Rui (00:21)
Hi, Tristan.

Tristan A (00:22)
Hey Ray, how's it going?

Rui (00:25)
Great, it's Sunday and it's sunny outside, so I'm really happy.

Tristan A (00:30)
it's a good day when it's sunny in the Pacific Northwest always.

Rui (00:35)
Thank you for first ever episode of Floating Questions.

Tristan A (00:41)
you for having me.

Rui (00:42)
you like to give the audience a little bit self -introduction?

Tristan A (00:45)
Yeah. Tristan. other things, I work professionally as a data scientist. I kind of have like an inclination for -oriented work, which is sort of my last couple of years of focus have been. I know you from our time at Deloitte together, we had a chance to work on some projects

And we find that we had a lot of overlapping interests. when I'm not thinking about work, love all the outdoor stuff. if you're confused about my accent, it's because it's French and South African.

Rui (01:16)
were born in South Africa.

Tristan A (01:17)
actually I was not born in South Africa. Nope, I was born in Mauritius.

grew up in many places including South Africa eventually. So I did boarding school ages 9 through 18 in South Africa. But prior to that, it was always between France and so I think I'd around a dozen countries before I started boarding school in South Africa. lived in like Tunisia, Yugoslavia, Morocco.

Mexico is included in the Caribbean, etc.

I joke that like my role is just the foreigner basically because if I'm in France I'm South African and if I'm in South Africa I'm French you know and if I'm in the States I'm like whatever so it's just like yeah yeah I'm just that guy that they grew up elsewhere

Rui (02:03)
you

did you move around so much and I presume it's because of your parents but like what made you guys move around so much?

Tristan A (02:19)
Yeah, my parents were spies. No. My parents worked for a French vacation company called Club Med. are a resort company that was, I they had their heydays in the 70s and 80s, but what they were very strong at was new holiday destinations and opening up resorts there.

Rui (02:22)
for sure

crazy, honestly. That's a lifestyle that's very difficult to imagine for me. Because I grew up in one place, did you feel as a kid growing up, lot of people moving from place to place, probably they struggle with maintaining friendship and have a sense of belonging you go through those struggles?

Tristan A (02:47)
I'm

think that answer is always difficult because you're sort of looking back and trying to infer. But I guess what I can say is that just your perception of things as a kid is, it's just not the same as an adult. lens through which you judge things are very different. ⁓ been told that happened like at least on one occasion where I had made like friends with of the

guest kids I was visiting for a week and then would cry when they left at the end of the week.

What I can say for sure is that it's made me very and adaptable to cultures. can drop me on earth and I'll be fine. I'll find a way to get on with people and travel and get my way around. So I think that's kind of one of my skills that I find in my career is that I speak to people in the boardroom and I can speak to farmers.

All the same, know, no problem.

Rui (03:52)
well since you brought up can you explain a little bit were you doing?

Tristan A (03:59)
in my previous role, I worked as the Director of Data Science and Engineering KOMAZA, which is a forestry social enterprise in Kenya that works with smallholder farmers to grow sustainable wood, quite a mouthful basically we were the biggest forestry company in Kenya.

we had this interesting model where instead of burning any land and growing, forests on massive tracts of land, like most forestry companies do, worked with farmers that had, half an acre, small size plots the idea that these trees could become a source of revenue for these smallholder farmers allowing us to grow wood.

in a more cost effective manner have other positive carbon offsets and all that. that was Kamaz as a whole. And then my role there was essentially involving everything data. managing the data sets from all of the distinct business units doing some of the modeling, forecasting, things like that.

Rui (05:08)
wow, there's a lot for me to unpack there.

just want to make sure that I got the business model It's a social enterprise. it has presume, two folds of.

One is one is the social impact The economic one is that you would, with the small farmers, they have small lots of land, and then they grow trees. And then once the trees mature, you guys would take the trees to sell on the market. And the social aspect of it,

is to help farmers to some returns on their land or what is the form of social return?

Tristan A (05:50)
Yeah, think you said it pretty correctly. bottom line, if you will. bottom line is just to run a profitable company. And the other is trying to have positive impact in the the impact they will focus on was in the field of economic development, lot of the farmers that we were working with, were specifically the kind of farmers that were working in

areas with degraded land or basically not getting like a lot of from other nonprofits that try to help farmers their yields And the idea was that, look, if you have this extra land like you typically putting in the time, it's just not worth the effort, returns on it, plant some trees there.

give you the trees, we'll give you the training to know how to manage the trees, all this for free. And once the trees are harvested, that one -time payments a rough sort thumb, would be enough to put a kid through school.

Rui (06:53)
see. you know what was the funding story?

