Some time near the end of 2010, Andrew Mason and Eric Lefkofsky did a little bit of subtraction, and some multiplication, and wound up with a $950 million check for Groupon’s Series G. The math worked like this: Groupon pays some customer acquisition cost to get a new user. Let’s say it takes $2 worth of advertising. A user signs up, and uses Groupon maybe once a month. On a $30 deal, Groupon collects about $18 in gross profit, spends around $15 (and dropping!) on acquiring and servicing small business users, representing an annuity of $3 a month. This implies a payback period of a couple weeks, and is a good reason for Groupon investors to dump some lighter fluid on the fire and fund the company’s wildest user acquisition dreams.

Having raised another ~$960m in IPO and secondary sales, Groupon’s value now sits at about two thirds of their pre-money valuation from their late 2010 round. Why didn’t the Groupon math add up after all?

The problem Groupon faced was that their initial success validated a model that anyone could copy, and everyone who copied it increased both Groupon’s cost of customer acquisition and its churn rate. Instead of paying $2 for a $3/month annuity that lasted for years, they ended up paying $20 for a $3/month annuity that lasted just a few months. (A friend who worked in the industry — and who swears that he will never, ever be involved in or invest in group buying again — says that the average cost per lead went from $1 to $20 in under a year when the industry got hot.)

In theory, Groupon could have just stopped spending until the craziness died down. The problem was that the craziness had embedded optionality: the public markets had enough demand that the #1 group buying company, and probably the #2 and #3, would be able to IPO. And given their growth rates (Groupon grew revenues 23x the year before it went public), every company with material revenue had a meaningful shot at cashing out in an IPO.

With an industry in flux like that, Groupon’s only option was to at least keep growing revenue faster than the next-biggest competitor in dollar terms. And that meant paying the market price for new users even when that market price made no sense.

But growth has to come from somewhere, and who better for a group buying site to target than the people who are already using Groupon? LivingSocial, Gilt City, and BuyWithMe all shamelessly copied one another’s ad copy and ad targets — so the users Groupon competitors acquired were often coming from Groupon itself. This pushed up churn enough to completely wreck the original logic of the business.

Groupon did, in fact, manage to make it to IPO, and they successfully buried their competitors under a pile of marketing overspending. But in all probability, the gold rush of group buying permanently reduced the size of the market: too many users got burned out on daily deal emails, and too many merchants got burned, period. Groupon exists, and it’s a viable business — GAAP profitable by 2018, if analyst consensus is to be believed — but Groupon is less of a success than it could have been.

The Groupon Scramble, Redux

Groupon comes to mind because of this memento mori about LivingSocial’s valuation going from $6bn to $0 in the Washington Post. But it’s also notable because the story is happening again, with a different cast of characters but the same math and the same incentives.

Now, the stylized model is: if users sign up to pay $60/week for three meal-in-a-box deliveries, and we make an incremental profit of $10/week, and just half of users like it and end up subscribed for another year, then we’re at breakeven paying ~$250 to acquire new users. And that’s before we upsell people on wine! A $250/user LTV covers a lot of marketing sins.

(We’ve long since passed the days when a Groupon marketing intern could sign up for a $50 Facebook ads trial account and net 50 new customers by lunchtime.)

But Blue Apron, the leading meal-in-a-box service, faces a familiar set of challenges:

• A competitor for the next big IPO: HelloFresh filed for an IPO, then withdrew their filing because their valuation was below the valuation their lead investor wanted. It is a truth universally acknowledged that in a sufficiently hot sector, the #2 player will be valued at the #1 player’s price/sales multiple, plus a couple turns for good measure. So the hotness of Blue Apron’s IPO is a predictor of how torrid HelloFresh’s marketing will be in the following six months.
• There are endless, super-specific clones: Groupon had to deal with local clones, demographically targeted clones, vertical-focused clones, even a gluten-free clone. Blue Apron has to deal with local clones, demographically-targeted clones, vegan clones, and even — yes! — a gluten-free clone.
• Since the group-buying bubble, ad options have narrowed. Facebook and Google have gained share, and any user acquisition tool that really scales has to scale on those platforms. The only new factor is the advertorial networks like Outbrain, Taboola, and Zergnet — the ones responsible for, say, this:

Tragically for Blue Apron, HelloFresh, and the rest, this new ad market is an even more efficient market than it used to be: what distinguishes Facebook/Instagram ads, AdWords, and advertorials is that they all appear side-by-side with — and compete with — organic content. This tends to make the ads more visible, and easier to track. With fewer sites controlling more traffic, competitive intelligence is easier: you don’t need to track a thousand publishers to see what ad Blue Apron is running. You just need to look at Taboola ad units, which will be identical across all the big publishers.

Will It Burst?

For a bubble to get crazy, you need two things: bad incentives and slow feedback. Mercifully, the meal-kit industry has pretty fast feedback. If Blue Apron starts losing users, they’ll be able to identify and plug the leaks.

Really, if you wanted proof that there was a bubble, you’d look for a refrigerated warehouse REIT. This is, as a business, not such a bad idea. REITs are popular because they’re tax-advantaged; they’re unpopular because the IRS only grants these tax advantages to businesses that are, in some sense, buying real estate and renting it out. But it turns out that, if you squint just right, you can have a datacenter REIT, or even a billboard REIT. A meal-in-a-box REIT would be a sight to behold: it would rent warehouses outside of major cities, and then lease space in those warehouses to whichever meal-in-a-box service could pay the most. And since warehouse space is never the biggest cost for these services, but schlepping boxes around without spoiling food is their biggest headache, the REIT would have extreme pricing power. Pricing power plus a tax advantage, plus downside protection (when the bubble bursts, you can always sell your warehouse to a grocery store, a food wholesaler, or Amazon), all adds up to a theoretically responsible way to gamble on a bubble.

That hasn’t happened yet. In fact, the ecosystem of meal-in-a-box-adjacent services is surprisingly weak; so far, each of these companies has rolled their own. So we’re still early, but if recent history is any guide, results will be disappointing.