrate it was 'the right choice'. My concern is that when someone takes a 'good action' such as replacing high intensity animal protein with low intensity plant protein there is no evidence from your side that it made any difference. You are divorcing the actual choice that was made from what is perceived as the outcome.

The dollars-to-emissions relationship is just not as simple as is being represented, and for those who are not specialists there might be a false sense of progress.

"We cut our daily emissions from transport by having employees purchase Uber rides in off-peak times."

"We cut our emissions for business travel by setting up a policy that flights must be purchased at least 2 months in advance."

"We cut our emissions from our regular food purchases by looking at the local newspaper for coupons and signing up for a customer loyalty account."

"Maybe we all should fly to Las Vegas (tickets are cheap) rather than have Linda fly to NYC (an expensive ticket)."

Each of these might be smart business choices, but they have absolutely no real world effect on emissions that should be attributable to a company, but that isn't what the company is being told.

As a first pass to estimate sense of scale and where to look into unsustainable practices and prioritize better data collection, Bend seems to be valuable.

[tedpower]

Thanks, ya, fair points — as orgs level-up and get more sophisticated, you might start to use Bend as the first pass, identifying hot-spots, and then doing deeper dives on the specific 'SKUs' that are being purchased.