We study the winners. We reverse-engineer narratives about why they won. We skip the thousands who did similar things and lost. I started writing these “Munger’s Follies” pieces to fix that. I’ve been building a mental model toolkit for the last few years. Munger, Hormozi, Kahneman, Cialdini. Frameworks for thinking about decisions, incentives, and the ways smart people fool themselves. Reading about a bias once doesn’t fix it. You have to see it in the wild, over and over, until your brain starts pattern-matching on its own.

Blue Apron is one of the best classrooms I’ve found.

The founders were very smart people. Matt Salzberg went to Harvard Business School. The team raised $199 million in venture capital from serious investors. The product worked. People loved getting those boxes. Forbes put Salzberg on their cover.

And the company sold for $103 million in 2023, roughly what they used to spend on marketing in a single year. A 95% decline from IPO valuation.

CAC and LTV: The Two Numbers That Tell You Everything

I need to explain two terms because the entire Blue Apron story runs on them.

CAC is Customer Acquisition Cost. How much you spend to get one new customer. The full calculation includes everything you spend to acquire: ad spend, marketing salaries, software, agency fees, content production. Take the total and divide by the number of new customers in that period. If you spent $500,000 across all acquisition costs last quarter and signed 1,000 new customers, your CAC is $500.

LTV is Lifetime Value. How much total profit a customer generates before they leave. If the average customer stays 8 months and you make $15 per month in profit from them, the LTV is $120.

Those two numbers run every subscription business. When LTV clears CAC by a wide margin, the math works. Spend $100, get $300 back. Do it again. Scale as fast as you want.

When CAC sits above LTV, every new customer costs you money. Growth makes the problem worse, not better.

Blue Apron reported a CAC of $94 in their IPO filing. That number was a lifetime blended average going back to 2014, when friends and family signed up for free. The actual math from the year before the IPO: $178 million in marketing divided by 387,000 net new customers. That’s $460 per head.

The number they told investors was $94. The number running inside the business was $460.

On the other side, Blue Apron had a 26% contribution margin. For every dollar of revenue, they kept 26 cents after food, packaging, and fulfillment. At $460 to acquire a customer, you need that customer to generate about $1,770 in total revenue before you break even. At their order frequency, that takes roughly 14 months of continuous ordering.

72% of customers churned within six months. 50% were gone after two weeks.

And it was getting worse. Every new cohort of customers generated about $7 less in revenue over the first six months than the cohort before it. The business was degrading from both directions at once: acquisition costs rising, customer value falling. Blended averages hid that. The cohort data didn’t.

Why “blended average” is dangerous. There’s a concept I picked up from Munger’s writing on self-deception: the most dangerous lie in business is the one you tell yourself with real numbers. A blended average mixes your cheapest customers (early adopters, friends, referrals) with your most expensive ones (paid acquisition in a crowded market). The blended number always flatters. In a growth business, the only number worth watching is the marginal cost: what it costs to add the next customer today. Blue Apron’s blended CAC was a rearview mirror. Their marginal CAC was the windshield. They kept looking in the mirror.

The Behavior Gap

The real lesson lives outside the spreadsheet.

Blue Apron’s product required customers to change a daily habit. Come home tired after work. Skip DoorDash, skip the rotisserie chicken at Costco, skip the frozen pizza. Instead, spend 45 minutes following a recipe card with pre-portioned ingredients.

Week one, it’s exciting. You feel like a chef. The Instagram photo gets likes.

Week six, you’re exhausted and the box is sitting in the fridge unopened. Week eight, the guilt of wasting food pushes you to cancel.

That loop ran for millions of customers. Blue Apron’s response was to spend more on marketing. Better targeting, more channels, bigger discounts.

62% churn in six months is not a marketing problem. It’s a product problem. The product was asking people to become someone they weren’t. At least not on a Tuesday night after a long day.

People wanted to want to cook. That’s different from wanting to cook. Blue Apron sold the aspiration. The aspiration doesn’t survive contact with a 10-hour workday.

Commitment and Consistency bias, running in reverse. Cialdini’s research found that small commitments build into larger ones. People who sign a petition are more likely to put a yard sign up later. The first small act shapes identity. Blue Apron needed that to work: order once, cook once, feel like a cook, keep ordering.

But 45 minutes of meal prep on a weeknight isn’t a small commitment. It’s a big one. And when customers failed at it (because life), canceling the subscription was the easy part. The harder part was admitting to themselves they weren’t the person who cooks from scratch on weeknights.

That’s worse than regular churn. Regular churn means someone found a better option or lost interest. Identity failure means the customer associates your brand with their own inadequacy. That emotional sting doesn’t fade with a win-back email. Blue Apron lost customers and left behind millions of people who felt bad about themselves every time they saw that blue box in their inbox.