Tristan A (06:58)
yes, it's interesting.

through different iterations. Initially, it was actually a nonprofit. So very much focused on like, let's see how we can help farmers and let's see what we can do. I think the founder, who was an American guy, Tevis,

on different things until he was like, okay, trees, is the way to go. I think after a while they realized that we do this as a social enterprise, it sort of opens up the opportunities for the kind of investors

can have and like maybe the scale of impact is going to be more important.

kind of when they switched to the social enterprise model. it's a forestry company, first thing they had to get right was seedlings, then they had to get planting right, and they had to get growth right. so around the time that I joined, they had small sales team.

And we were basically opened up the biggest sawmill Kenya and I think East Africa maybe. so there was like a big push on sort of developing that like latter part of the work and really successful as far as commercializing the sale of timber.

Rui (08:07)
Very interesting. long does it usually take to finish harvesting cycle? Because if the cycle is really long, might run into some problem,

Tristan A (08:17)
long answer is it depends, but I guess the short answer is eight to 10 years, at least in the areas that we were looking at. had like a very strong corporate finance team. And one of the things that they did is they created this vehicle, I believe is the term,

did an exchange where we will you the trees, we've planted them, at age two or three, so you can infuse some cash into the business And at maturity, the business can buy back the trees at cost so that they can harvest it and sell it.

think vehicle.

was created when the company was around to 10 years old So some of the very first harvests were to come to mature like the volume difference between year, and eight and the quality difference and all that would vary. not exactly a startup, there were some aspects startup -like where we believe this.

to be successful in the long term. So we're going to test this out, still running at a deficit

Rui (09:21)
the funding essentially is from this vehicle or there are other sources of funding as well.

Tristan A (09:26)
There were other sources of funding as well, definitely. investors sort of doing that model of like, let's grow and then get to profits.

Rui (09:34)
Interesting. enterprise has two goals, one economic and another one social benefits. Sometimes these two be at odds with each other. you seen such examples at Komaza?

Tristan A (09:47)
Yeah, that's, it's, it's a great point. the two can be at odds, most obvious one is that if you're just a nonprofit, would go and help everyone out no matter situation. Whereas in Komazu, there were farmers who were like, we can't work with you, unfortunately, because your land is too small. or your farm is

so far in such a hard to reach area that the cost of transport to get to it and extract the wood would negate any sort of profits people were very realistic about, we can't help everyone.

Rui (10:22)
Hmm.

I see. And

also mentioned like the specific role that you're playing there is the director of data science and engineering. a little bit of the kind of problems that you're trying to solve there.

Tristan A (10:38)
first big project I work on was getting Vera and VCS certification.

let me explain carbon credits really quickly. major corporations say they are trying to go climate neutral, of the main ways that they achieve this is buying carbon credits. And so what that is there'll be some of action or that a company does that

prevents a certain amount of carbon getting into the Komaza the planting of trees is what captures carbon.

capturing is great, but needs to meet a certain threshold before you can actually sell it out in the open market. And so this is where getting accreditation comes in. capturing all of this carbon and we'd like to be able to sell it in the open markets that people can offset the impact. companies like Vero and VCS are the ones that'll be like, all right, well,

These are the things you need show prove that your credits are sufficiently good enough a quality where you can go and sell it out in the open market. of the first projects was basically just trying to figure out much carbon

Komar's tree is actually sequestering.

Rui (11:50)
Fascinating,

how did you do that?

Tristan A (11:53)
guess the brief explanation is sending our people to sample collect data on a lot of trees and then extrapolating that using a specific formula presented in the certification documentation to then come up with a total estimate.

Rui (12:10)
I'm actually interested in piece of collecting the samples. Do you hold some type of equipment and go around the trees and try to measure the amount of that's coming out of the tree?

Tristan A (12:21)
No,

good thing forestry has been around for a long time. can use these you take one measurement and from that one measurement, you can pretty much extrapolate the rest. usually what that measurement is, it's called the DBH, which is the diameter at breast height. So imagine you ran a around the trunk of the tree.

at chest heights of grown man, measurements, the DBH, plus the species of trees, you can go from that one measurement to basically what the biomass is of the tree.

from biomass to carbon. And so main thing that we had to work out was all the trees are planted in the same area and not all the trees are planted in the same year. so our problem was need to do, sampling according to.

where the trees are planted, when the trees are planted. obviously we can't go out and measure every farm. So we need to go and measure subset of it that it meets the standards by, carbon estimation.

lot of the work was let's define our different strata. Let's figure out what the strata represents as the total volume of the portfolio. So obviously like later years had more trees, therefore we needed to sample more of the bigger strata. And then

Let's then coordinate this effort where we had to go and measure, think it was farms or something. All right. Let's coordinate the field teams to go and measure farms and how do we select the farms? the fun part is figuring out how to deal with all of the things on the ground that might go wrong for sampling.

like, the road has been completely over in a flood. can't get to the farm to, I'm trying to measure a tree, but there's big beehive on it. So it's not safe for me to get to it. a lot of that, a of that. So it's just, it was this interesting sort of conversation back and forth we've succeeded in

Rui (14:19)
You

Tristan A (14:29)
getting these data, but we didn't succeed in getting this. how do we get those two to work together so we can still meet the standard for the carbon project?

