As a CFO, I've always hated how much of my day is choosing between control and speed.

You either get real-time visibility or you get your team moving fast — never both.

My Brex card is different. Brex gives me the controls I need without slowing down important expenses, like Wally’s dog treats or Producer Ben’s cool gizmo sound things. With Brex, receipts are automated, expenses hit in real time, and categorization doesn't require three people and a spreadsheet. I can see exactly where money's going while we focus on growing the Mostly Metrics world empire (also, Sales Guy Matthew, just because you CAN expense and eat $32 of Chipotle doesn’t mean you HAVE TO).

Finance doesn’t have to slow companies down - it can and should push them forward.

See why 35,000+ global companies (and me!) use Brex to spend smarter and move faster:

Hey it’s CJ, CFO turned recruiter. Do you need to add someone new to your team before annual planning season starts?

I place finance and accounting hires at the manager, director, and VP levels.

The first time I built a comp plan for a sales team, it ran like a vending machine. Rep drops in a signed contract, out pops a commission check. Shazam!

It was very much price times quantity driven. And the only thing that mattered was getting that signature.

Consumption pricing broke my vending machine. The customer signs a commitment, and then the rep (and CFO) spends the next twelve months sweating whether they ever turn it on. In this model, commission and revenue recognition both follow usage. It's very possible to sign a monster deal in March and still be waiting on most of your check (and rev rec) in December because the customer hasn't gotten around to using what they committed to.

A while back I wrote about how Snowflake forecasts a consumption business. We went deep on their daily, account level forecasting models and ability to see the future ten years out. This was very much the finance view of the usage based model. I wanted to augment that view from the sales compensation and planning angle, so I called Ron Gabrisko.

Ron has been CRO of Databricks for ten and a half years. Yes. For that long.

For context, the average CRO gets about 18 months before everyone agrees it isn't working out. He joined in December 2015, when Databricks had 40-something employees and was scratching out $12K and $18K deals with Silicon Valley startups.

He reminisced on the early years, before they had an enterprise motion, which was the biggest unlock to their explosive growth:

The company is generating north of $5 billion in revenue, growing 65% y/y. And they’re valued at $134 billion, with rumors of a pre-IPO +$175 billion private round swirling.

Yes, Databricks is a brilliant AI product. It's also a brilliantly engineered GTM machine that has tuned the incentive levers like a Ferrari. Because how a company pays its reps tells you what business it's actually in, and reveals how it delivers value to its end consumers. Here's what Ron shared on DB’s model that's worth stealing.

Those sub $25K startup deals were going to take a long time to get to $1 billion in revenue. The founders had already promised investors they'd get to $10M ARR in the near future, and Ron ran the math to quickly see it was going to be a total slog. So he convinced them to fund a bet that cost money up front and would take a while to ramp. He asked them to believe him.

He hired a bunch of expensive, experienced enterprise sellers and pointed them at the banks, the retailers, and the big healthcare systems who could cut a big check. The founders weren't completely sold, but went along with it. Their worry was that enterprise reps wouldn't get the product, which was fair, since it was complicated and technical.

Ron's answer was to send those reps to talk to everyone already running Apache Spark (the opensource code DB was built on top of), find out what hurt, and report back on what they could actually sell them. A lot of the early product and positioning got shaped by that field research.

This is the part of the journey a CFO has to be willing to stomach. Enterprise is a different motion, with 9, 10, 11, and 12 month sales cycles, so you're pre paying senior-rep salaries against a pipeline that closes nothing for the better part of a year. It looks like a waste until it begins to take off. And it takes real conviction from the top to keep cutting those checks while the bank account says you're underwater.

It broke their way in year one with two multi-million dollar deals, both sourced through their investor a16z. That was the moment Ron knew the bet was right:

The long sales cycles that felt so expensive to fund are the same reason the base is durable now. Enterprises are brutal and slow to land, but once they're on the platform they adopt hard and keep expanding for years, and a giant company standing behind your tiny one serves as credibility that helps you muscle into the next big deal.

At my old seat based SaaS company, the land was a religious event. A rep would close a fat first-year contract, bang the gong we kept in the front of the room (super loud), and everyone would hoot and holler. Then the customer got handed off to Customer Success like a baby at a christening, and the rep moved on and never thought about the account until two months before renewal.

This all made sense at the time, since the contract guaranteed the revenue. A booking turned into a billing and we got to log the revenue equally across the calendar, regardless of how many people actually logged in.

