The electronic system used by the Iowa Democratic Party for the first time to compile its 2020 presidential caucus results was only counting “partial data,” IDP Chairman Troy Price said in a statement Tuesday morning, giving the most specific clue about what went wrong.

That partial data—called a “coding error” in the most recent national press reports—was most likely tied to three different sets of figures that the IDP planned to release for the first time after the caucuses ended, but did not due to what the IDP called unspecified “inconsistencies.”

The IDP announced at midday that the “majority” of results will be released at 4 p.m. local time. (As of mid-afternoon, the IDP said it may be “half.”) 

Those three sets of inconsistent figures—results details that the party has never released before—could only refer to steps in the process where the number of participants and the votes cast in two consecutive rounds of caucus voting did not all match.

The IDP has not said more about what went wrong with its tabulating system software, which was never tested before on the scale used Monday. But it is possible to identify one discrepancy in the numbers that would have been reported via the IDP’s app to its software and system and could have caused the “partial data” and “inconsistencies” in analyses.

A likely cause of “partial data” may have been the process itself, according to Voting Booth’s eyewitness observations and assessment (based on undertaking Iowa’s caucus chair training and numerous interviews with top party officials, including a demo of the app last week).

The “partial data” or data mismatch may have less to do with the app used by caucus chairs to report the winners in two consecutive rounds of voting and the resulting delegate allotments (although caucus chairs and campaign precinct captains had problems with getting online and logging into their reporting and tracking systems). Rather, not everybody attending a caucus voted in the second round if their top presidential choice was disqualified in the first round.

That pattern of drop-off voting could have produced the “inconsistencies” that were seen by the IDP boiler room. In the demo by IDP Executive Director Kevin Geiken, the app showed when too many participants were entered into its calculator (called an “overcount” on the app), but it didn’t report undercounting or intentional drop-offs in the second round.

This very scenario was seen in Polk County’s 57th precinct in Des Moines when 11 voters of the 385 attendees did not vote for another candidate in the realignment round—after their first presidential choice was eliminated.

These voters mostly came to vote for Joe Biden and were overheard saying that they could not vote for anyone else, especially after the race’s other centrist, Amy Klobuchar, also had been eliminated in the qualifying first round. They didn’t want to cast a vote for Bernie Sanders or Elizabeth Warren in the realignment.

In other words, they came. They voted. Their first choice lost (wasn’t viable in that room). They didn’t pick another candidate. Thus, there were gaps between the number of participants in the two rounds of voting. That could account for tabulation software in the IDP’s electronic backend seeking balanced totals—via an app on each of 1,678 chairs’ smartphones—reporting figures that didn’t match or balance out.

Or these same voter falloff numbers could have also appeared in called-in results that the state party was receiving if the caucus chairs could not connect with their app to the IDP’s backend. That also happened in Polk-57, where Caucus Chair John McCormally could not get his app to log in before and during the event, trying several times.

So McCormally ran the caucus using math, pens and paper, and then called in results. In his case, McCormally reported the results quickly (possibly because his wife was a volunteer at the IDP boiler room. But other chairs across the state encountered waits of 90 minutes before talking to the IDP.

This quagmire deepened inside IDP headquarters when it had to assess the growing problem. Top state party officials had to find a solution or use a backup plan, which party officials were bullish that they would not have to use a day before. That backup entails gathering the paper summary sheets from every caucus chair. Those summaries list the voting round totals and delegates won.

The summary sheets, which are signed by precinct captains from all the campaigns, were to be turned in to IDP county chairs (along with presidential preference cards filled out by voters), according to the caucus training materials. The county chairs, in turn, were to turn in all their paper records into IDP headquarters either physically or by mail.

Before caucus night, IDP Executive Director Kevin Geiken was asked about worst-case scenarios in a demo of the caucus app that only two reporters attended—including this writer. It might take a day or two to physically collect and recount the full paper vote record, should the electronic system be jettisoned for whatever reason. He emphasized there would be a reliable and accurate count, but it might not be as fast as expected.

The IDP announcement mid-day Tuesday that the “majority” of votes would be released (before updating this to “half” the votes later that afternoon) suggests that it made a dash to collect and count as many ballot summary sheets as possible.

When Voting Booth covered this ‘what comes next’ scenario before Iowa’s caucuses, few state and national party officials imagined that the reporting system would melt down. They expressed great confidence in the party’s voting system and its private contractors. Caucuses are not run by government election officials but rather use rented voting systems.

Even hours before the caucuses began, these officials downplayed reports that some precinct chairs were having trouble signing onto the IDP app and the possible consequences.

But voting technology experts predicted these problems. Reliability issues are to be expected when a new system debuts, especially one that has not been tested at scale and when its users encounter access issues (insufficient bandwidth and inability to log in) atop software glitches. That is why government election officials like to debut new voting systems in low-profile races.

Iowans and everyone else will get to see accurate results eventually. That is a silver lining. The IDP will be using a paper trail, and that paper trail will be more detailed than in any past caucus.

The only other silver lining might be the national media will realize that it may not be possible to report fast and accurate results on election nights. Many other states will be debuting new voting systems in caucuses and primaries.

Steven Rosenfeld is the editor and chief correspondent of Voting Booth, a project of the Independent Media Institute. He has reported for National Public Radio, Marketplace, and Christian Science Monitor Radio, as well as a wide range of progressive publications including Salon, AlterNet, the American Prospect, and many others.