Rui (14:40)
is a fascinating one because I think, especially for people who are used to technology world, the data like it's automatically generated by your for -tech heavy enterprise, the data has to come from somewhere.

And a lot of logistics and operational processes needs to set up in such a way that you can easily collect the data

Tristan A (15:04)
could spend a whole podcast talking to you about the kinds of issues on the ground that we would run to in theory it should be easier to go to that farm made of those trees and every reason why that might not be possible.

Rui (15:19)
know this data isn't perfect. How do you determine the threshold where, this is good enough for calculation?

Tristan A (15:29)
did some resampling of some farms. So we specifically went and had teams, separate teams, I should point out as well, go and collect the same farm twice to try and understand how much difference were there in any the one -team's measurements and all that.

there isn't exactly a hard and fast line. there's no one telling you what's right or wrong. It's just sort of like, okay, the best you can do is just try to the scale of what's the potential range and then.

some maybe smart people on my team would be like, all right, we can actually use that to interpolate into the equation. Or we need to move forward and let's just use this data and if we do this next time, like what are some of the things we should be aware of, you know?

Rui (16:18)
I mean that totally makes sense at the same time because this is not financial metric you either have the money or you don't regulator they wouldn't have the correct answer for how much you are actually producing you also don't really know you can try your best know for people who have great ethics

who want to do the right thing, they will try their best, but at the same time, because there is such a lack of, almost like the market overseeing the accuracy of a financial report, do you think this actually creates a lot of problems the social enterprise?

Tristan A (16:54)
Yeah, I think what you're saying is that there's less objective truth does this create more of an issue? even the auditors that they send to go and verify your work, realistically, we knew can only actually go and sample so many farms and all that.

So yeah, there's definitely, unfortunately, more scope for

I think it is documented. believe there was John Oliver episode on credits and just how unfortunately, like a lot of that is being questioned because A, on the one side, it disincentivizes disincentivizes like the large corporations to change their

behaviors, right? They're well, we can just buy credits. the other side, you have places like, lodges for private hunting that are selling carbon credits, because they have a couple of trees their farm, the reality of it,

think it's the attitude of not letting the perfect get in the way of the good, is kind of the best we can do.

with all its flaws

Rui (18:02)
I see. world is deeply imperfect and then we have to operate within that. And maybe we can move on to a different project that you were working on besides the carbon,

Tristan A (18:15)
the satellite imagery was something we didn't fully get off the ground. But the idea was simply this. had somewhere on the order of 50 ,000 distributed small plots over a geographic area the size of half of Kenya. you have one big forest of trees with all uniformly planted and all that,

it's fairly easy to estimate like this square area and, density and then it's really straightforward to be what this is, how many trees and what I have. When you're working with many farmers plots that are all sorts of weird shapes becomes much more difficult problem.

You can't forever be sending out people with motorbikes to go and measure the trees. That's not scalable the idea was we wanted to develop a way to be able to detect our farms from the air.

is there a change in this polygon? we see it getting greener? Do we see, trees getting larger Or do we see that, of them are missing? know, the trees have been stolen or eaten these things have happened. that was the satellite project.

And then sometimes we just had much more simple operational stuff all the farmers had to have paper contracts for, reasons? Because you're working with people in the field and they need to have something to hold on to. And our field team also had an app that they used when they were

the farmers and capturing the data to location of the farms and whatever else. things, which should be speaking about the same people often didn't match up. And so what was happening is that year after enrollment, the data team or of the data team would spend month just basically checking these paper contracts against what's registered in the app to make sure information was the same.

I'll sort of value add to that is we the form so that it would be optimized to be by computer vision. So having boxes where we know certain data points would be. then being able to use that to sort of automate the process of extracting the data from those paper forms and comparing them against the database. that was, I

to 500 working hours saved just on that.

Rui (20:36)
Makes a lot of sense. you take one step back, are the other issues that are more systematic that you find

difficult to solve without a larger move in the picture.

Tristan A (20:54)
We discussed a little bit of just the algorithm kind of context of things.

some of the things are just

building a solution that needs to work. So I think you made the analogy where build a model and look at science as like building a car, the model is the engine and then you've got all the other things around it that need to work the engine to run. So we're talking about the wheels and all of that. And so I think for that, there's just a lot of

do you get people to use the tool? Right. How do you get to folks to understand like, all right, this is what data can do. This is what data can't do. just having ownership of the data, I think was just some of the big stuff. this one drove me out of the wall. had different business units

was using imperial measures and the other one was using metric and they were adjacent business units and then they surprised they couldn't agree on how much wood was flowing through. I'm like, are you kidding me?