Try running a re-engagement campaign against that.

The Moat That Never Got Built

Blue Apron spent hundreds of millions educating American consumers about meal kits. When they started, there were about 13 meal kit companies in the US. Today there are over 380.

Blue Apron funded the education. Everyone else showed up to collect the customers.

I think of this as “man with a hammer” syndrome. When your only tool is a hammer, everything looks like a nail. Blue Apron’s hammer was marketing spend. Every problem looked like it needed more awareness, more reach, more top-of-funnel. Awareness without barriers is a donation to your competitors.

HelloFresh launched the same year with the same product concept. But they did something Blue Apron never did: they put cold-callers on cancellations. When a customer tried to leave, they got a phone call. Not an email. A human being asking why and offering a reason to come back. HelloFresh fought for each customer individually. Blue Apron bought attention at the top and watched it leak out the bottom.

HelloFresh also did something subtler that gets less attention. They built a portfolio: EveryPlate for budget customers, Green Chef for health-conscious buyers, Factor for people who didn’t want to cook at all. Different price points, different commitment levels, different behavioral asks. Blue Apron treated every customer the same. One product, one price, one experience. If that experience didn’t fit your life, you left. HelloFresh gave you somewhere else to land inside their ecosystem.

Then grocery chains launched their own kits. Costco, Target, Kroger. No shipping cost, no subscription. You grabbed it while you were already shopping. Then Amazon announced they were buying Whole Foods for $13.7 billion, two weeks before Blue Apron’s IPO.

Amazon didn’t kill Blue Apron. Amazon revealed that Blue Apron had nothing defensible. No proprietary recipes worth protecting. No switching costs that made leaving painful. When you spend to build a category without building walls at the same time, you’re doing free marketing for whoever shows up next.

The Discount Spiral

Blue Apron grew partly through free first boxes and referral credits. Aggressive discounts. The kind of offers that attract people who want a deal, not people who want your product.

Data from 1010data showed 50% of Blue Apron customers churned after the second week. Half gone in 14 days. These were coupon clippers, not cooks.

Discounting to acquire is borrowing against lifetime value. You pull tomorrow’s revenue into today’s top line. If the customer leaves before you collect, you’ve donated that acquisition cost to nothing.

The spiral is self-reinforcing. Discounts attract worse customers. Worse customers leave faster. Faster churn means you need more marketing to replace them. The marketing gets more expensive because you’ve already reached the easy audience. Each quarter is worse than the last, but the blended averages hide it until they can’t.

By March 2018, Blue Apron had lost 81.4% of its market value since the IPO nine months earlier. Salzberg resigned by November 2017, five months after the IPO. That speed tells you how fast the board’s narrative shifted once public market scrutiny replaced private investor patience.

The Incentive Trap Nobody Talks About

Blue Apron’s board was stacked with venture capital partners. VC funds operate on fixed timelines. They need exits. The path to exit was an IPO. The path to an IPO was growth. The path to growth was spending. Every incentive in Salzberg’s environment pointed at one behavior: grow faster.

A CEO who slows acquisition to fix unit economics is a CEO who misses the growth targets the board set. A CEO who hits the growth targets gets a Forbes cover. Which behavior gets rewarded?

This is the mental model I keep coming back to: incentive-caused bias. Show me the incentive and I’ll show you the outcome. Blue Apron’s incentive structure made it nearly impossible for any rational CEO to pump the brakes. The system rewarded the behavior that killed the company.

I say this not to excuse Salzberg but to be honest about the structural forces. We love assigning blame to individuals. It’s satisfying. But when the incentive structure points everyone in the same direction, the more useful question is: how do you design incentive structures that don’t require heroism to produce good outcomes?

Four months before the IPO, Blue Apron announced a 495,000 square foot fulfillment center in Linden, New Jersey. You don’t break ground on a half-million square foot building when your unit economics are underwater. You do it when you’ve confused activity with progress. Or when every incentive around you rewards building bigger instead of building better.

Salzberg compared Blue Apron to Netflix and Uber after the collapse. He said the blitzscaling playbook failed in meal kits. He’s right about the outcome. But Netflix’s marginal cost to serve one more subscriber approaches zero. The content exists. The servers exist. Blue Apron’s fulfillment required real food, real hands, real cold-chain logistics for every single box. Roughly 80% of employees worked in those fulfillment centers. The variable costs never compressed with volume. Importing a Silicon Valley playbook into a food logistics operation is “man with a hammer” thinking in its purest form. When your only framework is blitzscaling, every business looks like it needs more capital and more speed. Some businesses need discipline and better margins instead.

In Part 2, I’ll try to answer the harder question: what could Blue Apron have done differently? And more importantly, what did the companies that solved this same churn problem actually do? Because the uncomfortable truth is that the playbook existed. Blue Apron just never ran it.