Due to the complexity of nuclear reaction models, current nuclear data evaluations must rely on experimental observations to constrain models and provide the accuracy needed for applications. For criticality applications, the accuracy of nuclear data needed is higher than what is currently possible from differential experiments alone, and integral measurements are often used for data adjustment within the uncertainties of differential experiments. This approach does not necessarily result in physically correct cross sections or other adjusted quantities because compensation between different materials is hard to avoid. One of the objectives of the recent CIELO project [M. Chadwick et al., Nucl. Data Sheets 118, 1 (2014)] was simultaneous evaluation of important materials in an attempt to minimize the effects of compensation. Improvement to the evaluation process depends on obtaining new experimental data with high accuracy and lower uncertainty that will help constrain the evaluations for certain important reactions. Improved experiments are accomplished by careful design with the objective of achieving high accuracy and lower uncertainty, and by designing new innovative experiments. New and unconventional experiments do not necessarily provide differential data but instead nuclear data that evaluators will find useful to constrain the evaluation and reduce the uncertainty. This also means that closer information exchange and collaboration between experimentalists and evaluators is important. For conventional experiments such as neutron transmission or capture measurements, it is important to understand the sources of uncertainty and address them in the experiment design. Such a process can also lead to the design of innovative methods. For example, the filtered beam method minimizes uncertainties due to background, and the Quasi-Differential Neutron Scattering method simplifies the experiment and data analysis and results in lower experimental uncertainty. A review of the sources of uncertainty in various experiments and examples of experimental techniques that help reduce experimental and evaluation uncertainty and increase accuracy will be discussed.