There's a natural distribution to these things, so actually an average of 3 does in fact say "something" about the tail behavior of concurrents (that is, the highest values you can expect under peak load).
As a rule of thumb: it's probably on the order of 5x-10x (absent outlier events); then look at your data to find out what it actually is (users are not evenly distributed across the day/week); then apply the usual 4x or so rule for capacity planning.
That is to say - it isn't necessarily a scaling issue. Especially when you're just starting out and no one is looking at your damn app anyway. If it's well designed and cleanly built you'll have plenty of time to factor for scale when the time comes (if it comes).
