When polls closed in Florida in November 2018, it appeared that the Republican candidate for Senate, Rick Scott, was the victor. With 100 percent of precincts reporting, he held a slim lead over the incumbent Democrat, Bill Nelson.

Over the days that followed, though, the race tightened. There are always ballots that trickle in more slowly in elections, more in some places than others. Normally, there's no urgency to tallying them since the results of the contest are clear from the votes counted elsewhere. In this case, though, Scott's lead was narrow enough that the votes did matter, and so political observers watched them closely.

Scott, who at the time was the governor of the state, held a news conference implying that there was some sort of fraud at play. His highest-profile endorser, President Trump, joined that claim, suggesting that the votes being counted were somehow fraudulent. At one point, Trump proposed throwing out any votes that were counted after the results that were announced on election night, for no apparent reason other than that he didn’t like what they said.

This suggestion, offered either in bad faith or from a shortage of familiarity with how elections work, was ignored. In the end, the additional votes didn't shift the result enough to cost Scott his Senate seat anyway.

It was a good reminder, though, of how incomplete election results can lead to incorrect perceptions of the outcome and even shape how the rest of the votes are counted. Elections officials in Florida — in particular in Broward County — were under enormous public pressure as the additional votes came in, pressure aimed at one particular result. Scott, like George W. Bush in 2000, was able to approach the count (and subsequent recount) from the position of having been the apparent winner based on the initial results, giving him leverage in any arguments about how things were progressing.

What happened this week in Iowa was much, much worse.

At stake was not a Senate seat but the Democratic presidential nomination. The delay in vote counting wasn't centered on scraping together the last few votes in a close race but for the votes in their entirety. It wasn't a one-off election, but the first contest in a lengthy series of contests, lending additional significance to what happened. The results in Iowa help shape perceptions moving forward among candidates, pundits and voters.

When Iowa Democratic Party officials announced on Monday night that they would hold off on releasing results from the evening’s caucuses until they were confident they could do so with accuracy, it was the right call. Imagine what would have happened in Florida two years ago if Florida had announced that Scott had a 2-to-1 lead, say, with 62 percent of the vote in! Anything from that point forward would have seemed as though it was working against an obvious Scott victory.

Then, on Tuesday, the Iowa Democratic Party decided to release the results it had, some 62 percent of the total precincts that caucused. And just like that, a political ecosystem that had spent months building up the significance of the Iowa contest had something to work with.

There were all of these big machines just waiting for data about the outcome in Iowa, machines into which you could dump the results and out of which would come news stories, shifts in candidate spending and new attention from voters. Some of the machines are literal, bits of software aimed at processing race results and determining the direction of the primary moving forward. Others were figurative, processes left over from past primaries or ones constructed with an eye toward the boundaries of the current primary field. There are machines that feed the machines, predictive tools with short shelf lives aimed at giving an answer to the “Who won?” question as soon as possible. All were ready on Monday night and then Tuesday afternoon for some fuel, any grist to start operating. The Iowa Democratic Party dumped in what it had.

The party, hoping to avoid criticism it faced following the 2016 caucuses, decided to release multiple sets of data about how the caucuses went: two sets of voter preference data and two sets of delegate calculations. Before the caucuses, the party’s website explained that this multi-tier release was because its role was to “not declare a winner, the party’s role is to present results.” But the party is certainly aware that the results it presents have enormous power and that what it presented would be parsed specifically to figure out who won. Saying it wouldn’t declare a winner was offered as a bit of self-aware objectivity but, in reality, was more of an exculpatory clause.

Those first results showed a slight lead for former South Bend, Ind., mayor Pete Buttigieg — on one metric, at least. Buttigieg led Sen. Bernie Sanders (I-Vt.) on one of those delegate calculations, state-delegate equivalents, while Sanders led Buttigieg on voter preference estimates. Since state-delegate equivalents were the traditional metric by which victory was assessed, though, Buttigieg had the space to declare some sort of victory. (Not that the lack of data had prevented him from doing so on Monday night.)

The campaign saw immediate benefits, according to top adviser Lis Smith.

Suffolk University, the Boston Globe and WBZ TV are conducting a tracking poll in New Hampshire, the next state to vote. In the poll conducted on Sunday and Monday, Buttigieg earned the support of 11 percent of respondents. In the poll conducted on Tuesday and Wednesday, he'd jumped to 19 percent.

These shifts for Buttigieg are inseparable from the data released by the party. Those machines were hungry for a winner so they could start grinding away, and now they had one.

Remember, though, that Buttigieg was only the leader on that one metric. On others, Sanders performed better. Those initial results did provide some useful information, including about how poorly former vice president Joe Biden had performed. But even if they’d been complete, the split result between Sanders and Buttigieg made establishing the “winner” trickier than it might otherwise be.

It’s important to interject that establishing a concrete winner in this case is an exercise in uselessness. Sanders and Buttigieg, as I wrote on Wednesday morning, essentially tied and will split Iowa’s 41 delegates — 1 percent of the total that will be awarded during the primaries — roughly evenly. We don’t need to know who won since it’s not as though only one of them is moving on to the U.S. Senate, as was the case in Florida in 2018. The Iowa caucuses, instead of being a discrete election, is more like one precinct in the broader, slow-moving national election that are the primaries.

The results coming in from Iowa that led to the Buttigieg burst were coming in chunks. That itself should trigger warning bells, since some chunks of data can be more representative of the final result than others. Imagine, for example, if most of the votes that had been reported in Florida in 2018 were from the big cities. It would have seemed that Nelson, the Democrat, was faring particularly well. In this case, a big missing chunk was the satellite caucuses introduced this year, a new factor in the contest that was still missing.

On Wednesday evening, a big chunk of those results was reported — and it gave Sanders a big boost in the state-delegate equivalents count. According to the most recent results, with 97 percent of precincts in, Sanders and Buttigieg ran about equally on final voter preference and on that delegate metric. What was functionally a tie on Wednesday morning evolved into a tie in near-literal terms.

The machines, though, were already chugging along.

What would have happened if that first 62 percent of results had included those satellite caucuses but not some area where Buttigieg did well? The first results we got might have shown Sanders with a lead across all three metrics — and made it seem as though his victory was clear-cut. He’d have gotten the benefit of those two days of the machines operating under that assumption. Buttigieg’s strong performance would still have been noteworthy and may have led to a surge in interest, but it would probably have been more muted.

The Iowa Democratic Party layered mistakes on top of mistakes. Its release of incomplete data was a predictable risk, one that had predictable effects. We know from incidents like Florida two years ago that the counting of votes can itself affect perceptions of victory and, in the Iowa caucuses, those perceptions are broadly a much more important result than the delegate assignations themselves. The party tried to wash its hands of analysis of the results in advance, but its power stems directly from the extent to which the world assigns value to what happens in the caucuses. Those machines steamed along with imperfect data, which is a problem, but the party knew the machines were there — needed the machines to be there — when it threw out what it had.

We can’t answer the question now, but it will be worth asking down the road: Which was a more important factor in the Democratic primary process, the views of Iowa voters or the way in which the party revealed those views?