There is no science, as long as the outcome can not be confuted.
There exist two types of research: scientific and non-reproducible. The difference stands in the way it is approached and communicated, not in its goals or applied methodology. Both are worth being pursued, as both can help the society progress.
Scientific research requires approaches that ensure reproducibility of results. As long as a research activity (be it a field experiment, or the outcome of a simulation or a mathematical algorithm) is performed, encoded and reported, so that anyone else might repeat the procedure and reproduce the same outcome (possibly, in statistical sense), it is science. As a consequence, it can be confuted, extended, or improved. According to Karl Popper (“The Logic of Scientific Discovery”, 1934), science is where results can be shown to be erroneous.
Therefore, a scientific publication can be claimed to be such, as long as all details of the experiment, simulation or numerical evaluation performed, are given, and any other researcher can find the same results by reproducing the model presented in the paper.
Sometimes research activities are performed without respect for the dogma of reproducibility, as long as this would require an effort not possible according to the resources committed. For instance, an experiment is performed on the field, without recording the full context. Or, a technical report about a simulation is produced, without full description of the settings and the model used. Results are useful (if they are) only per se.
These types of activities can however inspire new scientific research lines. Inductive reasoning is sometime needed to create new knowledge, and can be started even from non-reproducible experience. Moreover, resources (of human or financial nature) are often limited and researchers need to take simplified approaches.
The Radio Networks group by nature performs scientific research, in the context of the evolution of Mobile Radio Networks and the Internet of Things, though sometime we approach specific goals through non-reproducibile methodologies when time contraints require so.