You can't, really. There's absolutely no reason why someone would believe magic is at play, unless they're already primed to believe in magic (which, mind you, plenty of scientists are guilty of!). Heck, if magic was real, we would simply let it join the rest of the stuff we call "science", because science is not any particular set of technologies or tools - it's a toolset for learning about what is real. Real magic is no different. No cheap tricks like transmutation or FTL communication are going to change that - it would require us to take another deep look at the things we took as fixed (or what the magic actually did, as opposed to what we observed it did - e.g. is it really a violation of the conservation of energy, or did the energy come from somewhere else?).
The magic that would be by far the hardest to swallow would be something complicated. Transmuting a chunk of one element into a chunk of another element doesn't qualify - it's interesting, but simple. What would be really complicated, but is taken for granted as simple in human imagination? Minds and bodies. People take them for so simple that they routinely "grant" human-like minds to rocks, animals and weather effects. But in reality, they are insanely complicated.
So for the most plausibly magical trick, I'd go with a polymorph. Change a human into a cat. Have the cat send a message while being a cat. Change it back. Have the human report back. Give the opportunity to every other scientist who claims it's just a trick. Is it fool proof? Hell no. Humans are susceptible to suggestion, and it's plausible enough that you used a drug to enhance that. We already know drugs who do something like that. It's also a lot more plausible you used some illusion (a perfect free-floating 3D hologram is far beyond what we can do now, but far more plausible than magic). You will always have people who couldn't experience it on their own. And even if they did, how could they trust their own experience? There's just too many ways that people routinely fool themselves, even without assistance. But at least it's something really complex, and not just something that could be answered with a simple technological device. Not impossible, but complex. Almost all magical tricks are really, really simple. This isn't.
Really, in general, go for things that sound simple to a human (because our brains do it automagically), but really aren't. Things like "walking through walls" sound really simple to humans (and are common magic in both fantasy and soft sci-fi), but answers to questions like "how are you walking in the first place if you're intangible?" are very easy to answer in human-language ("the spell knows the difference between a floor and the wall, duh"), while being nigh impossible in reality-language. But rather than just scientists, you really want practicing magicians (the trick-making kinds of guys, like James Randi) in on the demonstrations. You need people who are as non-gullible as humanly possible.
This still doesn't guarantee anything, of course. There's nothing you could do that would be absolutely unexplainable with science. Again, science is just about finding out how the universe works. If magic works, and it has rules, it will become a normal part of science. That's one of the points of Clarke's Law - and of course, Agatha Heterodyne's corollary, that sufficiently analysed magic is indistinguishable from science ;) In the end, if magic works, and people can use it... who cares whether you believe it to be supernatural? It clearly isn't, no matter how weird it is.