Rethink pricing before AI forces your hand.
As a CFO, one of the hardest challenges I’ve faced is connecting revenue to the actual value our product delivers.
Legacy billing systems weren’t built for usage-based or AI-driven models, and I’ve felt that friction firsthand.
Seriously… I’ve patched together some SKETCHY homegrown billing systems in my day.
That’s why I lean on Metronome’s Monetization Operating Model guide. It’s a practical framework to align finance, product, GTM, and engineering around pricing, packaging, and billing that scale. Whether you’re managing hybrid models or preparing for AI disruption, this guide offers concrete templates and lessons to get it right.
Trying to build a 5 year forecast on 5 months of customer data.
How Forecasting Changes as You Scale
Understanding why your forecast is wrong is one thing. But knowing how much effort to put into forecasting at different stages? Quite the conundrum.
Realizing that the amount of time in doesn’t always equal the level of accuracy you get out is where many finance teams get stuck.
I spoke to Kevin Drost, former finance leader at Reverb, which he sold to Etsy, about the shift in forecasting rigor as companies mature.
This is the story about the beautiful model that nobody really wanted.
The Beautiful Model That Nobody Wanted
Kevin learned this lesson the hard way early in his time at Reverb, a marketplace for second hand musical instruments.
Fresh off an MBA and three years of management consulting, he decided to build the most rigorous forecasting model he could conjure. He spent two weeks incorporating cohort performance, decay rates over time, CAC, LTV, and every variable he could think of.
"I sat down with the CEO and was walking him through it and he's like, 'Oh yeah, this is fabulous. Now what happens if we put more money into Germany?'"
Kevin paused.
"Well, I don't know. I didn't build that into the model."
"Well, do it."
"Well, I can't. I can't edit this thing in real-time. I need to go back and take another week and build that in as a variable."
The CEO got super frustrated.
"This is a waste of time.
I don't want to see this big beautiful model. I want something that I can play with in real time and get a sense for what's going to be the impact of the business."
Kevin’s boss when he looked at the model.
That’s when Kevin began to hone his skills in what he calls "cowboy forecasting."
Cowboy Forecasting: Two Minutes to Build a Forecast
Kevin, pictured here as a young Clint Eastwood
"He's like, 'you have two minutes to build a forecast because I say I want to grow about 200% next year… what does that mean for my business?
How many people do I have to hire?
How many transactions is that going to be?'"
The logic was simple: If you feel relatively confident about transaction volume and average purchase value, that's your revenue. Now work it down to what your cash flow looks like. That’s what you can use to hire.
"That turned out to be way more valuable use of time than a three statement integrated financial model that our VC loved… but is relatively useless in operating the company day to day.
The “aha moment” was a single-page set of key metrics from your P&L and balance sheet, all integrated and able to work with in real time. That would be way, way more valuable.
(Yes, I’m fully aware that in the GAAP accounting world Revenue and Cash Burn don’t live on the same statement, but they do in startup world.)
Kevin added,
"And then you can spend those three weeks you would've spent building it actually acquiring a customer or building product features, or doing shit that actually builds the business."
Simplicity = Understanding
Ari was the type of guy to hide the most sensitive variable in the model on tab 32 because he didn’t really understand it
Kevin also realized something profound about simplicity and models:
"If you're able to build something simple, it demonstrates that you're able to take something complex and get it down to something that's impactful.
If you are able to use 18 to 20 tabs, it gives you a lot of surface area to spread out stuff you don't understand.
If you can get it down to one page, it's like you own it."
Forcing that simplicity also makes you prioritize what's actually impactful for the business. For a marketplace like Reverb, the growth drivers are marketing spend, return on that marketing spend, number of customers, customer acquisition costs, and product development costs.
"You don't have inventory, you don't have a warehouse, and you don't have a lot of other cost structures or things that drive customer acquisition and retention.
It's going to be your marketing spend and the engineers that you're hiring, and the return on that product performance.”
Early Stage = Goal-Setting, Not Prediction
Kevin realized that in the early stages of a business, a forecast is a goal-setting tool, not an accurate prediction of what the business is going to do.
"In the early days, it's less 'this is set in stone' and more 'what's the hypothesis of what the goal is going to be.'
You are making guesses.
And the earlier you are as a company, the more those are just pure guesses."
The reality is your dead simple model is going to be wrong. But you use it because you know it's going to be wrong. But you can toggle it to get closer and closer as the business moves in real time.
When Rigor Actually Matters
But Kevin outgrew that simple model. By the time Reverb was acquired by Etsy, a publicly traded company, and Kevin became CFO of the business unit, everything changed.
"You have to develop a set of projections that are dead accurate and able to be given to the street. The level of rigor and analysis going into that is very different than Series A scrappy startup trying to just really focus on building the business."
At that stage, Reverb was using sophisticated analytics to determine every customer's behavior, acquisition costs spread across channels, and marketing mix. They could project not just the return on marketing investment, but the marginal return on the next incremental dollar spent.
"The trick is spending every single dollar humanly possible up until you're at that one-to-one or slightly higher return ratio and then stopping. If you're spending less than that, you're leaving money on the table. If you're spending more than that, you're burning cash."
To get to that stage requires a huge dataset and enough analytical rigor to have real accuracy.
But (and this is critical) that sophistication is only valuable when you have the data to support it.
The Danger of False Rigor
But here's the trap Kevin warns about: Don't manufacture precision when you don't have the data to support it.
