A Primer: Subscription vs Usage Based Pricing Models

More upside: “Core to the UBP (Usage Based Pricing) strategy is the idea that you can’t predict your largest accounts. When you implement UBP, you’re making a bunch of bets with the hope that some of them will pay off spectacularly.” (Source: OV). In other words, you aren’t capping your upside, regardless of their initial spend.

Net dollar retention rate is like a video game cheat code: Customers start at a lower level than subscription, and ramp their activity over time as the product becomes a core part of their workflow. This puts the majority of their growth into the expansion bucket, which gets big time net dollar retention credit. If I were writing a book report, I’d cite Snowflake in my bibliography.

You can sell large commitments: It’s easy to tell a story to customers about why they’d want to buy up lot’s of credits at a discount now so they can realize a better rate at scale later. I don’t have any evidence of this, but anecdotally I see more multi year commitments for usage based contracts than subscription. Feel free to argue with me, though.

Easier access: Since you don’t really care about the number of seats or licenses being used, you can make the account accessible to a wider audience within the company. More use cases will inevitably evolve as a result, and when your product becomes ubiquitous, it’s great for revenue.

Easier to scale back during macro headwinds: Dad might come home from work and turn the thermostat down. The same volatility on the way up, also exists on the way down. When times get tough, it’s much easier to “optimize” your consumption spend than it is to get out of an ironclad, annual ass Subscription contract.

Misconfiguration mistakes can frustrate customers: When you set up a usage based product incorrectly, it’s much more catastrophic than the mistakes you can make when assigning subscription licenses to the wrong folks. Many a developer has left the faucet on over night when testing something on AWS. I remember one time an infra developer blew up my GCP budget by like $18K in one day from a small mistake. This has been a common occurrence with new AI models that run on usage based pricing:

More prone to calendar seasonality: As an FP&A analyst, you pick up on weird fluctuations in cost during the year. You’ll notice how cheap it is to run your cloud infrastructure during the holiday season - your hosting bills mysteriously dip. Why? There’s less “usage” or activity or queries or whatever, because everyone is on vacation. I see the same thing with our BI tool usage during holidays.

Easy to forecast: There’s more predictability from period to period. It’s the easiest model to build an ARR waterfall for. You already know what they agreed to pay you. Duh.

Easy to understand: More customers are familiar with how Subscription works. It’s like a gym membership, but for software. See, that was easy.

Churn mitigation: It’s harder for a customer to shrink in-period.

You have to wait longer for a compelling opportunity to upsell: It’s more difficult to increase spend in-period. Subscription companies often circle renewal dates on the calendar to upsell additional products. With usage based pricing, the conversations with sales reps are more frequent, as customers look to optimize the core product and then add-on additional modules. DataDog, who relies predominantly on usage based pricing, is great at this. Peep their multi product attach rate:

Net Dollar Retention has less room to run: Subscription customers usually start at a higher commitment amount than consumption customers, and mathematically have less room to run when it comes to expansion.