Understanding the American electorate has never been more important or more challenging, especially for journalists.

We know the country is polarized and that voters in both parties are angry. But knowing exactly why, or what might shift elections, is increasingly difficult. Candidates and campaigns tailor their messaging to specific voter groups in ways that are often hard to track. The issues that motivate voters largely depend on curated news sources and social media algorithms.

These dynamics have fractured what was once a reasonably shared understanding of the issues at stake, to the point that it sometimes seems like voters walking into the same polling place are barely voting in the same election.

At the same time, journalists trying to explain voters’ motivations and election outcomes are forced to rely on data that can’t tell the full story. Pre-election horse race polls tell you who people say they’ll vote for but provide little depth on voters’ motivations and concerns. Election returns tell you who won the votes and where the votes came from, but not why. Academic surveys explore the “why,” but they’re typically not available for months after the election.

For a great many people, the result was simply unbelievable. It was unbelievable because the press and the polls—the mechanism by which we make elections make sense—got it wrong, and we got it wrong because we were relying on thin tools and not enough time spent listening to voters. We missed underlying signals in the public mood, especially the deep malaise and distrust among working-class White voters who were engaged and motivated by Donald Trump’s messaging.

The costs of that failure do not stop when the election is called. When nobody can say with confidence what motivated a result, any winner can claim a mandate for anything and govern accordingly. And in the absence of a real explanation, half the country is free to conclude that the other half has lost its mind.

A vote is not an opinion. It is a choice, often a complicated and begrudging one. The reasons behind it deserve more than a single crosstab.

NORC at the University of Chicago, a nonpartisan, not-for-profit research organization, is building a tool called the Civic Health & Elections Project that we hope will help address the problem. In the week before Election Day, we are surveying more than 85,000 registered voters, with at least 750 in every state. This breadth of coverage provides a thorough picture of the entire country, not just swing states and competitive districts. The core dataset, along with interactive dashboards and mapping tools, goes public within hours of the polls closing.

You can ask whether voters in your state who backed the winner also think the economy is improving, whether they trust the institutions that will carry out the agenda, what they expect from government on health care or immigration, and how any of that differs from the state next door. Because the sample is large, those questions still have answers when you narrow to a region or a subgroup.

The detail and timeliness of the project’s data let a reporter in a deep red state like Oklahoma or a deep blue state like Vermont move past the fact that a district went for a party by 15 points. Right now, that result arrives with an implied explanation attached, and most of us accept it. With these data, that reporter can examine segments of the state’s electorate, look at state-specific questions, and compare their state with ones that resemble it and ones that do not. The national press tends to treat such places as monoliths, or to skip them because the outcome was never in doubt. These data let us explore what those results actually mean.

The Civic Health & Elections Project is not designed to help you call a race. But with results available within hours of the polls closing, it can help you understand the why, in every state, while the result still matters.

Before the data arrive, write down two questions about the voters in your state that you have never been able to answer. The reporters who get the most out of the project will be the ones who knew what they were looking for before we release the data. The data will be available at 12:01 a.m. Eastern Time on Wednesday, November 4 at civichealth.norc.org. You can also watch the video tutorial for newsrooms that want help working with the dataset. You will not need a statistician or a data desk to find and tell the important stories.

Understanding the electorate is not a five-day assignment that ends when the last race is called. It is the precondition for holding politicians accountable to what the people in their state want. One story raises the questions that become the next one. That is how a country starts becoming comprehensible to itself again.