Square’s S-1 shows that Starbucks transactions cost the company $77.1M, but Square only charged Starbucks $62.9M. How was this ever a good idea?

Stepping back a bit, I think it’s important to not put excessive pressure on any one marketing initiative.

Basically, you’re going to end up testing 3 types of marketing initiatives:

• Stuff that supports free leads you already have.  This typically is very high ROI.  Most customers at the end of the day come from (mini-)brand, word-of-mouth, virality, referrals, etc.  The CAC here can be close to $0.  You need to invest here, too, to support this revenue.  But it’s high ROI.
• Stuff that generates raw new customers — always expensive.  My goal here is $1 in ARR for $1 spent.  Now Square isn’t recurring revenue so it’s not 100% on point here.  But acquiring new customers is expensive.
• Stuff that just doesn’t perform.  You have to test stuff.  Some of it either won’t generate any new customers or will be so crazy expensive it’s impractical.

So you have to look at the mix.  Starbucks was an expensive experiment that did yield brand awareness and customers, but at less than a $1:$1 ratio.

If it helped put Square on the map (which it may have), and wasn’t the only source of new customers (which it clearly wasn’t) … it’s OK.

You’re supposed to have a mix like this.  You have to judge it as a cohesive whole, and then peel back or kill stuff that underperforms.