Navigating Consumption-Based Pricing with Mike Scarpelli, CFO of Snowflake

Back when I started out in venture in 2015, companies like Workday and Zendesk were changing the game of enterprise sales with their subscription pricing models. Now, we find ourselves talking with many of our Lightspeed portfolio companies about whether or not they should incorporate usage-based pricing.

With usage-based pricing, a customer’s cost is directly proportional to their consumption of the product or service. A number of recent public companies — Snowflake, Twilio, Agora, JFrog — have successfully leveraged this pricing model.

Given the general interest in this subject, I’ll be sharing a series of interviews over the next several weeks with executives in the Lightspeed network who have scaled usage-based companies, starting with Mike Scarpelli, CFO of Snowflake.

Last year, Snowflake had the largest software IPO in history. One part of their appeal to customers is that Snowflake only charges for the volume of data stored and compute resources consumed. The company grew revenue by 124% and posted a 168% net revenue retention (!). This is hands-down best in class when compared to any public software company today.

Mike Scarpelli, CFO of Snowflake

Anoushka: Mike, you have built your career working with iconic software companies including EMC, ServiceNow, and Snowflake and have seen both SaaS and consumption-based pricing. How should a business evaluate usage-based pricing?

Mike: Over time, I believe most software will be delivered in a usage-based model. But it will take some time. As a CFO, what I constantly hear from customers is that they only want to pay for what they are using. No one wants shelfware. Yet today, I negotiate with software vendors all the time who increase deal sizes each year even when our company is not utilizing its seats or licenses.

A usage-based model can work very well for any software that has a lot of transactions and compute. If the software product is an HR tool that people are looking at once a quarter for bonus allocations, then a usage-based model doesn’t make sense. However, there could be many SaaS businesses that can be priced on a consumption basis instead. For example, reps are constantly pinging and updating their Salesforce CRM — this could be a priced on usage-based model. Service Cloud, where customers are logging tickets, could be usage-based. For our customer-facing applications at ServiceNow, customers wanted a usage-based model. However, they did not know how to forecast their growth in usage, making budgeting a challenge.

Anoushka: Is there a “why now” that makes you believe most software will be delivered in a usage-basis over time?

Mike: We now have the ability to really track compute resources per query that any customer is running. If you went back 5–10 years ago, this would be much more challenging. In the days of on-premise software, you were buying the hardware. It did not matter how much compute you were using. You had to do everything on a license or per-seat model.

The cloud vendors have already adopted the model. If you want to compete with the big guys, you will need to evolve your pricing strategy.

Anoushka: What are the challenges of moving to a consumption-based pricing strategy?

Mike: For the business, you need to build an ability to predict your customers’ future consumption to forecast your business on a quarterly and annual basis. At Snowflake, we forecast usage on a customer-by-customer basis and have predictive modeling around future use. There is no out-of-the-box module or product for doing these calculations. We built this system in Snowflake. Our ERP does the invoicing and then the calculation that goes into usage all gets billed within Snowflake. We reforecast revenue daily at Snowflake.

With usage-based pricing, a customer often can’t forecast their own usage growth. Thus, it becomes much more challenging to figure out how to budget for the software. A per-seat license model can be much easier to budget for. Your software needs some type of capability that lets customers model out their growth in consumption

Finally, governance is important. You want to make sure only the people in your customer’s organization who should be using the software are using it. Customers want alerts to see if certain groups’ consumptions are spiking so they can plan around it.

Anoushka: How has this pricing model impacted Snowflake from a team and organizational structure?

Mike: It changes your compensation plan for sales reps. You can’t pay everything on new deals. You have to pay on consumption. Every one of our reps has a big consumption quota. We fundamentally want our reps to be involved in customer success.

We don’t believe companies should have a separate customer success function. The first thing we did when Frank joined Snowflake was we blew up was the customer success function. You are either going to do support, sales or professional services. Customer success is not accountable for anything.

You may hear some pushback from your salespeople and you will also have some portion of your salesforce focused on landing new accounts. However, a bigger percent of your commission dollars are going to be tied to consumption.

In the SaaS world, your FP&A group can develop a forecast in a silo in isolation from salespeople. The days of FP&A doing forecasting in this manner are gone. Our revenue and FP&A team is aligned with sales and our sales reps. The group has a regular cadence review of our top 50 to 100 customers. Our FP&A group spends a lot of time with salespeople to understand every element of our large accounts and the trends happening with consumption.

Anoushka: What are the core KPIs you think about? Have these changed at all given Snowflake’s usage-based model?

Mike: In the SaaS world, billings is the key metric you look at. Snowflake lets customers pay quarterly, monthly or on-demand so we don’t focus on billings at all. We focus on remaining performance obligation — for all your booked business, how much is going to roll up into revenue. It is the best leading indicator for our investors. Naturally, we care about revenue and revenue growth as well. Investors also want to believe there is leverage in your model. It is important to show a path to higher gross margins and pounding that with investors as that continues to grow for us

Anoushka: How have you managed pricing and pricing discounts which becomes very important for scaled customers in a consumption model?

Mike: You target the margins you want to meet. There are different discount bands based on the size of the deal. We spent a lot of time around this. As a new company breaking into new geographies and verticals, you need to do discounting for those customers. As you become mature, customers talk and you need to be very disciplined around your discounting. You need to make sure to treat your customers like for like because they all talk.

We have become much better disciplined around discounting like for like customers. We put in place a sales finance function to ensure discounting discipline and to help customers model their spending based on the discount. We also put in place a deals desk function that can quarterback a deal process to ensure salespeople are not spending their time pinging various internal groups.

Anoushka: Any final thoughts?

Mike: Many startups today espouse the belief in growth at all costs. I totally disagree with that. Growth is important but too many young companies waste money on things that there is no ROI. You need to spend the money wisely.

Lightspeed Possibility grows the deeper you go. Serving bold builders of the future.