EP 47: The Truth About ESG Proxy Data: Benchmarks, Risk & Real-World Use

Ella Williamson: Welcome to Sustainable Intelligence, where we discuss all things sustainability for the private markets. Today I’m joined by Caitlin Pentifallo, Head of Sustainability and Regulatory Strategy at Novata. Caitlin sits at the intersection of data, regulation and real world implementation.

She also stewards our own internal sustainability strategy. Caitlin works closely with sustainability leaders who are under pressure to deliver credible reporting, forward-looking strategy, and measurable progress, often without perfect data.

And that’s where today’s topic comes in: proxy data. For many sustainability officers, proxy data has shifted from being a stop gap to becoming part of the operating model. So let’s start at the foundations. Caitlin, from a sustainability perspective, what is proxy data and what role should it play when companies are trying to understand their real world impact?

Caitlin Pentifallo: A good question to start with. So when we’re talking about proxy data, we’re talking about something that has been modeled, estimated, benchmark derived, and that’s being used in place of real direct, primary source data. And that could be on a number of topics. We see a lot in the environmental area related to greenhouse gas emissions but also lots of proxies being used in the social side as well.

Ella: And why do you think proxy data has become so important for sustainability teams at this very moment in time? 

Caitlin: I think we’re seeing what a lot of our clients feel, which is collecting data is hard. It’s hard to have good coverage into your portfolio. It’s really hard to have direct connections if you’re in a situation, for example, with private credit or funds of funds or LP reporting. There’s lots of times where you might not have such a direct connection to those you’re looking to receive data from, and that can lead to coverage gaps. I think something else that’s also happening that we see with the rise of omnibus and the changes related to SFDR 2.0, which don’t panic, we’re not expecting those until summer ’27, but for the first time, we’re really seeing regulators open to the idea that proxy data does have really strong value. That it is allowed, that it is supported, that it does align with PCAF. And so with all of these things, we can now better support our clients to enable something that could be like an early-stage risk screening. Say we have a client in the early stages of a deal and they would love to know what they should expect relative to emissions, relative to other factors, relative to SFDR reporting. And so this proxy data really allows people to do that in a way where previously we might have said they were shooting in the dark. So this gives us more portfolio level comparability. It gives us some directional signaling before we have portfolio companies that are mature enough for data collection. And so it allows us to hold the line until we can graduate into our more preferred forms, which are of course, primary source. But it does serve a really important role in the interim.

Ella: I couldn’t agree more there. And that piece about using it as a stop gap for now, but with the history and backfill data that we have, it is so much higher quality than it ever has been. Now, in your experience, where are the biggest data gaps that you see in practice when you work with clients?

Caitlin: Yeah. Certainly it’s gonna be Scope 1, 2, and 3 emissions. And so I can also tell you those are our most commonly requested metrics. We see those in our top 10, across geographies, across industries. It’s the data that everybody wants. It also means it’s really hard to get your hands on. And so when there are gaps, there aren’t, I would say, great reliable external sources where people can say, oh my gosh, amazing, here it is, private market benchmark data that I can now use as a proxy. What we’re seeing instead is that we might have really large web scraped sets of data that we might not have a lot of transparency into where they came from, how they were derived, what methodologies were used. And so for the first time in Novata’s history, we’re at a point now about four years into this journey where we’ve hit that critical mass of around 1.2 million data points coming from nearly 14,000 users means that we have that information coming back. And so knowing and acknowledging that collecting primary source carbon data is really hard, it also means, though, that we have a lot of data points coming in that we can now fill a really important gap for. So I see that as a key step on this journey, not just for us as a company but for our users as well.

Ella: And based on the discussions that we’ve had in the past, I know that benchmarks power most proxies. What makes a benchmark really credible?

Caitlin: So there’s a number of things that we’re always gonna look at from a data quality standpoint. First thing we always wanna look at is sample size. So we have, depending on the type of benchmark it is, we have minimum thresholds for how many people must be in a particular cohort in order for that benchmark to proceed. For example, say we have a GP with a really large portfolio and they have a lot of data requesters, we will never produce a benchmark that’s exclusive to one portfolio. That’s not a benchmark by our definitions and by almost anyone’s industry standard. We’re really looking for that diversity to ensure that what we’re reporting on is not just reflective of any one corner. We really want to see other data points such as response rate, answer rate. So for answer rate, and you can find this in all of the metadata that comes with any of our benchmark or proxy products, that’s telling us what is the percentage of companies that were asked this question that did respond. So this means that we’ll see really high answer rates on questions that I don’t wanna term as easy, certainly the work behind them is not easy. But it is quite reasonable for a company to respond to whether or not they have a code of conduct policy or an AI cybersecurity policy. We’re gonna see answer rates in the eighties and nineties.

