Leave a comment
Get the GH Bookmarklet

Ask GH

It seems like a given that you should A/B test pricing, but I'm not sure whether it truly is the case or not. I haven't seen anyone write or talk about when you might not want to consider A/B test pricing. It's fair to say that pricing is not black or white thing so it, at least, seems plausible that there may be instances in which A/B testing pricing shouldn't be a (primary) concern? Anybody have any experience with this?

  • SJ

    Sebastian Johansson

    over 4 years ago #

    The only two reasons I can see is one, upsetting customers and hurting your brand. And second, if your country has some laws that forbid it (read in a blog post some years ago that some countries do, but not sure if its actually true and/or acted upon).

    A possible third I have experienced is a bit of an admin/coding headache when it comes to invoices, refunds etc.

    • TD

      Tarek Dinaji

      over 4 years ago #

      You can always use tools like stripe to avoid admin headache for invoicing and refunding.

      My thought is when you are small and you offer two pricing for the same product you often get penalized by users who comes to know that they were charged higher than another customer.

      But I think offers and discounts are common practice for a small to large scale companies now. So there should be little window when you don't want to offer two pricing and it is the time when you have just launched your product.

  • SE

    Sean Ellis

    over 4 years ago #

    Hey @Patticus do you have any comments on this one? I too am curious about your thoughts on A/B testing pricing.

    • PC

      Patrick Campbell

      about 4 years ago #

      Just seeing this, so sorry for the delay on responding, @sean. (for context, I'm the CEO of Price Intelligently - we use our software to build pricing strategies for a bunch of SaaS companies)

      In reality, I *wouldn't* A/B test your pricing if you fall into 80-85% of companies out there, because while potential PR debacles are definitely not great, in reality you likely don't have enough traffic or segments to run proper multivariate tests.

      It's not that A/B or multivariate testing is bad (it's actually ideal); it's just that there are too many factors, even in a SaaS offering, that need testing. As @JoeMurfin alluded to, there's not just the price point, but there's the packaging, framing, tiers, number of tiers, positioning, design elements, etc.

      This is what makes all of those "we reversed the order of our tiers and saw 12% growth" or "we changed our prices from 9s to 0s and say 21% growth" articles extremely frustrating, because those little tactics are masking the reality that the positioning/packaging/pricing are probably still really off.

      In reality, you should be running more of a cust dev/research process (stuff we talk about extensively on our blogs) to circle around to most of your answers and align your customer personas to your tier. Then you'll likely boil your entire testing framework down to a few axes ($50 vs. $100, 100 integrations or 80 integrations, etc.) that you can test with enough statistical integrity.

      Super quants hate this answer, but it's the most practical, because you don't want to run into a dunning kruger situation where you feel smarter because you ran an optimizely test, but in reality didn't really find anything out.

      All FWIW and happy to answer more. Practically, I'd quantify your buyer personas first and get your packaging aligned properly (through customer interviews and surveys). Then I'd A/B or multivariate test the prices.

  • JG

    jordan gutierrez

    over 4 years ago #

    If you have clients paying you already and you are testing a price that is lower than what they are paying...

  • JM

    Joe Murfin

    over 4 years ago #

    I think you'd be better off testing the framing of your pricing. Not just the pricing itself.

    Eg.

    3 for 2 on tickets to XYZ
    Or
    Buy 2 get one free to XYZ
    Etc...

    My guess would be the one using free* would win but always worth testing...

    Not forgetting comparing it to C amounts of cups coffee.

SHARE
13
13