In general, the company that ends up owning the customer relationship is the one that expects to make recurring sales. If there's a product that's purchased intermittently, like snacks, socks, or toothpaste, it rarely makes sense for the manufacturer to somehow have a direct way to reach their customer. It does, however, make sense for Walmart or Target to do this, and to think hard about how they can get that customer to come back a little more often, and buy a bit more when they do. Most software products are distributed directly to the customer, and within that business there are wildly different ways to think about the value of those customers, mostly driven by who they are. As a general rule, outside of usage-based pricing, the larger the initial purchase, the more likely it is that a given customer cohort will spend more year after year, even accounting for the ones who leave.

• At the low end, there are subscription products sold to customers and small businesses. These tend to have a finite life: every month, some fraction of Netflix and Spotify users will either stop streaming or stream from somewhere else, and each month, some fraction of GoDaddy or Cimpress's users either go out of business or concede that they didn't really have a business in the first place.

• In the category of software contracts where phone/Zoom contact is almost mandatory and an in-person meeting with a sales rep isn't out of the question, retention is a bit higher, and there tends to be some room for price discrimination: if these products are hard to onboard, then something that was a $20,000 purchase the first year is probably a better bargain at $22k the second year, just because it's such a pain to move over to an alternative.

• And at higher levels of spend, what's being sold is often a bundle of products and services—Palantir can afford to have someone spend a while on-site with customers, and Bloomberg is happy to offer an overkill level of customer service for its help desk.

At each of these tiers, the customer relationship is a bit different. For low-priced subscriptions, whether they're sold to consumers or small businesses, there's just limited scope to raise prices very much. It's close to the classic micropayments problem in that the cognitive overhead of deciding whether or not to spend, say, $0.40 to read an article is more costly than the $0.40. An all-you-can-eat subscription is partly a way to pay not to think so hard about individual instances of consumption. But that same math comes into play when prices change: most people don't want to calculate whether something that they liked at $10/month is still a good deal at $11/month, so what companies generally have to do in this space is set their initial price so that it's an astoundingly good deal—all the music you want, a default TV/movie option every night, the feeling that your business is a Real Business because it has a website, etc.—all at a low price. The best companies with this model have been able to raise subscription prices over time, in part because the customers they're losing are the ones who were probably going to cancel anyway. If $2/month makes the difference between whether or not you were going to stay subscribed to Netflix, it's not really about the $2, just a reminder that you weren't watching all that much in the first place. (Or that you feel guilty about how much you were watching; Google search volume for "cancel Netflix" usually has a small uptick right around the start of the year, though it's not as visible as the more explicitly self improvement-oriented searches.)

At higher price points, there are two levers for growth: first, when prices are high it makes sense to have more continuous customer feedback, which means the features these companies launch are directly related to what their customers are asking for. It's usually a good idea to pair a price increase with some kind of demonstration that the value proposition has improved. These products also get priced differently because their customers are typically doing some kind of seat-based pricing, and a useful tool a) tends to warrant more seats, b) will hopefully increase growth at companies that use it, thus increasing the number of seats that can be sold, and c) will often find ways to encourage the company to add more peripheral users: if you're selling a tool that enables some kind of collaboration, the more departments that can plausibly collaborate with it, the more seats you can sell.

And at even higher price points, vendors like to describe themselves as "partners," and that's true in many ways: they're making enough money from the customer relationship that they need to pay close attention exactly what that customer wants, they're not just enhancing existing products but adding new ones, they may give customers visibility into their roadmap (and vice-versa)—and, like a true partner, their revenues are going to be directly connected to the overall economic upside that they generate. If a large company has a seven- or eight-figure contract, it's because they're buying something mission-critical from the seller best positioned to offer whatever the service in question is. And that seller has a lot of latitude for price discrimination, but also knows that every year, they're deciding how much trust to monetize through immediate revenue and how much to monetize through lower churn and a longer growth runway.

There really isn't a best or worst version of these. In general, larger contract sizes have a hard time delivering high gross margins, because they'll have ongoing sales, implementation, and customer service costs. On the other hand, those relationships can be incredibly sticky; there are sometimes technologies that continue to exist and be maintained because there's literally one company left that uses them, and moving off is never a priority. Higher-churn products generally assume that there will be few to no direct customer interactions (and the few that used to exist are the sorts of things that AI chatbots handle well). It's a pain to know that your business will only work if you keep paying to acquire new customers, but at scale you know roughly what you're paying for and what you're getting—if the LTV is three figures rather than eight, the CAC/LTV calculation can have more significant figures. So this ultimately ends up being a tradeoff between two kinds of revenue quality: large contracts tend to get larger over time, but the details are uncertain; small contracts trail off to zero sooner or later, but at scale, you can precisely estimate when.

Read More in The Diff

In The Diff, we often talk about these CAC and cohort dynamics, at a variety of scales and in many different contexts. For example:

• Companies like Box and Dropbox use a kind of K-nearest-neighbor approach to adding features ($), which is why two companies that solve the same problem for two different kinds of customers can evolve different sets of features.

• We looked at DoorDash’s retention-based economics around the time of their IPO ($).

• Snowflake is a great case study in big contracts getting bigger.

When SaaS companies sell to other SaaS companies, the entire sector’s growth becomes a reflexive phenomenon ($).

Join the discussion!