"If we tried doing that at year one of Reverb, it would be a completely wasted effort because you'd manufacture all this kind of false rigor into things.
You'd say, 'Well, because I have all these data points, I feel really confident about my forecast,' but all those inputs are either made up or statistically just not significant. It's a trap."
Kevin once built a model that could "do everything but cook you breakfast in the morning." But if you toggled the take rate by 2%, the entire thing would blow up.
"Okay, good to know that that's a really sensitive input into this model, but if I am trying to say what am I going to do a year from now on any level of confidence, the confidence level is 0%. So don't waste time on it, waste time on negotiating a better rate on your payment processing or improving your conversion rate on your marketing spend."
The 80/20 Principle
Kevin's a big fan of the 80/20 principle, particularly in the early days.
"The earlier you are in a company, the less benefit you're going to have by having that extra 20%. So it's like, well, what gets me close enough, and what assumptions and estimates can I make in the spirit of back-of-the-envelope math?"
What's the incremental benefit of going from 80% to 85% accuracy? It's usually almost nothing. So forget it, go with that and move on from it.
The goal isn't perfection. It's about decision making. The more you can quantify the elements that go into your decision, the more unbiased your decision-making process can be.
Sidebar: Building Estimates with No Data
So what do you do when there literally is no data?
Every investor wanted to know: How big is the used instrument marketplace? What's the TAM?
"Absolutely no data in the world will tell you how many used guitars were sold at garage sales in the world in the last year. But you still have to have some sense of like, all right, is this a $10 billion industry or $100 billion industry?"
Kevin's approach was to find analogs and work backwards:
Decent data existed for brand-new instrument sales (~$10 billion)
Reverb had great relationships with guitar stores across America
They asked: "How many brand new instruments do you sell and how many used instruments do you sell?"
The ratio came out roughly one-to-one
They extrapolated: If a store does $1 million in new sales, they likely do $1 million in used sales
Bottom-up from there to estimate the total market
"My approach is: What do we have some insight on? What can we extrapolate into what we're trying to understand? And then at a certain point, when does it make sense to say this is good enough, this is 80% of the way there? What's the incremental benefit of going from 80% to 85%? It's usually almost nothing. So forget it, go with that and move on from it."
This is classic consulting case study thinking (not three statement modeling, which he originally pulled out of his toolkit, in error). You're not trying to be perfectly right. You're trying to use logical assumptions to get directionally correct, then validate those assumptions by talking to customers and partners.
Building Intuition Through Experience
One final insight from Kevin that stuck with me: You can't forecast well if you don't deeply understand your customer.
At Reverb, every new hire received $1,000 each quarter with one mission: Buy and sell as many things as you can on Reverb over the next three months. Whoever had the most money left over and had transacted the most wins.
"We wanted to militantly ensure everybody working at the company could really put themselves in the shoes of somebody using the product. If you haven't bought a guitar online, it's going to be really hard for you to pretend you understand the mindset of a customer.
That's not just for a marketer or somebody in product management. It's just as critical for the accounts receivable person or the lawyer.
You have to know what your customers want and the only way you can do that is by being a customer."
The Jump Finance Leaders Need to Make
Kevin's journey from building complex models nobody wanted to developing the art of “cowboy forecasting” shows the evolution every finance leader needs to make.
Early on: Keep it simple. Build models you can play with in real time. Spend two weeks acquiring customers, not perfecting a forecast that will be wrong anyway. The point is to built a tool, not a “take it to the bank” forecast.
As you scale: Let the data dictate your sophistication. When you're public and giving guidance to the street, that rigor becomes necessary (and finally possible).
But through it all, remember the CEO's frustration with that beautiful two-week model: "I want something I can play with in real time and get a sense for what's going to be the impact of the business."
That's still the North Star… measuring the impact on the business. And that’s valuable for any business size and maturity.
Run the Numbers Podcast
Kevin Drost, a true renaissance man. He’s CFO at legal startup New Era ADR, and former finance leader at Reverb, acquired by Etsy. He joined me on the podcast to talk about his rise from music scout for Sony to tech CFO. On this episode we discuss:
Why simplicity beats complexity in financial modeling
how a hands-on understanding of the product and customer builds real financial confidence
The unit economics of marketplace models, and the mistake of taking on inventory
Why pricing and packaging in the legal industry is broken, particularly when it comes to arbitration
Mostly Growth Podcast
Kyle Poyar and I launched a new podcast. It's called "Mostly Growth"
This week we jam on:
Tips for giving a keynote presentation
The ROI you should expect on conference season
Where startups are actually spending money on AI (courtesy of a new report from16z + Mercury)
How legacy companies are shoving AI down our throats through bundling (we didn’t ask for)
A fascinating story about J. Edgar Hoover and what it teaches us about managing our direct reports with clear directions and expectations
If you’re not sick of me yet… (I’d be)
I’ll be at the Spendflo AI summit on October 15th. It’s at 12 PM EST.
I’ll be hanging out with my friend Sid Sridharan, Spendflo’s Co-founder and CEO, and giving a brutally honest take on:
Agent-washing (get it… like green washing?)
Could a startup credibly run with a ‘one-person finance team’ powered by agents?
Will agents negotiate with each other across functions (treasury, procurement, FP&A) to coordinate outcomes?
Wishing you stern, yet fair, debt covenants,
CJ
