It seemed hard to believe that the SNP could improve further on the dizzying heights that they'd reached in the previous two post-referendum Ipsos-Mori polls, and yet the logic was inescapable - the last Ipsos-Mori poll was way back in January, and every polling firm that has reported recently has detected a further additional swing from Labour to the SNP. So, although the figures you're about to see are truly extraordinary, they also have a touch of inevitability about them.
Scottish voting intentions for the May 2015 UK general election (Ipsos-Mori, 22nd-27th April) :
SNP 54% (+2)
Labour 20% (-4)
Conservatives 17% (+5)
Liberal Democrats 5% (+1)
Greens 2% (-2)
UKIP 1% (n/c)
Again, we have a continuance of the pattern of firms that use a 'real world' methodology reporting a bigger SNP lead than firms that rely on volunteer online polling panels. Ipsos-Mori collect their data by telephone, and are now showing a gap of 34 points, while the face-to-face pollster TNS showed a gap of 32 points in the poll they released on Monday. Every other firm has been using a volunteer online panel, and all of them are showing a smaller gap. The difference isn't huge in some cases, but we've yet to see the 'Great Convergence' that occurred just before the referendum. (Admittedly TNS have shown a much more pronounced trend than anyone else, but it would be hard to call that a 'convergence', because they've rocketed from the Labour-friendly end of the spectrum a few months ago to the SNP-friendly end now.)
In a way, this is counter-intuitive - you might expect a sudden transformation in the fortunes of two parties to be magnified in polling samples populated by the political junkies that make up a significant chunk of volunteer panels. Instead, the reverse seems to be happening, and online polls are lagging somewhat behind the tidal wave being detected by the 'real world' firms. This is also, incidentally, a complete reversal of the pattern we saw during most of the referendum campaign, when Yes did far better with online firms (with the notable exception of YouGov, courtesy of the Kellner Correction).
It's tempting to assume that 'real world' polling is closest to the truth, but annoyingly there are other methodological factors that set Ipsos-Mori and TNS apart, and those may be muddying the waters. Most importantly, neither firm weights by recalled referendum vote, which has become a new orthodoxy among online firms. In fact, Ipsos-Mori doesn't weight by any form of past vote. That might conceivably mean that enthusiastic SNP supporters are more likely to agree to be interviewed, and aren't being appropriately downweighted. (The counter-argument is that people who are enthusiastic about their choice are also more likely to turn out to vote.)
Sticking with that theme, the other factor that makes TNS and Ipsos-Mori a little different is that they both use a particularly extreme likelihood-to-vote filter for their headline results. That did have a big impact on the TNS poll earlier this week - a 27% SNP lead was transformed into a 32% lead after the filter was applied. Intriguingly, though, the same thing hasn't happened with Ipsos-Mori - the SNP were ahead by 34% even before the filter. So no alibi for Jim Murphy there.
STV, who commissioned the Ipsos-Mori poll, have projected that it would give the SNP a clean sweep of all 59 Scottish seats. In reality, I struggle to imagine the SNP taking Orkney & Shetland - the evidence from several recent elections is that the Northern Isles are still substantially insulated from the national trend. You never know, though - it's possible that the aftermath of the referendum may have flicked a switch there as well.
As the Ashcroft constituency polling has amply demonstrated, all of the other 58 seats are undoubtedly up for grabs. That's not to say the SNP will win all of them, but if they fall short in a few, it might not be by much.
By far the biggest oddity in this poll is the five-point increase in the Tory vote. Predictably, the right-wing press have leapt on that finding as if it's unquestionable truth and a stunning vindication of Ruth Davidson as leader, but I'm afraid there is precious little evidence of a Tory surge to be found anywhere else. The Survation poll on Monday had the Tories down 2%, TNS showed a static position, and Panelbase did show them gaining, but only by 2% (and therefore well within the margin of error). Until another firm replicates the big increase, it's probably safer to err on the side of thinking that it's a freakish result.
Over the last week or two, some of the more thoughtful unionist commentators have begun to recognise what has been obvious to the rest of us for quite some time - namely that Labour are suffering partly because of Jim Murphy's fabled "activeness", and not in spite of it. This poll will further assist the driving home of that painful truth, because Murphy's net satisfaction rating has slumped from -4 to -19. Unfortunately it's not possible to tell what his rating is among people who voted Labour in 2010, but we do know that among respondents currently planning to vote SNP, it's a catastrophic -53. Someone so hellbent on antagonising former Labour voters who have drifted to the SNP never had a prayer of making good on his early boast of "holding every seat". Murphy's days as leader may now be numbered (possibly even in single figures).
On the supplementary questions that seek to probe how voting intentions might yet change before polling day, there's the familiar mixed picture. Although a large number of people will vote both for and against the SNP on a tactical basis, that phenomenon looks like being a net negative for the party. But it's offset by the fact that people currently planning to vote SNP are much more certain that they won't change their minds - 86% of prospective SNP voters have already made a firm decision, compared to 76% of Tory voters, 66% of Labour voters, and just 37% (!) of the small group of Liberal Democrat voters.
The simplest and yet most important detail of the poll is the fieldwork dates - it was conducted between last Wednesday and this Monday. Many respondents will in effect not have been saying how they plan to vote, but how they have already voted by post. If the headline numbers are anything like accurate, the die may therefore already be cast in a large number of constituencies.