He won. She lost.

That sinking feeling that my industry was due for a thrashing hit me in the gut during the early morning hours of November 9 when it became clear that Donald Trump would defy most polls and become the 45th President of the United States.

My fellow pollsters weren't the most popular bunch as it was. We were heavily criticized for failing to predict the results of the United Kingdom's vote to leave the European Union. Our reputations were dinged after the 2014 midterms when Republicans beat expectations and won the US Senate.

Trump's victory would make those stumbles seem relatively small.

I knew what would come next: Questions. So. Many. Questions.

The big one -- how did nobody see this coming? -- persists eight months after voters went to the polls. There’s a strong chance that question, or some variation of it, will come up over your summer vacation. After all, 56% of you said you talk politics with friends or family “very often” in a March CNN/ORC poll.

This election will be studied and debated for years to come. But as I gathered with colleagues recently in New Orleans for the American Association for Public Opinion Research's annual conference, some things became clearer to me.

For one, the trouble with the 2016 election wasn’t about the polls. It was about how we used them to tell the story of this campaign, and how we can do it better.

Just in time for your summer getaway, here's my take on the election that stunned the world.

Nothing I’ve seen since the election illuminated the story as clearly as a presentation from University of Michigan political science professor Mike Traugott.

Using data collected by Gallup in daily tracking surveys conducted from just before the party conventions in July through Election Day, Traugott and a team from Gallup, the University of Michigan and Georgetown University visualized nearly everything you needed to know about the election in two word clouds.

Each day, Gallup’s researchers would ask whether, in the last day or two, people had heard, read or seen anything about Hillary Clinton, and separately, whether they’d heard anything about Trump. If the person being interviewed said they had, the interviewer asked them what they recalled about the candidate.

For Clinton, “email” pops out of the word cloud, the most frequently mentioned word or topic by far. For Trump, there is no dominant word or theme.

Breaking it down by week, the dominance of the email scandal throughout the Clinton campaign is jaw-dropping. In the final weeks of the campaign, literally every word that registers with enough people to be visualized is about it: Email, FBI, investigation, foundation, reopening, scandal.

Further, when these words are connected to favorability ratings for Clinton and Trump (the only measures of attitudes toward the candidates that were tracked in Gallup’s polling), it becomes clear that Americans’ recall of stories related to the email scandal and doubts about Clinton’s honesty were most impactful of all.

If you strip away the names and all the context that came from the incessant focus on horserace numbers and look at the results as what Americans heard from two presidential campaigns, one thing is clear: Clinton's message never had a shot.

The few recalled words closely tied to warm feelings of Clinton were largely about the process of becoming president rather than about being one: Campaign, debate and speech. Hardly the stuff a presidency is built on.

Clinton’s lack of a resonant positive message meant she never gave voters who didn’t like her or Trump a reason to choose her -- at least not one they heard.

She offered plenty of reasons not to vote for Trump. Several seemed to stick -- “woman” and “sexual” were the two words most closely correlated to negative views about Trump. But she failed to provide meaningful pushback to the negatives that were driving Americans’ views of her.

And despite all the disarray around Trump’s campaign, Republicans did make a successful case for their candidate. The words most closely connected to positive impressions of Trump were the words he wanted voters to remember: Economy, Jobs, Make America Great Again.

So how come we didn’t detect Trump’s impending victory?

First, it’s hard to see in real time. These Gallup findings were being released during the campaign, but much of the interpretation focused on the amount Americans had heard about each candidate rather than the substance. More often than not, that result was tilted in Trump’s favor, leading to more stories about Trump’s earned media and fewer about what people were actually hearing.

That conclusion seemed borne more of expectations than an objective read of the data. The poll result wasn't about news coverage, but what people were hearing from any source. That included news media, but also Facebook feeds and conversations with co-workers and around the dinner table at home.

Second, election analysts and poll interpreters made several assumptions that didn’t hold up:

--That the difference in perceived qualification for the job would outweigh ratings of honesty.

--That voter turnout patterns would yield an electorate that reflected the increasing diversity of American society.

--That the Democratic advantage in the “blue wall” states – those that had voted Democratic in every presidential election since 1992 – would probably hold.

--That those voters who had unfavorable opinions of both candidates would kind of behave like your average undecided voter.

None of it was right.

The Gallup data point to an explanation for one of the most interesting findings in this year's exit poll, and could be the counterbalance we needed for at least one – and possibly more -- of those critical assumptions.

The nearly one in five voters who held unfavorable views of both Trump and Clinton broke overwhelmingly in Trump’s favor, much more so than those who said they made up their minds at the end of the campaign. These “disgusted” voters chose Trump over Clinton by 17 percentage points nationally, and were likely decisive in critical states such as Michigan, Pennsylvania and Wisconsin.

Simply put, the data suggest Trump gave disgusted voters a reason to choose him, and Clinton never crossed that threshold.

“If one thinks of a campaign as a learning experience for voters about what candidates stand for and what’s likely to happen if someone’s elected, that wasn’t there,” Traugott told me. “A significant portion of voters was willing to take a gamble on a relative unknown just because of fatigue and dissatisfaction with the status quo.”

That fatigue surely left some of these voters out of the electorate entirely, even if they were counted in our pre-election polls. The Census Bureau’s Current Population Survey collects data on whether people voted each election year, and if they didn’t, asks why. In 2016, 25% of registered voters who didn’t cast a ballot said it was because they didn’t like the candidates or the issues. That’s nearly double the share who said the same in 2012, and about three times the level in 2000.

We don’t know whether the disgusted voters who chose to stay home would’ve balanced out the ones who did show up and broke in Trump’s favor, but pre-election polling suggests they might have. Throughout much of the campaign, CNN/ORC polling found these voters breaking in Clinton’s favor by a small margin.

It’s easy to focus on the horserace when you get a poll result. But by allowing those findings to merge with expectations without gut-checking them against the type of startling findings that defined the 2016 campaign from the beginning, we, and everyone following the news, missed the real story.

Traugott, who literally wrote the “Voter’s Guide to Election Polls,” said what often gets lost in poll analysis is “the ‘why’ part that polls are especially good for, because that coverage gets reduced by emphasis on the horse race.”

Pollsters and pundits were handed the most interesting campaign of our lifetimes. Instead of allowing the data and the electorate to speak to us, we made it fit into the mold we've used for every campaign since the dawn of modern polling.

Researchers who don’t have a pre-fit mold for every project solve this problem by using qualitative research techniques -- focus groups, in-depth interviews, and cognitive interviewing. This allows the people being studied to speak in their own voices and what emerges from those conversations can become the basis of new lines of inquiry rather than forcing public opinion to fit into the questions we’ve already written.

Gallup's project goes a long way toward highlighting the value of this type of research in political polling.

Some might argue that this election was too unique to make it worthwhile for the whole universe of political analysts to change their methods. But it doesn't take a unique set of candidates to make for a confusing narrative.

Take the special election just a few weeks ago in Georgia’s 6th congressional district. Almost all of the handful of polls conducted there focused on the horserace alone, with few supporting questions to explain why voters were backing one candidate over the other. Without that context, the story of the election is told by the people running it and covering it, not by the people casting the ballots.

Was Karen Handel's win a referendum on Nancy Pelosi? Did Jon Ossoff outpace past Democrats because it was really a referendum on Trump?

We'll never know, because no one bothered to ask.

Jennifer Agiesta is the director of polling and election analytics at CNN. She produces all of the network's polling and leads its Election Night decision team, while guiding reporting on polls. She was previously the director of polling at the Associated Press. She has worked on election polling since 2000.