Scott Tranter is Director of Data Science at Decision Desk HQ, the only provider of real-time election results, race calls, and data to cover US elections from the presidency and Congress to the county and city level. We previously spoke with Scott about his experience building and selling his previous company, Optimus.
In this episode we dig into the nuts and bolts of making race calls on Election Night in an accurate and timely manner.
Scott Tranter:
We need to see the votes. In fact, we're fervently or fiercely against anecdotal evidence when we look and call races.
Eric Wilson:
Welcome to the Campaign Trend Podcast where you are joining in on a conversation with the entrepreneurs, operatives, and experts who make professional politics happen. I'm your host, Eric Wilson. Our guest today is Scott Tranter, director of Data Science at Decision Desk hq, the only provider of real time election results, race calls and data to cover US elections from the presidency and Congress to the county and city level. We spoke previously with Scott about his experience building and selling his previous company, Optimus. In this episode, we dig into the nuts and bolts of making race calls on election night in an accurate and timely manner. Alright, Scott, so let's just lay the predicate here. Why is it worth all of this effort to have election result projections on election night rather than waiting for the votes to be counted? The
Scott Tranter:
Big answer of that, if we waited for the votes to be counted in most places, you'd be waiting 30 to 45 days because most people will realize if we don't have federal elections, we count the votes 4,000 plus different municipalities. So it's a township. If you're in a place like Maine, it's a county. If you're in most of the rest of America, these independent, we call them voting count units, they take their time and they should take their time. They got to do this right. But their process is, is they take several weeks to count everything and then post it and then it has to be aggregated up just because I know how one county goes, they're statewide and national elections. And so that's where our service comes in, is we go out and collect it on election night. We piece it all together and when these things take longer to count or longer to certify, we follow this pretty closely and aggregate it up and give it out to the new services.
Eric Wilson:
Got it. So we are able to get sort of proximate or likely data before it's official, official, official. That's why I think in your language, you always call it a projection.
Scott Tranter:
Correct. And I do want to nuance that a little bit. There is no modeling, there's no polling in this. We literally go out, sometimes we send a person to a county might send 'em, I know you're for Virginia, Eric. We might send a person out to Fairfax County. They walk into the election office and they say, Hey, can you give me a vitality sheet? And counties are usually very responsive. The vast, vast majority, in fact, almost all of them are. And if they're not responsive, it's just because, hey, it's nine o'clock at night, come back tomorrow morning at 9:00 AM and they provide these vote tally sheets and then we collect them. But we have to aggregate 'em up and sometimes don't count everything at election nights. They take a while and it's just getting these different updates
Eric Wilson:
And even setting aside this sort of commercial applications of race calls, which I want to get into next, there's research that shows that getting timely election results to voters increases confidence in elections. And obviously with some of the changes with counting mail-in ballots, we're seeing that there's even more efforts to just prepare voters say, look, we'll have an idea, but we won't know for sure and we're going to have to look at these mail-in ballots, especially if it's a close selection. And so outside of the customers and clients that you service, there's this value to the public as well?
Scott Tranter:
Yeah, no, I think that's what we strive to do is we want to make sure we count the exact totality. We want to be quick about it, we want to be accurate, but that's the goal is to accurately report what's going on in a timely manner so that people understand what's going on. I think we get a lot of questions, and you kind of alluded to is like, well, I've seen them count ballots faster and freelance and I've seen you count the pallets faster in UK and that is absolutely true, but they have a federalized election system and it's streamlined that way. We don't have it that way and that's not a good thing or a bad thing. That's just a thing.
Eric Wilson:
So who is relying on these election night calls and results from Decision Desk?
Scott Tranter:
So we have a number of clients, some of the scripts in their news stations, we do The Economist, we do all host digital publications like that and television stations. We do News Nation on table and things like that. Primarily they rely on us for two main things. One is the vote count, so they got pretty members to put up on the screen so they can help contextualize it. And two of the race projections, this is highly specialized in the sense that you've got to really know what you're looking at. You've got to understand that yes, this is a county that air, this is a county that Trump is, the presidential ones are kind of lots of information on, but we call over 40,000 races down to there's a mosquito control board race that was very close in South Florida,
Eric Wilson:
Important job.
Scott Tranter:
Those are the things we look at. And so I think that's kind of what we're here for is to help provide contextualization and race projections because race calls and certification is done by the state local officials usually a month later so the news media can report it out.
