Why (& How) You Should be A/B Testing Your Pricing?
Oh, the question thou fear the most:
What price should I offer my new product X at?
Determining right price for your product is perhaps one of the most difficult tasks when you are launching a new product or service. Pricing for physical goods is simple. For example, if you are manufacturing staplers, all you need to do is to calculate cost of production and distribution, slam 20% margin on it and there you have the price you can sell your shiny stapler machines for.
Economics 101 (price elasticity of demand)
But for digital goods with no cost of production, it is not that simple. This zero cost of production complicates pricing decisions because then you need to price the product according to how much the market values your product. At the core, idea is quite simple: higher you price, lower the demand is. However, if your price it too low, you won’t make a lot of money even though you might sell a lot. Similarly, if you price it too high, you won’t make a lot of money even though each unit sold brings you greater amount of money. This is the basic principle of price elasticity of demand.
So, as you can see in the graph above, increasing price by 10% reduced quantity bought by 15% which reduced total revenues. Every product has a price point in the graph at which revenues become maximum. Price more than it, revenues will fall. Price less than it, revenues will fall. Of course, you can’t sit over coffee one evening and draw this price-demand curve for your product. It has to be discovered. Your market determines this curve and A/B testing is an excellent way to find out which price-point maximizes the total revenue.
How to set a price range for A/B testing
Theoretically, the price-demand curve is infinitely long. Price runs from zero to infinity (Y axis) and so does demand (X-axis). But, of course, practically you need to have a price range in mind which you think is suitable for your product. For example, if you are selling an eBook you need to see if $15 gets you more revenue than $9. And you would probably be wise enough to avoid testing selling it for $100. The key question here is: how to set initial price range for price testing?
The answer is: don’t just roll the dice. Even though the science says that it is about software pricing, I have found it applies to many types of digital products. The basic gist is this: look for other similar products in the market and also look at the value your product is delivering. Set a price range accordingly. Once you have a price range in mind (say $50-$150), next step is to use A/B split testing to determine the exact price which maximizes revenues.
The Dark Art of Price Testing
Price testing is definitely one of the most difficult projects you can undertake. There are so many things that can go wrong. Consider this worst-case scenario: an influential blogger (say Mike from TechCrunch) is trying out your product and somehow gets to see that you are doing price testing. He writes about it on the blog (because, hey, it is fun to write about something controversial). Your customers read the post and get angry at you. Worst-of-the worst, one of the customers turns out to be idle lawyer and sues your company. It is a worst case scenario but quite plausible.
To avoid OMG, we got sued due to price testing, you should be doing price testing according to following rules (which I classify as the good, the bad and the ugly — in reverse order).
The Ugly: never offer exactly same product / service at different price points
Yes, you read it right. This is perhaps the way many companies do price testing but you should NEVER show different prices to the visitors for exactly the same product or service.
It’s illegal and can lead to huge potential lawsuit.
The Bad: have slightly different offering for different price points
This is a less nefarious version of plain-old price testing. Instead of showing different price points for the same product, you show different price points for slightly different product offerings. You can vary product offerings tested at different price points by adding or removing some trivial features. I will give you an example, if you are selling a backup service you can create one version where you offer 5 GB storage for $20, in another version you offer 5 GB storage + SSL (trivial feature) for $30. So, practically both offerings are similar but technically there is a difference and if anyone ever questions you, you have grounds for justifying the difference in price. After all, you are offering different products (no matter how trivial the difference is).
But I consider it immoral. Yes, you can evade potential lawsuits but anyone will know that you are fooling people.
The Good: offer different products (or plans or solutions) at different price points
This is the most ethical way to do price testing. Ideally, you should offer completely different product plans at different price points. Taking backup service as an example again, if on your pricing page lowest tier offers 5 GB for $20, test a version where you offer 10 GB for $40 and 2 GB for $8. You are trying to gauge sensitivity to price here. If your conversion rate 10 GB is same as that for 2 GB, this means your service is so compelling that people want don’t care if it is $8 or $40. So, in the next update you ramp up price as $40 for 5 GB (while still grand-fathering old customers). This way you would know what is the best price point for your service.
Of course, not all digital products have luxury of offering pricing plans. What if you are selling an eBook? In that case, you need to add some extra value (e.g. 15 minute consultation with author) if you are trying to test a higher price and remove some value if you are trying to test a lower price (e.g. shorter version of ebook).
The key lesson for using A/B testing to determe ideal price is this: offer different value at different price points to gauge price sensitivity of target market. Then whatever price offers maximum revenue, start offering your main product at that price point (while grand-fathering old customers).
Final Gospel: measure revenue, not conversion rate
I have suggested it earlier in the article but will make it clearer here. During price testing, you should measure revenue (not conversion rate). Because even though you may end up selling less (hence lower conversion rate) at higher price points, your total revenues may actually be higher.
VWO lets you measure revenue by integrating with Google Analytics and Omniture SiteCatalyst. So, if you are measuring revenue in one of these analytics tools, you can easily see which price variation resulted in maximum revenue. (Even if you measure revenue in internal dashboard or excel, it should be quite simple to measure it for different variations)