Well if all you have is 170 people then you're not sampling but testing the population.

I'd go with the A/B testing option just to get more experience with it. Having data at a low confidence interval is better than having no data at all. Maybe it's a wash maybe it's actually very important. Worst case is there is no overwhelming winner and you're back where you started.

Use the following link to determine the sample size you need. http://www.surveysystem.com/sscalc.htm