It is difficult to give the exact answer without knowing the type of experiments and variables. However you can start categorizing and prioritizing the variables and tests, respectively. At a point you might be able to understand to reduce the number of tests.

I am making a sensor .... various variables affect the sensitivity of the sensor which I believe they should all be included, isnt that right ?

I always use Plackett-Burman design to screen most effective variables

You should be able to reduce the number of treatments substantially. This would require a screening design. The type of screening design can be matched to fit your objective(s). For example, the Plackett-Burman was mentioned - these are used when one is confident that there are no interactions between any of the factors. Because of this property it is a good choice for "ruggedness" testing.

Could you provide more information on the number of factors that you would like to test?

All fractional factorial designs (FFDs) can be used for screening based on our purpose/requirement. To my knowledge, Plackett-Burman and Taguchi methods are effective screening designs. Both have its own merits and demerits. But with FFDs, you will be able to able to screen more number of variables with minimum trials. Also, you didn't mention about the levels of factors.

in order to decrease the number of experiments you can use fractional DOE that is useful for decreasing the number of experiments.if you are interested I can help you.

This is my email address : [email protected]

You can use "DX9" and its tutorial.

You can use Taguchi design

Try to use 2^k and 2^(k-p) designs already exist in DOE softwares such as design expert and minitab. This helps you to find significant linear effects. If the statistics are not reported due to small degrees of freedom, try to use normal probability plots (again exist in softwares).

If you have historical experimental data, you can check the effects without conducting any new experiments by applying generalized linear models.

let me know if you have more information about your experiments. In some real cases, expert judgments can lead to more focused analysis or merging factors.

According to Pareto principle some factors will be more important than other. You need to perform an educated guess. For some guidance of this problem I strongly recommend the following article: http://www.sciencedirect.com/science/article/pii/016974399280096M

If you don't want to reduce the number of variables you can make reductions by reducing the number of levels each variable is measured. In addition you can use various design strategies. Using an experimental design strategy like fractional factorials will require some assumptions about how neglible you believe the interactions between factors will be. The other thing to remember in doing so is to aim for all levels on a common base like all at 3 or all at 2s and 4s. This makes it more likely to find a main effects plan with a smaller number of runs.

As mention by all our friends here, basically you first need to conduct 2 level factorial design to screen out significant factors from all the variables..then you can conduct full factorial or fractional factorial with more level on the reduced number of variables.

How many factors do you have? What kind and number of levels do you have for each factor?

I really like something called an Optimal Factorial/Response Surface design. If you have say 8 factors to test, and each factor is at 3+levels, and they are continuous, an Optimal Response Surface will work well. You will want to test it for main effects at first to screen out some factors and levels. You will want to add in some Lack of Fit Points and a few replicate points. Optimal designs minimize the number runs you need to test all the factors you want to test.

I suggest you to used super saturated designs if you are sure that some of your factors dont have effect on the response. For more details you can see the wu and hamada 2001.

As mention by Abd Rashid Amirul, basically you first need to conduct 2 level factorial factorial design to screen out significant factors from all the variables with main effects only..then you can conduct full factorial or RSM with scren out variables that have impact on response variable.

First of all,  you must decide whether the factors are dependent or independent. In your case it seems that they are independent. Generally, screening design can be suggested, however, keep in mind that screening design requires some assumptions about importance of the factors. In my opinion based on experiences with experimental designs, it is better to decrease the quantity of the factors. If it is impossible, try to combine two factors whose effects on responses are very similar. Always try to eliminate or combine the factor without interaction between others. 

I agree with the suggestions from Yahya kaya. In my experience, I always review papers concerned with my experiments and then I do some one-factor experiments to find those factors that really influence my experimental results. Finally, screen the factors that I found to selects most effective factors. Plackett-Burman and Taguchi orthogonal analysis were suggested by me.

you are studying 7 factors using full factorial design. Firts try to find out all factors should be independent varibale. Second you can do first screening with main effects only. From screening results findout which factors have significant impact on your final result and with only these screened factors you can do full factorial or RSM. your number of experiments will definitely reduce. If all factors are important and independent, fractional factorial design is option to reduce number of experiments. with n-1 fractional factorial design your experiment number will be reduce to 64.

It seems that you run a full factorial design of experiments. i suggesst to used a fractional factorial design with 64 or 32 run.

If you can do it through design of experiments 2k fractional

You can do a split with the appropriate number of replicas to avoid such a large experiment and have the required power experiment. Choose the appropriate fraction to reduce the number of combinations tested (1/4 suggest fraction with 2 replications).

It is an important topic, I would like to learn a little more about the "too many factors" problem.

Any published references?

If you don't use regression analysis, you can use Taguchi orthogonal analysis or fractional factorial design.

The way is fractional factorial designs

There are a lot of way to reduce the number of experiments. You can use the fractional factorial design or super saturated design or many ohther kind of desisgns. You should select the proper design based on the goal of experiments.

You can refer to the following link :

http://www.itl.nist.gov/div898/handbook/pri/section3/pri33.htm

Basically, a screening test should be conducted; and then you can use Fractional Factorial or Taguchi design to reduce the number of experiments (Design expert software may be helpful for you).

In my opinion full factorial is the most accurate one.

May I guess that you do some experiments in physiscs? Then the standard approach to reduce the number of physical quantities in the relation which you are seeking and thus the number of experiments is "dimensional analysis" to obtain dimensionless terms. Often the physical relations can be "guessed". From Wikipedia: "Dimensional analysis is used to express the system with as few independent variables and as many dimensionless parameters as possible." Only thereafter I would consider any statistical method.

For what it is worth.

I agree with the above authors, try a screening DoE to reduce the number of dominant variables.

If the total number of variables still is prohibitive consider replacing experiments with simulation, i.e. if you can model your problem sufficiently well - then you can use a 'digital twin' to learn more about it. More elaborate DoEs are doable when using software to drive other software, e.g. http://openmdao.org/ or https://www.noesissolutions.com/.

/Claes

There's an expression in the tech world, "Let's not try to boil the ocean", meaning: trying to accomplish too much. I would suggest that you are trying to test/experiment on too broad a level, and focus in on some aspect of your problem that is better defined, or reconsider the outcomes you are testing for. Again, by "tightening up", reducing the quantity of outcomes, the quantity of H0 hypotheses you are trying to negate, you can probably cut down the DoE load considerably.

Second, but related answer is Pareto's Rule: 20% of the factors product 80% of the outcome; are there a few factors that will control 80% of the results you are looking for?