Quarter to quarter, companies are evaluated based on aggregate metrics: free cash flow, EBITDA, revenue, etc., often on a per-share basis. And sure, those metrics can be telling, but in the long run, companies are best understood at a granular level. Deeply understanding a company often means approximately replicating Richard Feynman's observation that:

Figuring out a company's unit economics starts with figuring out the "atoms" of that company’s business. Last week's issue gave some examples: a restaurant might look at its unit economics in terms of the performance of individual locations, while a subscription software business might look at it in terms of seats.

So how do we figure out the right “atoms” to understand a business? There’s no silver bullet here; you need to do some thinking—those units should, ideally, be:

• Fairly homogeneous, either in size (one US Netflix subscriber is pretty similar to another) or in behavior (a five-person company and a 5,000-person company both have some deployment pattern for new SaaS tools, where they start out with a small number of users on cheap or free plans and expand their seat count and price until they reach a steady state where growth is bounded by customer growth).

• Not easily divisible, because if they're easily divisible the proper unit is whatever they divide into.

• Limited by discrete categories, ideally qualitatively: A Netflix model might build up revenue estimates by tracking ad-supported and non ad-supported users separately, for example. And that’s okay, but it's hard to look at unit economics trends for companies that divide up their "units" by size, because it means individual constituents are switching from one bucket to another. Specifically, shrinking enough to drop below one cutoff makes the average revenue per customer above and below the cutoff rise.¹

• And most importantly for valuing the business, it needs to be possible to match some costs to fixed unit-level costs, some to variable unit-level costs, some to unit acquisition costs, and some to expanding-the-set-of-acquirable-units or increasing-maximum-revenue-per-unit costs.

This last point is where unit economics really shine as an analytical tool, because they let you convert all the complexity of a company into a fairly simple formula: there's some cost to add a unit, whether that's the cost of building a restaurant or the cost of marketing to a new customer. There's some semi-predictable stream of cash flows from the acquisition of that unit. And that means growth can be converted into a net present value calculation: spend $1m building a restaurant that produces $150k in after-tax cash flows, growing 3% annually, and you can either a) say that you're investing at an 18% internal rate of return, or b) say that at a 12% discount rate you're converting $1m in expenses into an asset with a present value of $1.67m.²

Being able to do this repeatably is a wonderful trait for a business, because it gives them a nice set of strategic options: when there's a wide spread between what it costs to increment the unit count and what that's worth, they can just focus on increasing unit numbers. Restaurant concepts often end up in this position, especially if they're the first to scalably commercialize a cuisine. But once it works in a few regions, it will probably work in others. Then, as the business gets saturated, the question becomes: how do we reduce the upfront cost of creating new units³ and how do we increase the returns on each unit?

Thinking in terms of unit economics is also useful for understanding why early-stage company valuations can expand so explosively. If there's a large addressable market, and a company can acquire customers for much less than those customers will generate over a lifetime, then:

1. The company should raise money to expand as fast as possible in order to take advantage of this,

2. The company should raise money to expand as fast as possible lest someone else, perhaps with favorable but not quite so great unit economics, is the first to raise and ends up taking over the market, and as a result

3. A corollary to this is that a company that's convinced investors that it has good long-term unit economics will, as a result, see its revenue rise while its margins go down, often from -50% or so to -100% or lower, because it's spending so much upfront to capture that future revenue. This result seems paradoxical, but it’s true: in an efficient market, the better a company does at achieving product-market fit early on, the more unprofitable it will be in the early stages of its growth—which will hopefully pay off later on.

Step three can burn a lot of capital if either the unit economics estimate is wrong or the addressable market isn't as big as the company thought. And it can take just one hyper-optimist in an easy fundraising market to ruin the economics for everyone: if Uber overestimates the ride-sharing market's size by 2x, and Lyft gets the estimate right, the best case for Lyft is still to raise enough money to keep up with Uber, and then to hope that Uber eventually gives up. (This is more or less what happened, and Lyft is worth more than they would have been worth if their response to Uber's growth was to say "This is ridiculous—we're out.")

Step three also burns plenty of capital if those long-term unit revenue and unit margin numbers are off. That's been the story for many mini-bubbles in the last few years: group buying circa 2010, mealkits around 2016, 15-minute delivery in the last few years, true crime podcasts (better DNA testing and cheaper recording equipment means that we simultaneously have a surplus of podcast hosts and a dire shortage of serial killers) etc. In each case, growth was predicated on recurring revenue from newly-acquired customers, but increasing competition meant that customer churn went way up—it's hard to stay loyal to Blue Apron when HelloFresh offers a 50%-off coupon, and once every customer is conditioned to expect offers like this, their expectation is for a continuously cheap product.

And that explains yet another mystery: why is it that when companies report some tiny incremental shift in their unit economics, their stock sometimes takes a massive dive? The reason is that the valuation of a business is basically the gap between customer acquisition cost and customer lifetime value, minus the fixed costs needed to create the opportunity in the first place. A 2% revenue shortfall, compounded over several years and compared to more stable costs, can cut the valuation of a business by a third.

There is a lot of competition to invest in companies with simple unit economics. The mature ones tend to trade at high multiples to steady profits, and the emerging ones can trade at even higher multiples of sales as people extrapolate future growth. But digging into a business to truly understand its unit economics is worthwhile when they’re complex and counterintuitive, because that tells you what every incremental dollar of investment does for owners. And as an incremental owner, that's what you care about.

1. An example of this comes from the last quarterly call by Zeta Global, a marketing tech company: "Super scaled quarterly ARPU [i.e. revenue per customer for customers spending over $1m/year] was $1.3 million, up 26% year-to-year. This drove 34% year-to-year revenue growth in our super scaled cohort despite a small decrease in super scaled customer count due to the oscillation around the 1 million scaled versus super scaled cutoff point." If one customer goes from spending $1.1m to $900k, this will increase average revenue per customer for both the $1m+ category (they lost a below-average customer) and for the sub-$1m category (they gained one).

2. The net present value of a stream of cash flows at a know growth rate is just (annual cash flows) / (discount rate - growth rate), so in this case it’s ($150k)/(0.12 - 0.03). One corollary, well-known among growth investors, is that the closer the growth rate gets to the discount rate, the closer the valuation gets to infinity. And that also means the valuation of a long-term grower is highly sensitive to any minor shift in that growth rate. At that same 12% discount rate, an 11% growth rate implies a fair value of 100x earnings, and a 10% growth rate means a fair value of 50x, so a 1 point deceleration in growth leads to a 50% haircut in valuation.

3. Which increases the number of plausible locations—every time a restaurant reduces its cost of new locations it increases the number of places that can support a location.

Read More in The Diff

The Diff frequently uses unit economics as a model for understanding a variety of businesses. Some examples include:

• This Twilio profile ($) from 2021.

• A Squarespace S-1 teardown ($) from earlier that year.

• A profile of LocalKitchens, a unit economics story at the location and square-footage level.

• Snowflake has gobsmacking customer retention metrics; updating this post, a customer there who spent $100k annualized in mid-2018 would, on average, be spending $1.1m annualized today—and this average includes customers who stopped using the product or went out of business!

• This recent piece on referrals ($).

And Monday's free post will be a look at the complexities of choosing the right unit economics, in an industry that arguably has half a dozen different ways to think about the natural increment of growth. Subscribe here to read it.

Help us learn more about you!

We'd like your feedback so we know what to cover, so if you haven't already, please fill out this one-minute form.