Now, don't get me wrong, we obviously wanted them to log in. Because inactivity usually leads to churn, and churn is bad for enterprise value. But, you get the point. It was more or less a wam bam process.

Databricks runs it the opposite way. Their reps try to land small and land fast, like a $50K or $100K deal on a single use case (that’s small to them in the Enterprise). The goal of that first contract is very modest: just prove one workload runs, and build a relationship with somebody inside the account who'll vouch for the next deal.

Unlike the SaaS deals I was describing before, they are NOT trying to maximize sticker price on entry. It's more like they are running a paid, long term, POC.

And they grease the skids so the rep isn't selling a science project from scratch. Databricks pre-packages the common lands and bundles in services to get the customer live quick, which Ron calls "fast starts."

At a high level, spending services dollars on a tiny deal sounds like a great way to torch money. But you have to remember that nobody in a usage model gets paid until the customer actually uses the thing. So there is "mutually assured destruction" in a way that doesn't exist in the typical SaaS scenario. A customer live in week three is throwing off consumption, commission, and rev rec.

That's why the size of the first deal barely registers for them. Ron watched $100K lands grow into million-dollar deals inside the first year, then into tens of millions in committed spend down the road. Every step was riding on that first proof point and that first champion. If you give a mouse a cookie.

So as the finance team at a usage based company, when a rep comes to you all lit up about a massive first-year commitment, the question you've really got to ask is how fast it turns on.

The deepest change Ron made wasn't the comp percentage, but where the quota attaches to.

In the SaaS world I came up in, quota was a personal stat, like batting average for a baseball player. Hey, you're an experienced enterprise rep, here's your $2.5M number, go get it however you want, and we'll pay you extra if you can exceed it. The number was attached to the human, in both my deterministic model and in the sales team's management philosophy.

Databricks moved quota onto the account and tied it to incremental spend, so the rep carries the growth of consumption inside specific accounts rather than a personal bag they fill from anywhere. When the company divvies up billions in quota, it's slicing growth account by account, with the target set by what's actually going on inside each one.

He got there in two steps. First, they started paying on committed contracts. Then they went half commit, half consumption. Then they went all the way to consumption, with commits paid on top but usage as the thing that drives the ultimate check.

This move crystallized customer alignment:

If the rep only eats when the customer consumes, the rep stops caring about stuffing the contract and starts caring about adoption, which means really understanding customer painpoints that need to get solved to expand the account. It's impossible to sell vaporware.

So that's the math part. But they also had to account for how reps think and the habits they may be engrained with. Salespeople come up worshipping the close, legendary December 31st hero deals at 11:46PM, and the big commission check that pays off the home mortgage.

Now try telling them that money drips in over months as a customer ramps. In Ron's words, he was talking a lot of AEs off a cliff. He had to explain to them that the ceiling didn't move, just the path to it. He had to reassure them they could still be a millionaire rep at Databricks, you just got there by making a customer consume a ton, rather than by nailing one monster booking on the last day of the quarter.

This reframe only works if finance can accurately forecast the account's ramp, which is the whole reason the Snowflake-style daily modeling I previously wrote about matters.

This made sales reps just as interested in the forecast as FP&A. In many ways it promoted a data driven culture throughout the org, not just in finance.

A few years back Databricks split the sales force in two: Hunters who go win new accounts, and Core reps who grow the ones already on the platform. Before that, territories were mixed, and a rep owned both jobs at once.

Snowflake, the canonical consumption company and Databricks' closest comp, runs a similar set up. They staff dedicated "Acquisition" AEs whose whole job is landing net-new logos, sitting in a separate org from the reps who carry a book of existing accounts to grow. Their early CRO put reps on aggressive new-logo quotas and hired specifically for people who could open doors, not just farm what was already there. So two of the most successful usage businesses on earth landed on the same org chart, independently. It makes a lot of sense why the pricing model pushed them there (remember: staffing models are downstream, not upstream from pricing decisions).

Mixed territories fail in a usage model in a specific, predictable (and kinda obvious) way. If you give a rep new-logo targets plus a book of existing accounts, they're going to gravitate toward whichever accounts are ripping and park their incremental effort there. Why grind through a cold POC and usher a customer through a small deal to ramp, when you've got a live one compounding right in front of you? Ron saw it even with his best people:

And you need to activate those five or ten other accounts in the rep's book because those are the ones that will feed you in three years.

So the two motions got their own people and their own comp.

Hunters get paid on lands, with bounties rich enough to make chasing a brand-new logo worth it, because Databricks knows a chunk of those lands turn into ten and twenty year accounts.

