There was a rumour doing the rounds on Twitter last night that the No campaign recently commissioned a poll from Ipsos-Mori, and have kept the results to themselves. I've no idea where the rumour is coming from, but it did remind me of knowing comments that Channel 4's Gary Gibbon kept making about the internal polling for McDougall Central -
"You sense some exasperation at the top of the 'no' campaign and a bit of bafflement that the average of published polls is narrowing quite a bit more than their own private polling with bigger surveys."
You 'sense' here that Gary is 'sensing' exactly what he was supposed to 'sense', so I would take the reference to bafflement with a heavy dose of salt. However, if the No campaign are at least partly conducting this large-scale private polling via Ipsos-Mori, that would tally up with the rumour and the rest of what we know. I must say I've been very surprised by how easy it's been to find people who have been asked the referendum question by Ipsos-Mori. If the published polls were the only ones to have taken place, only a miniscule percentage of the electorate ought to have been contacted so far. But if Better Together have been commissioning telephone polls with a much bigger sample size, then there could be tens of thousands of interviewees out there.
By the way, don't be fooled into thinking that the private polling must be much more accurate due to the higher sample sizes. The standard margin of error for a published Ipsos-Mori telephone poll is 3%, and even if the internal polls had sample sizes of 10,000, the margin of error would still be 1%. So the difference is not huge. In any case, margin of error is a theoretical concept that assumes the methodology is absolutely correct, which to put it mildly is always open to question in this campaign.
There was a separate rumour a few weeks ago that an anti-independence trade union commissioned an Ipsos-Mori poll and withheld the results after it showed a Yes lead. I would give absolutely zero credence to the latter detail UNLESS it was an online poll rather than a telephone poll. It's perfectly conceivable that Ipsos-Mori's online panel would produce completely different results, but as a telephone pollster the firm is by some distance the most extreme No-friendly outlier, and without an overhaul of their methodology it's highly unlikely that they would be showing anything even close to a Yes lead.