It's never too early to measure. Measurement is pretty much a universal good as far as I'm concerned. Especially since starting early gives you the chance to get good at measurement before being good at measurement is really important (if you see what I mean).
What you need to be concerned about is how you make decisions based on the measurement.
It might be too early to (usefully) rely on statistical confidence for minor tweaking - it depends on the significance of the results that you see. VWO have a handy online tool for seeing whether things are significant or not statistically http://visualwebsiteoptimizer.com/ab-split-significance-calculator/ - experiment with the numbers to look at the kind of thing that might be useful for you.
What might get you some more bang for your buck is doing some quick and cheap user testing with, y'know, actual people ;-) If this is not something you've done before I'd thoroughly recommend getting a copy of Steve Krug's Rocket Surgery Made Easy.
Also - remember that you don't always have to rely on statistical significance. If you see two people walk past and trip on a loose corner of carpet - you don't have to wait for N other people before knowing that it's a problem and you should fix it.