It's just a percentage in each precinct. This isn't the most useful, though I may be able to reverse engineer some results with some effort.
The precinct results are the percentages of delegates to the county convention. You will notice that the percentages are the decimal equivalent of small integer ratios (eg 2/7 equal 28.6%). One problem is that a lot of precincts are 50%:50%, which means that they could be 1:1, 2:2, 3:3, ... etc. Somewhere there must be the number of delegates per precinct, but I haven't come across it.
The allocation is proportional to the Obama+Hatch results in the 2012 and 2014 elections. So you should be able to figure it out (the Iowa SOS does have precinct level results).
But the popular vote is much higher. For example, the Des Moines precinct that was shown on NPR had a popular vote count of 232:224, with 3 persons missing. But there were only 9 delegates awarded (5:4). So the precinct will be reported as 55.6% for Clinton, even though the popular vote was 50.9% Clinton.
2008 ratio of caucus attendees to state delegateState delegates per county - 2016Democratic resultsIt appears that the ratio of the Obama+Hatch popular vote to county delegates is not constant across counties.
For example:
Taylor 1789 Obama+Hatch :: 30 county delegates
Ringgold 1738 Obama+Hatch :: 80 county delegates
Decatur 2559 Obama+Hatch :: 70 county delegates
But for precincts within a county, the ratio of Obama+Hatch :: county delegates is a constant.
So for example in Taylor the quota is 1789/30 = 59.63
And by precinct:
Blockton 92/Q = 2 (rounded), Clinton 2, Sanders 0
Bedford 403/Q = 7, Clinton 5, Sanders 2
Clearfield 161/Q = 3, Clinton 3, Sanders 0
Gravity 127/Q = 2, Clinton 1, Sanders 1
Legion 183/Q = 3, Clinton 2, Sanders 1
Lenox 649/Q = 11, Clinton 7, Sanders 4
New Market 174/Q = 3, Clinton 2, Sanders 1
Total: 1789/Q = 30, Clinton 21, Sanders 9
Clinton is shown as receiving 70% of the vote in Taylor. That may be roughly correct, subject to rounding in each precinct, and assuming attendance per delegate was constant across precincts.
Taylor County will elect 2 delegates to the state convention. I think a 21:9 split will come out 1.4 : 0.6, and Sanders will get the favorable rounding, so that in the state totals, Taylor will be reported as 1 each.
A similar occurrence happens in Ringgold where a 53:27 Clinton majority at the county convention will produce a 1:1 split at the state convention.
But in Decatur. a 35:33:1:1 split may produce a 2:1 split at the state convention since Decatur has 3 state delegates (O'Malley and Uncommitted each have one county delegate). There could be some serious courting of those two delegates.
Since the highest number of attendees per state delegate has usually been in Johnson County (Iowa City and University of Iowa), I suspect that Sanders actually had a popular vote plurality. Likely not a majority since O'Malley was probably getting squeezed everywhere.
Because the size of the county convention varies, and is not proportional to the Obama+Hatch popular vote, it would be meaningless to produce statewide totals. I suspect that county conventions in larger counties are smaller in a relative sense. The Obama+Hatch popular vote in Polk was 116.44 x that in Ringgold. If the size of the county conventions was proportional, then Ringgold would have 80 county delegates, and Polk would 9300 county delegates.
Note that Polk does have 228 state delegates which is 114 x that of Ringgold.