By James Kwak
Groupon plans to go public later this week. According to the latest leaks, things are going well: the IPO valuation, scaled back from $30 billion to about $12 billion, may be raised because of a successful road show. Apparently even after the company conceded that the amount they pay to a merchant does not count as revenue, investors have decided they like what they see.
But there is still something fishy about Groupon’s business model.
The company’s basic story goes like this: We have been losing money (more than $300 million so far this year) because we are investing in building our customer base. In essence, we are paying for customers. But once we have a customer, we can harvest a long-term revenue stream from her, and the value of that revenue stream is greater than the amount we pay to acquire her in the first place. In an effort to show that this model eventually results in profitability, the company severely cut back on marketing expenses in the third quarter. That’s how its operating loss fell by $101 million while revenues only increased by $38 million. (See the quarterly results, p. 60.)
This is the most compelling evidence that their business model will work (p. 81):
“The Q2 2010 cohort is illustrative of trends we have seen among our North American subscriber base. The Q2 2010 cohort included 3.7 million subscribers that we initially spent $18.0 million in online marketing to acquire in the second quarter of 2010. In that quarter, we generated $12.8 million in revenue from the sale of approximately 1.2 million Groupons to these subscribers. Through September 30, 2011, we generated an aggregate of $92.8 million in revenue from the sale of approximately 9.4 million Groupons to the Q2 2010 cohort. In summary, we spent $18.0 million in online marketing expense to acquire subscribers in the Q2 2010 cohort and generated $92.8 million in revenue from this group of subscribers over six quarters.”
In other words, they spent $5 to get each subscriber, and have since earned $25 in revenue from that subscriber. Sounds good.
But there are still a couple of problems with this story. First, out of every dollar in revenue, Groupon incurs 64 cents in non-marketing costs (cost of revenue and SGA). So from that $25 in revenue you have to subtract not only $5 in marketing costs, but another $16 in other costs, leaving only $4 in profit (plus 36 percent of any future sales to those subscribers).
Second, Groupon’s long-term profitability depends on its ability to harvest revenue from existing customers. The problem is that, on average, sales to a given customer will follow a decay function. In the quarter in which you become a customer, by definition, you will buy at least one Groupon and quite possibly more. For example, Groupon entered Boston in Q2 2009 and sold 26,032 Groupons to 8,545 customers, or 3 per customer. On average, that number will fall over time as people lose interest, unsubscribe, switch to competitors, etc. The big question is how fast and how far it will fall.
I don’t have enough data points (and Groupon hasn’t been around long enough) to estimate that function. But I can do a quick-and-dirty estimate of sales to old customers by making an assumption about how many Groupons the typical new customer buys. This chart shows how many Groupons were bought by the average old customer in Boston, with different assumptions about how many were bought by the average new customer.
What you end up with is a number that is below 0.8 and falling. This means that when Groupon has saturated Boston (and it’s pretty close, with more than a million subscribers and almost 400,000 people who have bought at some point in the past) it can expect to bring in about 0.8 sales per quarter from its customer base. That’s not bad: with the current customer base, that would mean 360,000 Groupons per quarter. But what’s worrying is that that number has been falling for more than a year, meaning that the productivity of the existing customer base is decreasing. How far that will fall is anyone’s guess. That fall could be part of the decay function, in which case it will level off as the business matures. Or it could be something else, which might not level off.
Another big question mark is that Groupon’s business model depends on its ability to make money off existing customers without spending a lot on marketing. In theory, this makes sense, since email is cheap. But is it true?
We know that Groupon slashed its marketing expenses in Q3 in order to pretty up its financials. That should have had no impact on its existing customers, if you believe the company’s line. (“Once acquired, subscribers have been relatively inexpensive to maintain because our interaction is largely limited to daily emails and our mobile applications,” p. 80.) But in both Chicago and Boston (the two North American cities that Groupon breaks out numbers for), the total number of Groupon sales fell—by 11 percent and 5 percent, respectively. Since Groupon added new subscribers and customers in both cities (though not as many as in previous quarters), that means that sales to the existing customer base did particularly badly. This implies that sales to the current customer base are affected by marketing spending. So we really don’t know how Groupon will do without a continuing blizzard of marketing spending. The only data point we have to go by is Q3—when the company was getting plenty of free publicity because of the run-up to the IPO—and that isn’t encouraging.
Then there’s the fact that revenues as a share of billings (that is, the share of what customers pay that they keep) have been falling steadily, which could be an indication of increasing price competition.
Now I can’t prove that Groupon won’t be profitable in the long run. It’s quite possible that as they saturate markets, they can find a level of marketing that generates enough sales to old customers (since there won’t be many new ones) to cover their expenses, and that they’ll be able to keep prices high enough despite competition. In fact, my guess is that they probably can. But there’s nothing in the S-1 that proves that that is likely, either, because we can’t see the numbers that matter.