tl;dr - It looks like it would take over a year to determine price elasticity with significant numbers. What am I doing wrong?
----------------
I have an existing freemium SAAS product with three paid tiers targeted towards different customer segments. I want to determine the optimal prices for these paid tiers. Let's assume I am locked into these three products/tiers and can only adjust their price.
This already seems overly difficult, so I'm thinking of optimizing them one at a time, starting with the most profitable tier first. I realize this would mean we can only achieve a local maximum, but at least it gets us started.
So what I am learning is that I need to first determine the price elasticity of this product, which is done by measuring the impact on demand of a 1% increase in price.
Here's where I run into problems. ABTesting has taught me to always look for significance before trusting any data. But as a freemium product, the overall conversion rate of people who see these prices and convert to purchase this tier is low in real terms. Because of this, the minimum sample size I would need to detect a 1% impact on conversion (therefore a 1% impact on revenue) with a reasonable significance is so great that I would need to run this test for over a year before I can trust the numbers.
All the literature I am reading starts out with something like "given that you sell X widgets at $100 and Y widgets at $101, this is how to calculate price elasticity." But in a practical sense, how can I trust that these numbers aren't just noise?
I'm open to the idea that I am thinking about this in completely the wrong way. Links to articles or books that offer practical guidance are very appreciated.
For my own SaaS pricing we didn't display prices for months and provided custom quotes. Then set prices (standardized email) and saw at least 50% of users thought it's too expensive. We adjusted a bit but realized however low there will always be users you can't please.