I'm going to assume radio frequency transmissions were meant in the question, since there are other possible electromagnetic wavelength signal carriers, e.g. laser communication, and all kinds of other technomarkers, but repetition on itself is not the best indication of intelligent communication. And neither is periodic repetition, because we know of natural radio frequency emitters that can be both repeating and periodic - pulsars.
The first detected pulsar in 1967 was actually designated LGM-1 (Little Green Men) by Burnell and Hewish, because it wasn't yet clear at that time that the radio signals received were a naturally occurring phenomena, yet the pulses followed sidereal time so it was confirmed they're not of local origins. And there are other naturally occurring radio signals, perhaps better said radio interferences since they don't carry any structured data and are just more-or-less random noise, like e.g. whistlers, shorter of which can be difficult to distinguish one from another, and some other such noise can be periodic too, e.g. radio interferences or intermissions caused by occulting natural satellites. I explain some others in this answer to the Jupiter FM question on Astronomy.SE.
Better demonstrators of signals being of intelligent origins are signal clarity (strength limited to a narrow band, i.e. spectral width), if they are carrier waves (polarized or modulated) and can contain readable data. This is perhaps better explained by SETI in their FAQ answering the question How do you know if you've detected an intelligent, extraterrestrial signal?
The main feature distinguishing signals produced by a transmitter from
those produced by natural processes is their spectral width, i.e. how
much room on the radio dial do they take up? Any signal less than
about 300 Hz wide must be, as far as we know, artificially produced.
Such narrow-band signals are what all SETI experiments look for. Other
tell-tale characteristics include a signal that is completely
polarized or the existence of coded information on the signal.
These are just some markers that the signal might have been sent by some intelligent civilisation however, and the signal origins would first have to be confirmed to have originated from outside our Solar system (or in the frequency range and position within the Solar system where we wouldn't expect any of our deep space probes to communicate). This can be done like already mentioned in the story about the first pulsar being detected, with an extrasolar signal following sidereal time, and using parallax (triangulation) to establish distance to the signal origins. And signal strength can tell us direction it is being transmitted from.
Ideally of course, the signal would be decoded and we would detect some intelligent structure in it even before we'd be able to actually understand its message, for example use of metadata in the packet headers that precede actual data being sent and describe some of its properties (e.g. message length), use of BOF/EOF (Beginning of File / End of File) markers, use of large primes that might be used for data encryption, and so on. Any such data structures would clearly be of intelligent origins and all we'd have to establish is that they're not being transmitted by us. Which includes radio signal echo of past transmissions as they bounce back towards us off some celestials in our cosmic neighborhood.
Single such transmission with only a few such signatures might not be enough though, you're right in your question that many of such signatures could be put down to chance, a random fluke radio transmission that was of natural origins, and somehow managed to confuse us into believing it might be of some distant civilisation. It wouldn't be the first time (LGM-1) either, so we would be looking for a certain total complexity, either in a single message, or many successive messages. At some point, we'd be able to either decode / decrypt the message, parts of it, or conclude it can't possibly be random noise, because its total signature's Shannon entropy is too small in relation to its total complexity (internal repetition, i.e. its compressibility is a good measure of that) and it simply must be a data carrier signal.
With regards to the Wow! signal specifically, since we've never been able to detect any repeat transmissions and no other sense has yet been made out of the single one, besides that it was a curiously narrow-band and strong SNR signal, if we were able to obtain more data, that would obviously help determine its nature with greater certainty. For the time being though we don't have any conclusive evidence to explain it, and Jerry R. Ehman that circled that print-out and wrote Wow! next to it said it best:
"I choose not to draw vast conclusions from 'half-vast' data"