Core reps run the call-up motion: more use cases, more lines of business, more champions inside an account that's already consuming. They know where the bodies are buried when it comes to workloads. They navigate the org chart like water.

And the bounty does something useful beyond rewarding the land. When a Hunter gets paid well for landing, they stop trying to cram an oversized commitment into a brand-new account that hasn't proven anything yet, because the payout doesn't ride on the size of that first deal. And to keep Hunters from grabbing a bounty and running, the account still has to actually consume before it converts and rolls to a Core rep.

For finance, this is at the heart of staffing and comp. The split you come up with tells you how many of each role you're funding, what your business model can support, and what a land is worth versus what expansion is worth. It also tells you where the next two years of consumption is coming from. If you get the ratio wrong, you either run out of new accounts to grow, or you pile up Hunters with nothing left to hunt. It's a balancing act.

If you grew up in SaaS finance, December is your Super Bowl. The quarter and the year close on the same day, and a third of the number can show up in the last two weeks as everybody scrambles to get deals signed before the ball drops. I've watched a forecast swing from 85% achievement to 95% achievement in 24 hours because of it.

In a consumption business, December is a low point from a revenue perspective. People are on vacation, fewer people are in the office running workloads, and usage just sags. The revenue follows the humans, and over the holidays the humans are at their in-laws' arguing about politics, rather than querying data. Same thing happens in the summer when half the company is at the beach.

Yet none of this surprises Databricks. They've seen enough cycles that the dips are baked into the model. Ron can call his consumption number within a point or two using data science, holiday slumps included, because usage is just behavior, and behavior repeats. From his perspective, the vacation trough isn't a risk to manage, just a line item they continually predict.

Consumption also solves something that keeps bookings-based CFOs up at night: concentration. Ron mentioned, almost in passing, that he's never had a single deal worth 10% of revenue. Not one. With thousands of accounts all ramping at their own pace, no individual logo and no single quarter-end signature can punch a hole in the quarter if it slips. You stop sweating the one whale that has to close by the 30th. Many in the SaaS world wish this were the case.

There's a discipline carryover here that sales-side readers will love (it was new to me as a finance person). Databricks still runs MEDDPICC, the old enterprise qualification checklist, on its commit opportunities. The difference is they aim it at use cases instead of deals:

It’s the same rigor they always had, pointed at consumption now, which is part of how Ron can call the number so tightly.

At the same time, the quarter-end deal doesn't completely disappear. Databricks still signs big three-to-five-year commitments at the end of a quarter. That's when procurement teams are trained to act. The difference is they don't sit around waiting for the calendar to create urgency. They manufacture compelling events all quarter long by driving usage, so a customer who's ramping burns through their current commitment in May and signs the bigger deal then, not because it's the last day of June. So they have the ability (and motivation) to do big deals during and at the end of quarters.

Ron still recognizes:

But in a usage model it doesn't necessarily mean Santa is bringing cold hard GAAP revenue. Yes, they do big deals at the end of the quarter. But they are commitments to a floor. You can land a monster three-year commitment at quarter-end and recognize almost nothing from it that quarter, because the revenue only shows up as the customer actually consumes against it. The booking and the revenue have been pulled apart, which is exactly why the consumption forecast, not the commitment, is what you run the business on.

The sales success Databricks experienced started with a move to the enterprise. And it evolved its staffing model and comp structures to best support those customers, who wanted to migrate workloads over and pay as they consumed.

While you aren't Databricks, and don't have 5,000 reps and a daily ML model for account level forecasting, the principle holds at any size: before you start talking about accelerators or payouts, you need to understand account level activity.

Which is really a way of saying you start from the back. What will this account actually consume, and what has to be true to get it there. You build the comp plan, the staffing, and the fast starts off that number, because that number is the revenue.

And that means finance can't hand over a forecast and clock out. You own the ramp as much as the rep who signed the deal.

If you want to be on the rocket ship, you have to have true belief in your model’s unit economics so you can appropriately back the time to ramp. Because change is the only thing that’s constant, especially if you are compensated on that change.

In a consumption model, sales isn't the only one carrying a quota.

Trying to measure ROI on tokens? Me too.

So I talk with Rogo President Rahul Rekhi to unpack
token economics, why adoption without ROI is a trap, how vertical AI wins, and why domain expertise still matters.

Plus: forward-deployed bankers, outcome-based pricing, and what CFOs should measure
before AI spend gets out of hand.

Hoping your enterprise sales cycle has a faster than modeled ramp

CJ