For others where we know people are working really hard to get their data together and to really report, we might see that as a little bit of a drop off. So on, renewable energy used on Scope 1, 2, and 3, you’d be surprised, these are things that really are difficult to capture. So anything in that quantitative direction, we do sometimes see answer rate fluctuation. But I think one of the most remarkable things about our dataset is how consistently high those answer rates are which in turn allows us to derive really high and really strong sample sizes.

Ella: And like you just discussed, it’s balancing what is actually possible based on how hard it is to collect that data, but then also being able to build credibility and build a forward-looking view of your business so you can start to plan towards those. So if you were a sustainability leader trying to decide when to use and integrate proxy data, how would you advise these people to do this?

Caitlin: Yeah. I think the first caveat is that what we’re doing with proxy data, it’s not a substitute for primary data. We’re not saying that it is. But really what we can think of this as is a scaffolding mechanism for decision making under uncertainty. There are lots of times where we’re going to see uncertainty and proxies really are giving us a sense of a direction of travel and that itself can be really helpful. In the same way where we have seen clients using our Benchmark Explorer to great success, it’s good to know what you’re expecting. So are you expecting what these normalized emissions are going to be by millions of revenue or by FTE? What can I expect before I make this investment? What should I expect to see back when I am able to eventually get primary data?

There are a lot of use cases because I think, for many, it is effectively coloring in the gaps. So where we know there are holes, where we know people are missing some things, proxy data can still be forward looking. It’s still quite reliable. And it gives us a really strong sense of direction of travel and helps investors align expectations.

Ella: Really sound advice there. And building on that, do you have any rules of thumb based on materiality and specific use cases in terms of board reporting use cases versus internal learning versus investment decisions, for example.

Caitlin: Number one, the best thing you can do in any of these cases is be as transparent as possible. So everything that I produce and, partially from coming from academia, is going to be super conservative. I’m never going to say with a hundred percent confidence they’re gonna be under this range or this is what we could expect for them. We’re always going to couch that in confidence intervals. We’re always going to park our assumptions under a more conservative lens in order to keep us from using proxy data in ways that are not applicable. I think in terms of what are the best times to do it. As soon as you have something that’s really material, you’ve gotta ask yourself, if I can’t get this primary data, what should I be doing? And there I think it becomes less of a nice to have, more of an operational necessity. So depending on how material something is, how important that is to what it is that you’re trying to communicate and trying to achieve, that can be a great time to leverage proxy data.

Ella: Great, and in terms of materiality and use case, are there any specific governance measures that should be put in place around proxy data when companies do use it?

Caitlin: I think with anything what we will learn and what our users will soon learn is that there’s a tolerance for proxy in all of these cases. There might not be any tolerance at all when we’re talking about third-party verification and audited reporting. Not acceptable. We might say, in target-setting, we have a greater tolerance because we know that there are not tons of available data points that we can look to, especially in that forward-looking set. We could say as it relates to risk and pre-investment diligence, we have a really high tolerance because we’re really looking for that measuring stick as opposed to actually making any kind of claims around it. So I think with all of these things, it depends what that tolerance and what that appetite is for proxy data. However, if you’ve got something that is not really material to you and there isn’t great data around it, that might not justify bringing in proxies. However, if something is really material to what it is you’re trying to do, but there’s not a lot of data available, then that’s a great case for proxies.

So I think lots of internal teams will be going through this process to say, okay, when is this appropriate? When does this make sense for me? And when does it align with what it is that I’m trying to communicate? And so every investment team, every diligence team is all going to work through this at their own pace and time and have their own rationales.

And I think the best thing that we can do sitting in this space as a provider of these proxies is to be as transparent and guided upfront as possible so that our users can then make their own decisions about when it is and is not appropriate.

Ella: I really like the way you framed tolerance and appetite and really understanding where that sits within your business and within a certain area. Is there anything when you do start to bring in proxy data that you should keep an eye on in terms of when things could go wrong?

Caitlin: I do worry, I worry about a lot of things, Ella, and I think many of them have to do with a little bit of false precision. When we get into something like false precision and we are talking about greenwashing, I think that’s something that does, not worry me, but I think it’s a concern that we all have in the ESG data area, specifically as it relates to carbon emissions. So let’s say for example, we use our SASB SICS codes, and you are actually, at this very particular juncture in food and beverage. However, you might look back and say, ooh, at my subsector, I’m seeing lower estimated emissions. What if I just use that one instead? And so that’s something that I think I worry about is when do we claim precision and when do users have the ability to override or elect to report other kinds of proxy values?

I think those are things that are really worth being really straightforward on. I think akin to greenwashing is that you could have people trying to game normalization in a way. So when we’re normalizing based on our proxies, that’s something that does give us really good confidence.

