I am absolutely NOT an expert in this, but I can tell you our situation. First, we have concrete, too, but not on grade (3rd floor apartment). We also used heating mats, choosing Warmly Yours. (This decision involved the claims of electromagnetic fields reduction by someone else - I can't remember which one now - and coil distance and wattage per square foot.) I read a lot about trying to create a "thermal break" between the heating mat and our 12 inches of concrete. In the end, after reading a lot of posts all over the internet (reader beware, there is some pretty sketchy non-science out there), especially johnbridge and I think jcl, and after extensive discussions with my nephew who is a materials engineer, we did two things: 1/4" Wedi board and SuperTherm. Here's how we got to that decision (YMMV): So very many people told us that we didn't need any insulation, that insulation was stupid and unnecessary, that the heating mat would be absolutely fine without it. The two things we heard the most were: 1.) "I've installed 12 trillion-gazillion mats (or whatever extraordinary number they thought I would buy into), and I've never used it because you don't need it." 2.) "Heat rises." Regarding point #1, the thermal properties of wood underlayment versus concrete cannot be understated. What works over wood will be different over heat-sucking concrete. Even if you ask if all their other jobs were over wood, they very very VERY frequently lie. (Oh yeah, and I don't care what cocked-up thing you've inflicted upon other poor unsuspecting, non-researching saps....) Wood acts more as an insulator, concrete as a conductor of heat. Regarding point #2, lord this is stupid. Hot AIR rises, because it has more energy which makes the air molecules vibrate more, pushing the molecules apart, which takes up more room (for the same number of particles), which makes them less dense...so they rise above the cooler air. This is for AIR. When dealing with solids (and some liquids under some circumstances), heat acts more like electricity (both are forms of energy), and takes the path of least resistance. Meaning, that the heat from your mat will travel EITHER UP OR DOWN according to whether the concrete or your tile is more resistant to heat energy. The science behind which material involves density, heat capacity, etc. (It isn't exactly like electricity though, in that it isn't an all-or-nothing path decision. Apparently, with heat and different materials, it is a matter of what percentage - I'm over simplifying here - goes in which direction. Think about it: electrical conducting shields work absolutely, no matter how long the wire is live, but heat shielding can only slow down the transfer.) For anyone who owns a frypan with a handle that is made of a different metal than the pan body itself, this concept is in that pan. The body of the frypan is designed to transfer heat fast, the handle is made from a different metal that will absorb heat more slowly. The point: the handle WILL get hot eventually, just a matter of how much time. My nephew (a materials engineer, who doesn't work in this field but had more game than me for running the equations) figured out that my tile is waaaayyy more resistant than the concrete, that 70% of my heat would go into the concrete and only (about) 30% into the tile. All the people that said bare concrete would be "fine" were kind of technically sort of right: my tile would eventually absorb some amount of heat. I had two issues with this: how much more energy would I be paying for to heat the 12 inches of concrete AND how sloooowww would my tile heat up without insulation? So no contest, we needed a thermal break. (I wanted to be able to get my tiles warm in - at most - a couple hours. Without insulation, the mats would have to turn on about 12 hours ahead of time, just to hit 80 degrees by morning. No way.) I found a thread on johnbridge where somebody (I can't remember his screen name but I remember his real name is Eric, if that helps) actually mocked up several underlayment materials and measured how long it took for the top of the tile to move something like 20 degrees. (This was a while ago so some of these numbers might be a bit off, but the gist is the same.) So a good insulator would allow the tile to heat up faster because more heat is going up rather than down. He tested a couple products, and the clear, hands-down winner, coming in 40% less time than any of the others, was Wedi. (Interestingly, there was very little difference between the 1/4" and the 1/2". The guy that ran the test didn't understand why there wasn't a significant difference either, but we're talking about tenths-of-a-point level of small difference.) We went with 1/4" Wedi. The other thing we did - which was probably overkill - was because there are three different kinds of heat, and my nephew wanted to experiment with a ceramic (anti-reflective, anti-radiant) paint substance he read about. I understand the theory (waaayyy too involved to get into here), but the proof of this substance (SuperTherm) is not anywhere near substantiated yet. And dealing with the company sucked pretty badly (very unprofessional and difficult). But it is on half of my bath floor (we did a bunch of test to make sure it would adhere sufficiently to not screw up the tile) so I guess we'll see. The bath hasn't gone live yet and probably won't for another month. Don't let anyone tell you that you don't need it. Yes, the mat will work without insulation, but how well? How fast? (Yes, we are on a programmable timer. This was a big arguable point for the you-don't-need-no-stinkin'-insulation crowd, that the timed thermostat would cure all. That's fine as long as there are no exceptions to the original timer programming. But let me tell ya', my husband is ALL about exceptions. I swear he stays up late at night, trying to think up new ways that our routine won't suffice. He's adorable, but I need flexibility - in both lunch menus and floor temp.) I know this post is long, but your question involves some actual science (chem and physics), and knowledge is power, no matter what path you choose. And the amount of mis-information out there is staggering. And the surety with which people will tell you bad information is daunting. (And arrogant. And ignorant.)

you need to rent a compactor and go over the entire area to make sure all the gravel is compacted and nice and flat

The driver on basement floor is often the loss of headspace from putting insulation on the floor. Various type of insulation boards are often used since they have the highest R-value per inch allowing for smaller thicknesses. Wood sleepers are then placed about every 16 inches and a layer of plywood applied.

True. There will be heat lost to the ground. Nonetheless, do not conclude there is no point at all in having a layer of insulation: any break is better than no break at all. Think of wearing one layer of light clothing as protection against the cold: better than bare skin. Any small R (e.g. R2, R3) still directs more heat upwards than not. To place radiant heat directly on the slab means more and more heat is lost as floor temperature rises, not just 50% but a far greater fraction. "thin" , "not very" , "useful" -- terms undefined. A separation layer is better than none. Not to overthink this, just to drum it in. "Something" is more than nothing, and this is the significant point, not whether the R factor appears low or high. Once your floor temperature rises, you want to prevent sending most of it into the ground, which is what happens when there is no insulation / isolation / separation at all. Note that the definition of R rating testing does not allow one to distinguish between heat transfer occurring by conduction or by radiation. In many circumstances this is not critical. If you want to get into this level of detail, it will take a few dense paragraphs. Suffice to say that R ratings will be misleading and inaccurate for this application so do not rely on the apparent differences between R numbers. They are all small and that is OK. They are almost meaningless in this situation of a slab-floor separator with a thickness of from 2 to 16 mm. It is obvious that with electric cables you have more room for a thicker insulation layer. HTH -david