Real Talk: with James Jackson & Rachel Parrinello Consumption-Based Comp: What Nobody Tells You Before You Build It

Most companies moving to consumption pricing bolt on a comp plan as an afterthought. Then they wonder why their reps are optimizing for bookings and ignoring usage. In this conversation, James Jackson, VP of RevOps at Canva, and Rachel Parrinello, Sales Compensation Solutions Lead at Alexander Group break down what it actually takes to pay people for a model that never stops selling.

James, you've lived inside a consumption model at Snowflake. What's the thing people don't understand until they're already in it?

James: The sales motion never ends. In traditional SaaS, you close the deal and move on. In consumption, the deal is never really closed. You're constantly identifying new use cases, new workflows, new reasons for the customer to use more. And if your comp plan doesn't reflect that - if reps still get paid like it's a bookings game - they'll play a bookings game. Every time.

Rachel, how different are these models really? People throw "consumption pricing" around like it's one thing.

Rachel: It's not. And getting this wrong at the design stage breaks everything downstream. Committed spend is a customer locking in a million dollars and drawing down against it. Uncommitted is you have a contract and a price, but volume is genuinely unpredictable - common in fintech and adtech where transaction flows swing wildly. Pay-as-you-go is no contract, credit card on file. And hybrid - the most common right now - is a seat license with a consumption layer sitting on top, usually an AI feature or usage-based add-on. Each one of those requires a different comp design. If you're building one plan to cover all of them, you're already in trouble.

So what does a comp plan actually look like when consumption is real?

Rachel: For hybrid companies, I almost always recommend a dual measure plan - an Annual Contract Value (ACV) component for committed contracts and a consumption revenue component for ongoing usage. The question people always ask is when to separate them. ‍

Once consumption hits about 15% of total revenue, it needs its own measure. Below that, double crediting works - you reward the same dollar under both measures. But above that threshold, it gets expensive fast and reps lose focus on what they're actually being paid to grow.

Hunter-farmer - is that the right model for consumption, or does it depend?

James: We went hunter-farmer and I'd do it again. Hunters are 100% bookings. New logos, new committed deals - that's the job. The moment you start layering consumption expectations on top of a hunter role, you've muddied both.

Rachel: Farmers are where the consumption weight belongs. A typical split is 60-70% consumption, 30-40% bookings. Their entire job is to own the relationship post-sale and grow usage. Some companies in land-grab mode add a third metric - new logo count above an adoption threshold, not just a signed contract - to make sure new logos actually activate and don't just sit there.

Measurement is where this gets complicated. Total consumption vs. incremental - where do you land?

Rachel: Incremental, and more companies are moving there. Here's the problem with total consumption: a rep who inherits a massive account doing $3M in usage walks in with an enormous quota they barely had anything to do with. That's not pay for performance - that's luck of the draw. Incremental fixes this. You measure growth above a baseline, like this quarter vs. the same quarter last year, and reps are paid for what they actually drove.

And the baseline can't be set at 100% of last year's usage. Usage moves around for reasons reps can't control. If a customer had a slow quarter, that shouldn't be the rep's problem. We take 70% of the prior period as the starting line - that gap gives reps a realistic shot at earning before they've even made a call.

James: At Snowflake we paid monthly. Take the annual quota, divide by 12, and every dollar above that monthly number retires incremental quota. If a month goes negative - customer dials back usage - we zero it out but don't claw back against future months. You want to reward the trend, not punish the noise.

What does the data infrastructure need to look like to actually run this?

James: Snowflake had over 30 data scientists dedicated to quota setting for 2,700 sellers. We could forecast consumption to the percentage point. That's the deep end of the pool - but it tells you how data-intensive this model really is.

Rachel: Some companies are starting to use machine learning (ML) for quota setting, and the ones that do it well get better over time. One company we worked with had been running ML-based quotas for six years. They started by allowing manager overrides. Eventually they got rid of them - the model was consistently outperforming human judgment and the overrides were just introducing noise.

The real point is: before you build the comp plan, you need to know your data. Clean historical consumption data, real forecasting capability. You can't retrofit this. If that infrastructure isn't there, start there first.

What's the one thing people get wrong in the rollout?

James:

We didn't present the comp plan. We built it with sales leadership in the room. The moment it becomes a top-down delivery, you've already lost trust.

‍

We ran office hours, multiple explanation sessions, and engineered territories for simplicity. If a rep can't explain how their own comp plan works, something has gone wrong.

Rachel: Start with the contract strategy before you touch comp. What are you actually selling? How does revenue get recognized? What's the unit of measurement? Who owns consumption after the deal closes - Account Executives, Customer Success, someone else? Comp you design before answering those questions is just guesswork.

Want to go deeper on consumption comp design? Connect with James Jackson and Rachel Parrinello on LinkedIn, or join the conversation on Uncappd - where comp practitioners share what's actually working.