So we’re taking this median intensity value that we found for your sector that aligns to your geography, that aligns to your size, and now we’re trying to normalize that. What if we decided to normalize against FTE instead of revenue? How does that change the output? And so, anytime we start to sense that gaming or as users, if you start to see where that changes, I think those are things that might raise eyebrows or might just give you a little bit of pause to say, what is it that we’re trying to communicate? How important is that to be as clear as possible, and how do we really maintain the integrity of how we intend these proxies to be used? So we’re always trying to go back, layer in guidance, and also build our tools in such a way that we’re not allowing for such alternate views of reality, and that we’re being really upfront in our methodology and how different benchmarked proxies are derived.

Ella: Yep. And you spoke earlier about being as transparent as possible when you are using proxy data in whatever use case you are using it in. How could a CSO communicate proxy-based numbers externally, in a way that is very well received by external stakeholders?

Caitlin: I think don’t hide it. Being as upfront as possible is what’s going to get you the most credibility. So transparency in this case is everything. Put in factors like what proportion of this data was derived from proxy? So is it a hundred percent of your portfolio? Is it 10%? Is it select data points from a few providers that really struggled? Being really upfront, and we of course, will help our users do this in the appendices that we create through something like our SFDR reporting. But I think being really honest and upfront is the best thing that anyone can do. To say, “Hey, we don’t have this data yet, however we’re going in this direction, and in the meantime, this is what we expect it to look like.” I think the other thing that really builds and adds credibility is outlining what that trajectory looks like for improving primary data coverage. So you might not know if it’s going to be this year or next, but I think it can be really impactful to be upfront with what your state of affairs is. Have you just started using a tool like Novata? Do you anticipate being able to do this? How soon? All of that can go a really long way in building credibility, whether it’s with your own board or whether it’s with LPs or investors.

Ella: And how far into a company’s journey would you start to integrate proxy data in terms of, some companies have never collected ESG data before. Would you suggest that they start with the basics and then a few years later, once they have a good view then bring in the proxy data? Any guidance there.

Caitlin: I would say it’s never too early. In fact, one of our strongest use cases, say when you’re around venture, is when you’re trying to figure out what this is going to look like. So when you have pre-seed companies without revenue, it’s pretty hard to figure out. So we have no spend, we have no office, our employees are meeting in a coffee shop. That can be really difficult to get your head around and to get your hands around from a primary data perspective. So that I think is a great case for proxies is when someone is just starting out, they don’t have anything, where do I begin? Proxies is one such opportunity. It doesn’t mean that they should use them forever but if you were to say, hey, what should I expect as we start to build out our operations, as we get an office, as we grow those are all things that I think become quite observable over time and then there’s plenty of carbon accounting tools that can help. But that early-stage company, trying to figure out where it is they’re going to land is a really great use case for it.

Ella: And a big question for you. What do you think the future of proxy data looks like when it comes to sustainability data?

Caitlin: So it’s easy to get caught up, obviously, in the world of AI and try to use those tools to web scrape or try to do connecting all kinds of databases. It’s certainly easier than ever before to do any of those things. However, what I think people are also going to learn, in real time, is that these data ecosystems are also shifting—sometimes for the better, right? So we want more data to come into these ecosystems. We want people to be more data literate and more proficient in the ways that they use and apply and calculate their data. That’s all really important. But I think we’re going to see a bit of a hybridization period where there are going to be inputs coming from AI that have reviewed public-facing reports and people are going to want to combine one or multiple sources. So that hybrid data ecosystem probably coming in the future. 

And then let me tell you about what I would like the future of our own proxy data to look like. Every year we see a refresh of all of our benchmarks. So we are looking back essentially and saying, okay, across this now 2025 reporting year, what are the benchmarks that we can now collect? Be they universal, sector, sub sector, GP benchmarks, across all of those types be they size related on FTE or if they are revenue related. There’s lots of different ways that we slice and dice all of this data as it comes in. I will say it is a, it’s a really important process for us to keep and uphold. And because it is so important to our business, to our operations, to our proxies, to our benchmark products, we take a great degree of care in how those benchmarks are derived in the code, in where those notebooks sit, in how it’s connecting to our data sources.

And so as we, as always, increasing our sophistication every day, dynamically connecting to our own data pipelines as benchmarks can then be created and leveraged as reports come in. I’m not saying that’s what constitutes a benchmark, but I think we’re getting to a point at which we’re going to see data pipelines really become alive and help us understand our data in more real time ways than we previously have been able to across the last couple of years. But something we’re really proud of is the rigor that we put around our benchmarking process. And so expect in the future that this process is going to become faster, sleeker, more efficient, while also still upholding all of the rigor that have allowed our benchmarks to be as powerful as they are.

