You should start with a screening experimental design (DOE) where you just look at extremes of each factor. You need to use your understanding of your system to come up with the appropriate levels. e.g, if you believe temperature, stirring and time are important, you'd look at a low and high temperature, slow and fast stirring and short and long duration. This would all be part of one study. With three factors, it's pretty straightforward. Beyond that, you need to be smarter about the experimental design.

https://www.isixsigma.com/dictionary/screening-doe/

Then you'd keep the factors that come out as significant and do a response surface design which can determine interactions (e.g., temperature is only important at slow speed etc). At this stage, you'd look at three (or more) levels for each factor. This is where statistical experimental designs are important in order to reduce the testing required.

(Search the same website as above).

If you do this correctly then you can get away with many fewer tests than you'd have to do in a traditional way.

Although you mention size and PDI, there may be other bits of information that come out of the Zetasizer that you can use.

Presumably, you have an overall end goal in mind (entrapment? release kinetics?) Size and PDI may end up being irrelevant. i.e, you may find your factors are important but that the Zetasizer doesn't reveal anything. In that case you need to consider other responses. Look into this before you start. It's easy to run different tests at the same time if you plan ahead. If not, you'd end up having to repeat the entire experimental designs.

The above is based on my experiences where I had to find very efficient ways to study complex problems. If you only have a few factors, you don't have to go overboard.

If you have access to a statistician (or SixSigma experts), talk with them but make sure you emphasize the physical importance of which factors to focus on otherwise you may end up with a generic design that lets you down.

You should start with a screening experimental design (DOE) where you just look at extremes of each factor. You need to use your understanding of your system to come up with the appropriate levels. e.g, if you believe temperature, stirring and time are important, you'd look at a low and high temperature, slow and fast stirring and short and long duration. This would all be part of one study. With three factors, it's pretty straightforward. Beyond that, you need to be smarter about the experimental design.

https://www.isixsigma.com/dictionary/screening-doe/

Then you'd keep the factors that come out as significant and do a response surface design which can determine interactions (e.g., temperature is only important at slow speed etc). At this stage, you'd look at three (or more) levels for each factor. This is where statistical experimental designs are important in order to reduce the testing required.

(Search the same website as above).

If you do this correctly then you can get away with many fewer tests than you'd have to do in a traditional way.

Although you mention size and PDI, there may be other bits of information that come out of the Zetasizer that you can use.

Presumably, you have an overall end goal in mind (entrapment? release kinetics?) Size and PDI may end up being irrelevant. i.e, you may find your factors are important but that the Zetasizer doesn't reveal anything. In that case you need to consider other responses. Look into this before you start. It's easy to run different tests at the same time if you plan ahead. If not, you'd end up having to repeat the entire experimental designs.

The above is based on my experiences where I had to find very efficient ways to study complex problems. If you only have a few factors, you don't have to go overboard.

If you have access to a statistician (or SixSigma experts), talk with them but make sure you emphasize the physical importance of which factors to focus on otherwise you may end up with a generic design that lets you down.

As John suggested to you, you must perform a screening with design of experiments (DOE). I recommend to you a 2^k design if you have three factors (k=3), and a 2^(k-1) if you have more than three factors (k>3). With a 2^k DOE and results ANOVA you could calculate which factors and factors interactions have significant effect on your three response variables. With a 2^(k-1) for k=4 you could calculate which factor and some two-order interactions on factors have significant effect on responses. If you have k=4 and enough resources you can use 2^4 DOE (16 treatments with duplicate at least).

Paired t-tests /ANOVA