Tuesday, January 11, 2011

To ask or not to ask (for the user's credit card), that is the question

I've just answered a question on Quora and thought it might be worth cross-posting it here. Just in case there's still someone who isn't using Quora yet (admittedly unlikely given their current growth rate which is absolutely incredible).

The question was:
"For web apps, is it better to ask for the credit card before their trial starts (e.g., on the signup page) or after their trial expires?"
My Quora answer follows below.

* * * * *

I don't know the answer, and there probably is no general answer, but I recently put together a little model that helps to understand the determinants better:

Click here to open the Google spreadsheet

The blue values are sample (dummy) input values that you can change. The model is based on the following ideas and assumptions:

  • There are two types of visitors: Those who sign up whether or not a credit card is required (let's call them Group 1) and those who sign up only if no credit card is required (Group 2). There is of course a third group, those visitors who don't sign up in neither case, but we don't need them here.
  • If you require a credit card you have a certain visitor-to-signup conversion rate from the users of Group 1 (cell D8 and D32). And per definition, no signups from Group 2 users in that case (D33). If you move to a no-credit-card signup, on top of the signups from Group 1 (D9 and E32) you get a certain amount of signups from Group 2 users (E33) so your total visitor-to-signup conversion rate is higher (E34).
  • Looking at the trial-to-paying conversion rate, let's assume there's a baseline conversion rate of Group 1 trial users in the CC-required case (D12 and D38). If you remove the CC requirement I would expect that rate to drop (D13 and E38), because a) trial users who have provided their CC already may feel higher 'pressure' to try the product within the trial period, they feel more 'invested' and are less prone to procrastination; and b) doing nothing is easier than actively terminating your account. Some users will forget to terminate or just don't care.
  • Looking at Group 2 users, again of course no signups or customers in the CC-required case. In the no-CC case you'll be getting a certain amount of trial-to-paying conversions from those users (D15 and E39). I would expect that rate to be lower than the baseline conversion rate because Group 2 users are, on average, inherently less interested in your product than Group 1 users (more tirekickers).

So far so good. You can see the number of paying customers for each of the two cases in row 40.

The next thing to look at is churn, i.e. users who cancel after you've charged them at least once (I'm using charged-at-least-once as the definition of a 'paying customer' in the model):

  • I would generally expect the churn to be higher in the first few months following conversion to paying because some users may still be in their 'extended trial period', even if they're paying already. Also, the longer your customers use your product, the more value they will hopefully derive from it so they get less and less likely to cancel (D19 vs. D 21).
  • In the CC-required case I would expect that difference in early churn and later churn to be higher, maybe much higher, because many of the users who forgot to terminate within their free trial will terminate within the first months after subscription (D18 vs. D20).

Using the data on customers, churn and your revenue (or gross profit) per customer per month you can now calculate if you're better off requiring (D48) or not requiring (E48) a CC upon signup. One last factor that I've included is the cost that it takes to serve a trial user, e.g. bandwidth and time from your support team, which may or may not be significant depending on the nature of your business.

There are of course a couple of caveats:

  1. The sheet is completely useless until you've tested it and until can fill it with real data. The purpose of the model is NOT to replace real-life testing by making some assumptions and pretending that that lets you decide which option works better. Quite the opposite – the purpose of the model is to understand which parameters you should look at and measure.
  2. The model doesn't include all factors which may be relevant (their relevancy depends on your business, and I didn't want to make it too complex). For example, one question is if you consider tirekickers an asset (because even if they're not interested in buying your product yet, they may tell their friends about it or come back to you later) or a burden (because they divert resources away from the more strongly interested prospects). Another factor that I haven't included are different pricing plans – I've included just one price per customer per month.

Feedback very welcome!

No comments: