There’s a number from SpaceX that matters more than the trillion-dollar one everyone quoted last week, and almost nobody mentioned it: the cost to build a Raptor engine. From 2019 to now, the Raptor program has seen each version landing at roughly half the cost of the one before it, while producing more thrust.
SpaceX didn’t get to a reusable rocket by inventing one perfect engine. They built an engine, then built it again, and again — version after version, each slightly cheaper to manufacture and slightly more capable than the last — until the cost curve bent into something the rest of the industry couldn’t touch. No single breakthrough. A thousand small improvements to the same repeated thing. That’s compounding, and almost everyone misunderstands it.
Repetition is the operative word
We’re taught “compounding” as a finance concept: interest earning interest, the chart that hooks up and to the right. True, but it buries the word that actually matters — repetition. Compounding only happens to things you do more than once. The interest compounds because the principal sits there period after period, the same dollar getting another turn.
This applies to much more than just money, in business it appears in systems and process. The return appears as how much better and cheaper you get at a thing each time you do it again.
My one rule since the day I started building my firm is to get 1% better every day. It sounds like a fortune cookie. It also sounds trivial — 1% is nothing, you can’t feel it. Easy to say, easy to aim for, impossible to track. The gain on any single iteration is invisible, so people skip it and go hunting for the visible win: the big swing, the new logo. But a big win is a one-time event, that doesn’t compound when that 1% does. The entire discipline is tolerating invisibility and doing the unglamorous improvement whose payoff you won’t see for fifty iterations.
The value of the second deliverable
Here’s what that looked like for me, stripped down to the spreadsheet work: my first client was an early-stage dirt-bike protection company, and the financial model needed to forecast to cash. Given my time in PE building models for investment, I knew Excel, I knew how to read accounting, and I knew what an operating model was supposed to look like. It still took the better part of two weeks, call it thirty hours glued to the screen, to produce something usable. It was crude, oddly colored, and far too complex. But it forecasted cash, and it worked.
Then I got a second client, a startup turnaround. This is the exact moment compounding either happens or doesn’t. The lazy instinct, the billable hours instinct, is to build another bespoke model from scratch. Instead, I used my lessons from the initial iteration, what worked, what didn’t what was hard to update and spent twenty hours building a template: the three statements wired together, the cashflow dynamics, a hiring build-out, a capitalization framework, and a clean slot to drop the specific business model into. Infrastructure. After that, populating a new client’s model took about eight hours. A twenty-hour investment cut two-thirds off a process I’ll run dozens of times.
That’s the value of the second deliverable framed alongside the trap of the snowflake. If every client gets a hand-built, one-off model, you never compound; you just relearn the same lessons at full price, forever. The gains only show up when you let the first version teach the second.
Why I won’t bill by the hour
This has become the exact reason that I refuse to bill by the hour. Stop and look at the incentive. Under an hourly model, every efficiency gain is a pay cut. The template that makes me three times faster makes me three times poorer. So, the rational hourly consultant never builds the template, they’re paid to stay busy, not to get better, and the entire pricing model is at war with compounding. Linear, un-productized growth, billing more hours to more clients without the incentive to gain efficiency.
Retainers invert it. When you’re paid for an outcome, every minute you save is pure margin, and for the first time you’re aligned with your own improvement instead of penalized for it. This, more than anything, is why so much of professional services stays artisanal and unleveraged: the dominant way it’s priced punishes the exact behavior that would let it compound.
The process improvement loop
Concretely, I run the same loop on a spreadsheet that SpaceX runs on its hardware in process improvement. Five steps, borrowed from SpaceX, in order:
Challenge the requirement. For modeling, usually this is false precision. Do we need to know the exact cash balance in October? No. We need the sales target per rep that keeps this hire from bankrupting us. Two different questions, two very different amounts of work. Precision is a cost to interrogate before paying for it.
Delete the step. Building a three-statement model from scratch and then analyzing the business is the wrong process. The right one is: learn the business, drop its unique assumptions into the template, read what falls out. The from-scratch build was a step to delete, not optimize.
Find the design improvement. My templates hide reference columns to the left and top, flexible cells that let me reshape the balance sheet or P&L for a different business model without rebuilding anything. The improvement that quietly makes the next ten jobs easier.
Execute faster. Templated inputs, hotkeys for whoever’s driving the model, closing the gap between a decision and its entry. The unsexy speed work.
Automate, Last. Automation comes at the end because you have to automate the correct process, not enshrine a broken one in code. For me that’s become auto pulling the QuickBooks data into the model the moment the month’s books close and creating metrics that provide directional analysis, so we don’t need to dive deep each month. Automate first and you’ve only made the wrong process faster.
How a freelancer becomes a firm
This is how an independent consultant turns into a firm: not one big leap, but a few hundred 1% improvements stacked on top of each other until the practice can do things the freelancer couldn’t. And it only works if you’re building toward something durable. (That’s the other half of this argument, I wrote a companion piece on duration last week.) Chase the next billable hour or next bespoke job and you’ll never spend the twenty hours on the template; the math of compounding needs a horizon long enough for the marginal gains to matter. A firm is just the accumulated infrastructure of every small improvement you were patient enough to make.
Compounding is unglamorous, and that’s the moat. Anyone can have a good idea. Few do the same boring thing 1% better a thousand times in a row, with no applause along the way. The Raptor wasn’t a eureka; it was discipline applied to a repeated process until the numbers bent. So is a good firm.
I’ll happily admit financial modeling is not rocket science. But the method is identical, and that’s the part worth that you can apply in your business. The rocket and the spreadsheet run on the same algorithm. Most people just aren’t patient enough to run it.
If you’re interested in understanding how a financial model can be useful in your business, or just want to discuss process improvement in more depth, I’d be happy to hop on a zoom, email me to continue the conversation: Duncan@saorsapartners.com.