Terms like “need” and “better” are subjective, especially when it comes to memory techniques. What works well for one person may just not click for another. Blazingly fast speeds and accurate recall of thousands of numbers can be achieved with many different systems. You really have to try them and see what fits your goals and the time that you have available to learn and practice with a system.
A PAO will let you encode 6 digits in a single scene that’s made up of three elements that easily combine to form an efficient little mental animation. An advantage of PAO is that the system does nearly all the work in structuring what that scene will look like. If you learn to visualize each element, you can get to a place where you can combine them very quickly. The variety also cuts down on repetitiveness, so unless you run into a duplicate sequence of 6 digits, every scene will be different.
The tradeoff is that you’ll need to build and learn 300 unique elements for a PAO, which will require a larger time commitment up front. In my opinion, it’s worth that extra effort.
An object only system is serviceable, but something to consider is that you’ll be trying to visualize scenes using your list. If the only element in those scenes that encodes information is the object, there will be a lot of extra details that you’ll be adding that don’t carry any data.
Lets say you’ve got a MoP as your object for 17, a TiRe for 14, and a roll of TaPe for 19. If you want to create a scene that combines those three things in order, you might come up with “the mop is cleaning a tire, which then rolls away and gets stuck an a bunch of strips of tape that are in its way.”
Compare to a PAO structure: “MoBy is TeaRing TaPe.” This scene provides a clear subject and is naturally animated and can be easily visualized. It writes itself. There are no neutral actions or subjects. Every element encodes data.
You can always start with an object list, play around with it for a bit, then expand it to a Person-Object, then expand again later to a full Person-Action-Object.
With Person-Object, you can generate structured scenes but it also gives you some freedom to visualise actions and interactions that would more naturally connect those two elements. Again, you’ll have to pay around and see how you like those approaches.
One thing you’ll likely run into if you only have a single element learned for each number is the problem of sameness. If you get a string like 113245321177, you’ll notice that 11 appears twice and 32 appears twice. With just an object list, its very easy to confuse or swap places when trying to recall. You have to connect “ToaD MooN RoLL MooN ToaD CaKe.” (“I saw the ToaD and it did something with the MooN, but also did something with a CaKe… did the Moon go before the ToaD or after, or both? Which set went where? Argh!”) If you have a PAO structure, that same sequence is “TeD MooNing a RoLL” (11-32-45), then 'MaNNy TooTing on a CaKe" (32-11-77). Which would you rather try to visualize and recall?