Rui (21:57)
Right, our personal chats, I think we really talk a lot about like, how do we think about the picture at large, how to design, help morph a system to a place where it's also suitable for, you know, growing successful data science project. example,

information on the documents, maybe you run into problems like what happens when people don't have a good camera on their phone,

Tristan A (22:24)
mean, you say that, was literally had to tell them take the photo in the right orientation or please don't scrunch the paper.

Rui (22:31)
the supply chain with an entire ecosystem around data, And it's kind of mind blowing. Even for big tech that contacts with, I talked to them, data quality is a huge issue. that for

social enterprise that is not natively, in the sense that the core business is not product, It's very, very hard. do you see that we can shift that a little bit?

Tristan A (23:02)
talking about a data literacy problem, I'm being really open minded about this, I would say it sort of goes both ways. the data science teams also don't understand can't you get the tree measurements right? Like, it's not that hard to just run a piece of tape around the tree. know, our best tool actually to work against that was

Rui (23:14)
You

Tristan A (23:20)
sounds dead obvious, but it's so true. would send our data teams to go and do some harvest and tree distribution. So they would see okay. Now understand why these guys are complaining. I see why it's hard to get data. an extent we were trying to do more and more the field teams and showing them Hey,

These are the results of your data. This is why we sent you to collect the tree information. This is what we can understand from it. two way as far as like educating teams. I think that's kind of a big part of it. it's just the other people can't imagine what they don't know. It's really what it comes down to.

I think the best thing you can do is find the right people to show what's possible and have them sort of convert their teams to your side.

almost like a design problem. the one hand, you absolutely want people that are really strong in in algorithmic terms, right? Like it's literally the engine, as you say,

the team to really be successful, you sort of need to complement those with people that like, no, I need to understand how this algorithm is actually going to impact the humans on the other end and sort of all of the human things that are going to go along with it. That are going to make the make or break the differences to actually whether this model is used or it's used successfully even goes back to frankly, Deloitte. you could build something really great.

and folks would be like, yeah, but I know, existing process. I'm comfortable with it. unless you clearly show me why this is going to help me, don't care what so -and -so manager said, they're not doing this work. So I'm not going to use it. Plain and simple.

Rui (25:02)
now that I'm of the consulting world and taking a step back about the type of work that I was doing there and also the type of work now that I'm doing in order to, deliver something end to end. In consulting think the incentive structure and also the way that usually it's being operated, it's we come up with some type of strategy or proof of concept.

delivered to the client and you're like, use this. the, and then we're out.

Tristan A (25:29)
and then we're out. Exactly. So

what happens afterwards?

consulting was interesting is it just sort of inured you to this is where my project ends and everything that happens subsequently is basically somebody else's problem. reality of company is that know if you want this to go all the way through, the subsequent things are your problems as well. And, you know, sometimes it can be as much

to solve there as the model development itself

Rui (25:57)
for sure. Is Komasa still around?

Tristan A (26:03)
Unfortunately not.

Rui (26:05)
What do you think?

was the gravity that it down. Is it the business model? Is it incentives? Is it something else?

Tristan A (26:20)
I think there was a little bit of a disconnect between what the investors wanted and what the founders were aiming at. this is just my opinion, but I think there was also some disconnect created by the fact that there were different investors that had sort of different, points of view.

investors were viewing this really much as an operational forestry company that needed to pretty much be profitable from the get go. And I think some parts of the company were like, no, let's do like a Silicon Valley, startup model where we're going to run at a deficit for a long time. And once we reach a certain scale, then we're going to start turning profits.

Rui (26:51)
Mm.

Tristan A (27:05)
those views are just not compatible. so some point the that were sort of financing the operational company model were just like, we're not behind just throwing good money off the bed. So we're out.

Rui (27:22)
I see. sides have their own, and perspective. It's just unfortunate, are so many clashes and misalignment.

Tristan A (27:30)
Yeah, exactly. Both are models. they're not the same business model. going to have very different expectations of what good looks like.

at the end of the day, that sort of disconnect came to a forefront was viewed as money coming in, didn't come in. And then it was just the realities of the cash flows, dry pretty fast after that.

Rui (27:56)
at least you got so many interesting insights into the stories behind each data point. It sounds like a very adventurous experience, honestly.

Tristan A (28:08)
I really enjoyed my time there. really enjoyed the people I worked with and I thought I was incredibly privileged to be able to do data science work and get to up in and then hop out of Africa and get out to countryside and interact with farmers and stuff. just thought this, is so far from I was doing before know, traveling to

big cities and in boardrooms, very happy for the experience. think we did do some good. Obviously it's not the ending that have preferred. I think there was some good learnings and I hope the people that were involved in this can sort of do successful version of it, whatever that might be.