Eric Wilson:
Are there any other customers out there outside of the media category? Is anyone making trades and things based on election forecasts like this?
Scott Tranter:
Absolutely. I wish I knew exactly what it was. We used some financial services clients that fired data feeds and things like that. I don't really tell me what they do with it, but before it's fun. Basically US elections US is the largest economy in the world who controls it has some macro applications for businesses and countries around the world as well as businesses. The United States. I mean look at this election right now, we have the 2017 tax bill. We up p2 B up or not or modify next year. Who controls the White House, who controls the Senate, who controls the house, is greatly going to affect that. So there's a lot of business implications, economy
Eric Wilson:
Implications. And so you mentioned that you're sending people out to some of these key counties, but what are the various data sources that you are tapping into on election night that might be different from my command center where I've got the Secretary of State's website pulled up,
Scott Tranter:
The Secretary of States, let's say from California or something like that. They're usually half hour, hour behind because they're trying to aggregate from all their different counties. Whereas we go to all the counties in California and then we aggregate it up. So we're a little bit faster there. And then also, especially in some of these national ones like Pennsylvania is great, but you need to go up Pennsylvania relation to Michigan in relation to you need to string all this stuff together and I think that's where we're helpful. At the end of the day, we will not get a single of different tally than your local elections office, your state secretary state or anything like that. Our goal is not to get different numbers. In fact, we got different numbers and we very better our job. We just aggregate it faster and then put it in a way which the media can explain it. As you pointed out, you're a nerd, I'm a nerd. We like and look at state election spaces. They're not something that the casual person interested in who is in their government can go in and be like, yeah, this person won or this is why it is, or whatever, whatever information they're looking for.
Eric Wilson:
Got it. So speed kills in this case.
Scott Tranter:
Correct.
Eric Wilson:
And it's all the same data essentially, it's just different sensors or measurement devices I guess.
Scott Tranter:
Correct.
Eric Wilson:
Got it. I think that's really enlightening because in my mind I thought maybe you had some special secret sauce, but I mean you're right. It's like we need to know what those ballots say and there's only one source for the ballots. There's only one person counting the ballots. It's not like going out and getting a poll and I can do different samples.
Scott Tranter:
And Alex re said, there were other places, this is something I know DD HQ was started by a guy named Brandon Finnegan, and he was just sitting there and he's like, I think there's a faster better way to do this. And then he did it, right? This is the great thing. There's no secret sauce to this. There could be a competitor to us tomorrow where someone could say, no, no, no, I think I can get the information better. And that's the upside here is this information is freely out there. And I always like to point that out. If you're an American citizen, even if you're not an American citizen, if you're interested in arguing and you want to walk into a county election office at all China and say, Hey, can you show me the vote tally? They'll show it to you. If they don't have it, they won't have it. But this is public information. Vote is are public information. Our job is just to collect it at a wider scale than just the average Joe Schmo walking into a county office saying how many votes did Trump or airs?
Eric Wilson:
Yeah. And the campaigns are doing some of that work themselves too, where they're sending people, what role do these bellwether counties, we make jokes about them on Twitter all the time or certain precincts. How is this contributing to the election calls or projections that you're making in Virginia? We know what those counties are that we needed to win, or if the Democrats were going to win, we were going to have a bad night. Does that factor into it? Because you said there wasn't any modeling, but do you have those sort of pres signals?
Scott Tranter:
Yeah, we are trying to piece together enough data points so that we can get to be 99% sure of an outcome. The way I like to explain it is, look, let's say that candidate A has 10 votes, candidate B has eight votes, right? So you're like, Canada is two votes A, and then I tell guess what? There's one vote left to count and then I say, Hey Eric, who's going to win the race? And you're like, well Scott, of course it's Q 80 because even if that one vote goes to Q 80 B, it's them. You just went through the logic we're going through whether we're calling a presidential race or a Senate race, race control board races, all we're trying to do is figure out how many votes each candidate has, and we're trying to figure out how many votes are left, and let me expand that out.
