Adsense -> Imagine Google as the only master bidder. And sometimes third party bidders purchase inventory through adsense where Google takes a cut.

Header Bidding -> Incorporate third party bidders directly into the flow at initial page load. Bidder pays you directly where Google does not take a cut. This also means that, without other bidders aka third party network partnership. You are "header bidding" or auction with just adsense which is rather pointless.

Using Third party -> The third party may take a cut at higher or lower than Google directly. Resulting in less earning sometimes than direct partnership.

Header Bidding is not Adx. I see it more like auction flow, it's a standalone ad auction technique. Instead of waterfall cascade / daisy chain passbacks among multiple networks. The ad unit is passed and auctioned horizontally. This makes it superior than the outdated daisy chain technique.

Header bidding is an evolved form of daisy chain, which needs to be built on some form of ad server ie. DFP. Thus working knowledge / setup for DFP (for it being free initially) is generally required to start attempting header bidding.

Starting out, I would learn to optimize DFP with proper ad unit optimization and dynamic allocation, plus third party networks to learn about revenue and CPM optimization. Once you are able to get positive return on that front, header bidding becomes to next step to further optimize earning. Jumping straight from adsense units to header bidding can be quite hard, it's like two generation tech jump.

In your case, I would split the pageviews between third party and adsense, with dynamic allocation with CPM floors and frequency cap, and perhaps geo targetting using DFP. Learn through the process, then with that experience to further into header bidding.

In short, if adsense is the only monetization network, there's nothing to "DFP" or adserver, and thus there's no need to "header bid", there's no auction as Google adsense is the only winner in every auction. Same with DFP almost. If adsense is the only monetization channel, a straight tag is generally almost better, you avoid impression slippage through ad server calls.

And if you are already using DFP, well, there are a few big companies now making custom and easy to install "wrappers" that can be tested quite easily. It's about a couple hours of work to setup and test. It doesn't hurt to try and see how it works.

Header bidding brings good results, but isn't without drawbacks. The biggest is the work involved. That work brings good gains if you have enough inventory, but if you haven't got a reasonable level of inventory then your time might be better spend in other areas. Choices are really to either put it together yourself using prebid (interesting learning, but not quick job - especially in managing all the partners), work with a prebuilt wrapper (put together for you, usually by a demand partner who will introduce their demand at an advantage) or work with a third party who do the whole thing and take a slice.

I work for a GCPP who do the last option. We have deals with the best partners in place and handle all the technical and management, but take a slice. Publishers still do very well in terms of gains, but the economy of that model means that you need to be working with a few million pageviews a month to make it good deal for everyone.

With either the DIY approach or the prebuilt wrapper option do factor in management time. As well as managing the various demand partner relationships, you need to be monitoring and adjusting performance and checking that they pay what they bid.

I did #2 - work with partners who have pre-built wrapper and it was a breeze. They did pretty much everything, all I had to do was upload a couple lines of code into my header. The other great thing about some of the wrappers these days is that they will allow you to throw in other third party partners into their own wrapper so you can multiple bidders through one network's wrapper.

that's good if they let you add additional partners (and they are being given fair look). I don't generally see a huge benefit in a single partner, but testing a few and settling on anywhere 4-7 top performers seems to be the sweet spot in most cases.