Ella: Amazing. Well, I’m very excited. And as a CSO yourself here at Novata, what two to three investments would you recommend that any CSO out there would make in being proxy smart over the next year or so?

Caitlin: I think with anything, it’s getting your architecture right. Figure out what it is you’re labeling things, how you’re calling it, where your data is being stored. I think the most effective thing I can do in my role is to ensure that all the data that we have coming in across various sources is as standardized as possible. Whenever I’m reporting, I want that data to come in, in a really clean way. I want it to be really understandable because then I wanna take and transform that data, be it for our public benefit report, our B Corp recertification, our own internal reporting to our board. These are all things that we have become super efficient at. And I think when you’re efficient at using and leveraging that, it’s because you’ve made those early investments in how you classify, how you receive, how you input data. So all of that is probably the number one thing to do.

I think in terms of being proxy ready, there’s lots of things all of our users can do today in terms of diving into Benchmark Explorer. So we’ll provide you the tools right now to figure out what normalized benchmarks exist and allow you to really deep dive into those. And that’s a great place to gain and leverage that understanding as well.

And so, that’s a really powerful and low-cost way to start getting acquainted with it and then see where you can leverage that source over time to become richer and more powerful as it relates to your reporting.

Ella: Thank you, Caitlin. Amazing insight and the fact that benchmarks and proxy data are so closely tied. And the sooner you start implementing proxy data into your process, you start building that muscle and you can start reporting on that data. 

So we always finish these podcasts with a few quick fire questions. So what would you say is the hardest judgment call you have had to make in your role as a CSO when it comes to the data not being perfect?

Caitlin: That’s a good question. I think deciding whether something is meaningful and material and then if you need to use it can be a complicated one for anyone. And is what I’m looking at reflective of a sector average, or is this really indicative of what I can expect in future performance? I think all of us are always weighing that. That’s a daily judgment call of do I want to apply this? And what does that mean for me?

Ella: And what would you say is the most common myth out there at the moment about proxy data?

Caitlin: Probably that it’s inaccurate or it’s bad, or that it’s less good. Of course, I’ve said a million times primary data. Okay, that’s great. We want that number one. But being really aware that well-designed proxies are often more consistent and bring in less bias than self-reported data. There’s a million and one ways that we all know self-reported data can go wrong.

How many categories are there in Scope 3? At least that many ways. So I think really understanding where our proxies come from, for example, which is rooted in private company benchmarks that have met very specific thresholds around rigor and methodological stability can really give confidence.

So when people are saying, oh what’s proxy data? Does that mean it’s fake? No, it actually is based on real data. We’re just trying to derive that in a way that’s most applicable for a company. So that would be my myth that proxy data is necessarily unreliable or inaccurate. We might say the opposite is true.

Ella: I really like the way you looked at that. And actually there are two sides to every story. And finally, if we could take ourselves forward to 2030, say, what do you think great sustainability teams will do differently with their ESG data?

Caitlin: I hope that people will be reporting on it because the regulatory structures will be in place that require that. So right now we see so many regulatory regimes that are relying on TCFD or IFRS, and there’s really good structure around that. However, that’s not the end of the story there.

So I think as we move forward, I would love to see more climate-based reporting, more risk-based reporting, nature, social. These are all areas that right now it’s easy to point to and say, GHG emissions, Scope, 1, 2, 3. That’s the data point. We literally see that in lots of the regulations that are coming out.

However, the picture of the world we live in is much, much broader. So, my hope would be that in the future, where biodiversity, nature, data, impacts to communities are actually all being taken seriously. And that we aren’t overly relying on the quantitative, that we’re really bringing in the experiences of the world in which we live because undoubtedly climate is going to affect us all. And so I think it’s really important to ground that in all of our lived experience.

Ella: Absolutely, and I mean, 2030 is just around the corner, so every person that can add to that journey and start contributing to that journey, the more likely we are to get there. Thank you so much for all your insights today.

My key takeaways around proxy data. Proxy data is not a substitute for primary data, we all know that, and it enables you to fill those gaps that you wouldn’t otherwise be able to fill. Number two, you need to set your own company’s tolerance when using proxy data and be as transparent as possible when communicating the value of that data externally to your stakeholders.

And then finally, you need to get your data architecture right. How do you label it? Where do you store it? How do you standardize it? And what is material to your business? And therefore how should you use it? So really interesting. But I think, the bottom line is get started now collecting ESG data, if you don’t already do it in a secure place where there are benchmarks. And use proxy data to fill those gaps that you couldn’t otherwise do. Until next time, let’s keep building sustainable intelligence together. Find out more at Novata.com.