Let's say that the candidate has 10 votes, candidate B has eight votes, and I tell you there are four votes left. Then you're like, well, Scott, I mean, I dunno, maybe candidate Bay gets three of those four. It comes ahead, or maybe it gets all four. And then you're going to say, well Weat, where are those votes coming from? Are they coming from places that are helpful to candidate B, or are they coming from places that are likely to vote for candidate A? So that's when those bellwether counties and those regions and that knowledge comes is once you know how many votes are left. If you're Harris and you're trying to win Pennsylvania, if there's a lot of votes left in Philadelphia, statistically speak to your area. If you're Donald Trump and you're Pennsylvania and there's a certain amount of votes left and they happen to be from Western Pennsylvania, well that's going to be a little bit more advantageous to you. So that's where some of these bellwether counties come in and just these knowledge, it helps us fill out that problem of, alright, how many votes are left and where are they going to go?
Eric Wilson:
Got it. So this is the way on election night when we're watching this and say with 1% of precincts reporting, we can project that. So-and-so won this state. It's really just a math algebra problem.
Scott Tranter:
Correct. I always like to call it intermediate election algebra. We could all do. I can teach pretty much anyone. And you got a mini lesson right now on how to call a race. The hard part is knowing the ins and outs. You've got a lot of experience. How does
Eric Wilson:
Henrico County vote?
Scott Tranter:
Yeah, you can even pronounce Henrico. Correct. Little things like that. And having to know this across all 50 states, there are some regional experts, there's some ins and outs, there's some trends and things like that. That's kind of the fun
Eric Wilson:
Part, really valuable knowledge that you have here. You're listening to the campaign trend podcast. I'm speaking with my good friend Scott Tranter about decision desk HQ and how they make election night calls and projections. So I guess you sort of answered this in a way, but do media reports or social media activity impact your analysis or reporting at all? So for example, if you're like, oh, we're hearing really long lines in these parts of this county or the Philadelphia suburbs, people are lined up or there was a problem with voting or it's raining in Milwaukee, does that factor into any of your, or you just, you're like, I need to see the votes.
Scott Tranter:
We need to see the votes. In fact, we're fervently or fiercely against anecdotal evidence. When we look and call races, the high turnout, the rain, all those cliches are banned or made fun of among our race teams.
Eric Wilson:
Got it. So that's just the opium for political operatives when we're waiting on race
Scott Tranter:
Called. Yeah, but there is quite a bit of data you could look at, for instance, if you get hard data from Florida's very good about giving official midday turnout reports, those I would trust someone saying, Hey, I saw a long line at my precinct location. I don't know about that, but when a Florida County tells me, Hey, this is how many people voted this morning, those types of things, there are certain data points you can get that are not anecdotal. There's just not as many of 'em as people would
Eric Wilson:
Like. Yeah. I think Florida, at least my sense, has been the best state to get us act together. Obviously they had an unpleasant experience about 24 years ago, and so they really put their act together, but seemed to be beef fast and swift.
Scott Tranter:
Florida's one of the best out there,
Eric Wilson:
And for better or worse, your fate pretty early on in the night. Okay. You've got this tension between getting it right so people trust you and your clients keep coming back to you. You don't want the Dewey defeats Truman headline to go out, but people are paying you for your speed. How do you resolve that tension? How do you balance those two factors?
Scott Tranter:
Well, it's relatively simple. We need to be fast because that's our job. But we could be fast out of hundred races, call on race well wanted, it doesn't matter. So accuracy, you're right, it is a eternal struggle or balance between speed and accuracy. And again, that's why we have a business, why there is an art and science to going out to all of these 4,000 plus local elections officials and collecting the vote on election tonight and aggregating it up and fact checking and QA and then saying, okay, I know exactly how many votes candidate has. I know exactly how many votes candidate B has, and I have an exact amount of how many votes are outstanding and I know where they are and now I have enough information to make a race projection. So that is the art here is maintaining the accuracy because that's important. Head to our job anymore if no one thinks we're accurate, but join it faster than others.
Eric Wilson:
I'm thinking of Minority report. You've got the Precog sitting in their baths there. Is there three people or three people have to agree on a call? Is there kind of a check there?
Scott Tranter:
Yeah, so that's a good question. So our race call team is about 12 people. This cycle. We're divided into roughly groups of three on some of the bigger race calls. So Precogs,
Eric Wilson:
Yeah,
Scott Tranter:
Yeah, yeah. Preco groups of three. You're assigned various races. And then on some of the bigger ones, I'm part of the decision making process. I oversee the entire race called T, but by and large, when we're going to call 30 thousandraces, it's not even election, we can't do that many, but there's several hundred what we would call marquee races, presidential in Pennsylvania, senate in Ohio. There's a lot of races like that, congressional races, things like that. We need three people to say, Hey, yeah, I am looking at the data and I am sure that this race is going this way. And if anybody has any inkling that, yeah, I'm pretty sure, but I'm not a hundred percent sure, then they don't call it. So it's kind of like a rule of three. We need three people to say, yes, this is callable, and again, this is humans.
