After previous surprises (which some evidence suggests may have been election irregularities when you note that HRC won MASSACHUSETTS by 2 points but in hand counted paper ballot precincts Sanders won by 17 percentage points), the polls played with the all important assumptions of who is a likely voter in order to come up with their overall polling numbers. Sanders underperformed in MA and SC so they would reduce the weight of Sanders voting demographics like young folks and independents. That causes the Sanders forces to be underestimated greatly in Michigan. Overconfidence on HRC's part causes them not to pull out all the stops, and no big gaps between exit polls and election results are seen in MI like they were in SC and MA. (Nobody says Sanders won SC but the lead appears to be padded by 20 points in HRC direction, and look and behold polls were off 20 points in the other direction in MI)

Now internal polling for campaign is naturally combined with a sophisticated understandING of who will get out to vote (who the actual voters are going to be). Because campaigns have better info on likely voters than polling organizations (though their models can some times be self-deluding) internal models can be better than polls. Also the campaigns probaby didn't tinker with the likely voter assumptions like the professional pollsters did.

On edit: I'm sure you note that I disagree with your premises partially. Exit polls are traditionally best and we're so far off in just a handful of places. What gaps between exits and results mean is that it is certain that EITHER the election results OR the exits are wrong, OR BOTH. an investigation should proceed. But instead of following scientific method, lame media simply assumes it is only the exits that are wrong.

they only count as likely voters those who actually voted four years ago. And while young people are notorious for not voting in great numbers, the assumption that if you didn't vote in 2012 you won't vote in 2016 is incredibly flawed. Plus, if I recall correctly, a lot of young people who voted for Obama in 2008 didn't bother to vote in 2012, and a lot of them are showing up this time around. Again, some significant percentage of people who will show up and vote are being ignored.

It's my impression that "internal polling" is intended to do one of two things: either bolster that candidate's confidence, by showing results in his or her favor, or to give a truly accurate picture of what's going on out there. Which may not be favorable to the candidate. I know that on this site there has been a lot of speculation that Hillary's internal polling has shown that she's not doing as well as a lot of others, and that explains her increasingly desperate behavior. Since I'm not a party to what goes on in her campaign, I can't begin to offer an opinion here.