After the initial surprise and the head-scratching, pollsters are trying to sort out what went wrong in the way they handled the Alberta election.
Several said there was little they could have done to foretell Monday night's victory by Alison Redford's Progressive Conservatives over Danielle Smith's Wildrose Party.
There was, they said, a late swing in voting intentions that increased after the last survey was taken.
But there were also arguments that something more profound had been missed and that pollsters and media were too focused on horse-race numbers and failed to grasp the true dynamic of the Albertan electorate.
"There's been much discussion about how the polls could be so wrong," said Ian Large, vice-president, Alberta, for Léger Marketing.
"My first reaction was, `Oh my, maybe something went wrong', ," said David Coletto, chief executive of Abacus Data Inc., which conducted automated-call polling.
It was a late shift, nothing could be done
Mr. Large said Wildrose ran an effective campaign until late controversies -- comments on gays and race relations by two candidates, Allan Hunsperger and Ron Leech, and Ms. Smith's unwillingness to condemn them -- reinforced voters' fears about the party's social agenda.
"None of the polls were conducted late enough to capture the impact of those comments," Mr. Large said.
The other challenge, he noted, was that voting took place Monday and that the key shift unfolded during the weekend, when it is harder for pollsters to reach respondents at home.
Like last-minute shoppers, many Albertans made their voting decision during the last weekend, Mr. Large and Mr. Coletto similarly argued, both separately reaching for the image of voters making up their minds after hashing it out at family gatherings.
"The voters were moving so quickly . . . We finished our last poll Thursday night, four days before the election. We were seeing already movement away from Wildrose toward the Conservatives and that really accelerated in the last 24 hours," Mr. Coletto said.
In that survey, Abacus reported that Wildrose for the time had lost ground thought the upstart party still had a 10-point edge among decided respondents.
The last public poll, by Forum Research was an automated-call survey conducted Sunday. It found the gap between the two parties narrowing, with Wildrose at 38 per cent and the PCs at 36 per cent.
"I think in the last day of the campaign, maybe right up until election day, Wildrose cratered. We did get part of that [trend] but it continued," said Lorne Bozinoff, president of Forum Research.
There was a segment of voters who left the incumbent Conservatives for Wildrose and then were turned off by Ms. Smith's party in the last week, he said. "They couldn't go anywhere but to the Tories."
Albertans: More complicated than you think
Still, the PCs didn't just come from behind to win narrowly but by a nine-point margin.
"People don't shift that dramatically on the last day," said Brian Singh, president of ZINC Research, and a strategist for Naheed Nenshi's mayoralty campaign in Calgary.
"The pollsters were using some techniques that didn't provide the complexity and insight that they need to understand where the rhythm of the campaign was exactly going."
The morning of the vote, PC campaign manager Susan Elliott sent an encouraging e-mail to party members, telling them there had been a "seismic shift" in the last six days.
"Believe me, this race is far from decided," she said, revealing that internal polling gave them a four-point lead.
Mr. Singh said Wildrose has a 33 to 35 per cent stable block of hard-core voters, so the party's final tally (34.3 per cent, against the PCs' 43.9 per cent) underlined how it wasn't able to make itself more palatable to voters who didn't identify with its social agenda.
"The notion of what Wildrose brought to the table . . . wasn't properly explored," said Mr. Singh, whose firm didn't take polls during the campaign.
"If you look at what was going on with Wildrose, they may have had tremendous volume but their supporters were shallow in depth."
Another analyst was more blunt.
"We're going to be trying to figure out what happened over the coming days," Calgary polling consultant Janet Brown said. "Were the polls simply wrong? Or was there a big shift among voters?"
While there have been instances of polls getting things wrong – say, predicting a majority instead of a minority or vice-versa – she couldn't think of one where the result had been so different from the consensus among pollsters.
Those shy Albertans
Another factor was the likelihood that some poll respondents weren't forthright.
Ms. Redford's party was initially rattled by the no-meet-committee scandal, where MLAs got paid for being on a legislature panel that hasn't sat since 2008.
Some supporters might have decided to keep their views to themselves when contacted by pollsters, Mr. Singh.
"Early in the campaign, it was socially undesirable to support the PCs . . . so people were telling us they intended to vote Wildrose, that they intended to vote for change and then at the last minute voting intentions changed very quickly," Mr. Coletto said.
"So there may be some of that shy PC voter out there who doesn't want to admit it."
Mr. Large said future campaign polling might rely more on online panels, to circumvent the fact that fewer people use residential landlines.
Both Mr. Coletto and Mr. Bozinoff however said they didn't see a need to change methods.
"When six or seven polls all say the same, using different methodologies, one conclusion is that we were all wrong, the other conclusion is that we were measuring accurately what was happening at the time, and in a short period of time the entire environment shifted," Mr. Coletto said.
With files from Adrian Morrow and Dawn Walton