We're not looking at a screen, we're not waiting for a computer to tell us no. We may look at a screen and we have models and things that inform our, but there's always a human in the loop to say, yeah, I hear what this model says, but I still need to see more out of Henrico. Or you might even get two people to say, yeah, this is ready to call. And third is like, no, no, I need to see more. Well, there's not an argument then. There's no arm twisting. There's like, okay, well you get there, let us know and we'll circle back to it. I think that's the important part. That's how we maintain some levity in the calls and quality.
Eric Wilson:
How long do you think humans are going to be involved in the process?
Scott Tranter:
I think for longer than either of you are going to be alive, because people just trust humans more, right? I mean, let's be honest. I think automation and some of this stuff is probably a good thing, but people don't trust what they don't understand and they trust humans and humans are much easier to be held to account, at least in people's eyes. And it's efficient enough and there's accountability and there's training and things like that. So at least in our job in terms of vote collection and vote calling, I think there's, for the foreseeable future, there's humans.
Eric Wilson:
So worst case scenario, what happens if you guys call it for Dewey, but it ends up being for Truman, what do you do then?
Scott Tranter:
Well, that's an important question, right? Race calls from time down, it has to be retracted. And that is part of one of our things is when we make a retraction, we explain why.
So we had to make a retraction in a 2018 congressional race, and so did some of our competitors associated press. We retracted the call a week ahead of time, week before they did. And then we explained why and we explained why. Not in a statement. In a statement, but also on the record with our media clients and not media clients. I think that's important because statistically speaking, a missed race call is going to happen, right? If you drive a car long enough on the road, you won't get a flat tire, right? If you year and year out, you're looking at races, you will have a missed race call. So what's important even in mind is when it happens, how you handle yourself. And as you pointed out, trust is important. People, especially this election, it's important to explain to people why and hold yourself accountable from a business sense, and again, this is, I'm big fan of capitalism. If you make too many of these mistakes, then you will not get paid. And if you're not getting paid, then you probably shouldn't be doing the job. And so I think that's an important feedback and quality control mechanism in our industry is none of what we do can be bs. Our job is to publicly proclaim and put these things out so when we fail, it's out there. I can't hide it behind a, this is beta or you heard this happen? This happened in a dark hanger somewhere. No, this is very public.
Eric Wilson:
Yeah. So obviously Scott, we've got elections year round, but this is kind of like your Olympics once every four. You have a presidential race and all eyes are going to be on you guys on Tuesday night. What are you doing the other 1,460 days to get ready for that night
Scott Tranter:
Waiting for GTA six to come out? No, that's a good question. So look, there are about 40,000, 50,000 elections that are going to happen in 2025 starting in February. And then we got midterms where teams really small. As you point out, we can't maintain the largest team as we have today year round. And so we're about 10 full-time folks and there's local elections to have, but our job gets a whole lot less public profile, and we spend a lot of time analyzing the races and trying to get better. One thing I like to note is because the election process and how it's counted is so diverse and different between each county. Just keeping up with who's running this county and what's the group they use? Oh, they've got a new, all that kind of stuff. That's kind of what we do in the off season and stuff like that. But yeah, no, my life will be considerably less complicated hopefully in about 30 days.
Eric Wilson:
Got it. Well, my thanks to Scott Trainer for a great conversation. We got to let him get back to work. Preparing for election night. Make sure that you have Decision Desk HQ bookmarked and opened all election night where I'll be watching things. If this episode made you a little bit smarter or gave you something to think about, you've got a few fun facts to share on election night while you're waiting for some of these races to be called, all we ask is that you share it with a friend or colleague. Remember to subscribe to the Campaign Trend podcast wherever you get your podcasts, and that way you'll never miss an episode. You can visit our website@campaigntrend.com for even more. With that, I'll say thanks for listening. See you next time. The Campaign Trend Podcast is produced by Advocacy Content Kitchen, a media production studio up.