Module 1 · Foundations: The Recurring-Revenue Business · Chapter 3 of 32
First Principles of Recurring Revenue
ARR is not one number, it is three engines running on different math. Learn to decompose it, reason from first principles instead of last year, and read the shift before the board deck forces it.
9 min read
The growth plan landed on a Tuesday, and underneath the spreadsheet it was one line of reasoning: we hit the number the last two years by adding SDRs, so to hit a bigger number we add more SDRs. I had the same two years of data open on the other monitor. Pipeline per SDR had fallen by almost half across exactly that stretch. The plan was not wrong about the past. It was wrong about why the past had happened, and it was about to spend a year of hiring budget proving it.
That is reasoning by analogy: copy what worked, skip the question of whether the thing that made it work is still true. It is fast, it is usually right, and it fails at the worst possible moment, which is any moment the ground moves. Recurring-revenue businesses move the ground constantly, so they punish the analogy habit harder than most. The antidote is first-principles thinking, which Revenue Architecture borrows from sabermetrics: throw out the received wisdom, measure what actually predicts the outcome, and rebuild the plan from what is empirically true today. The Recurring-Revenue Factory chapter ended by promising to take the system down to its first principles. This is that teardown.
| Reasoning by analogy | Reasoning from first principles | |
|---|---|---|
| Starting point | What worked before, copied forward | What is measurably true right now, rebuilt from the ground |
| Pipeline coverage | We have always run 3x pipeline | Coverage = 1 / win rate; this motion wins 1 in 5, so it needs 5x |
| When outbound slows | Send more email, volume fixed it last time | Reply rates collapsed; more volume now makes the math worse |
| A falling growth rate | The team is underperforming, go fix the people | A constant dollar add is a shrinking percentage, so re-point the engine |
| Breaks when | The ground shifts, and in recurring revenue it always shifts | Rarely, because it does not assume the old cause still holds |
The pipeline-coverage row is the cheapest analogy trap to fall into and the most expensive to keep. An ownership-era deal converted about one in three, so “3x pipeline” was sound. A subscription deal that wins one in five needs 5x, and running the old ratio leaves you a fifth short every quarter without anyone able to say why. That single wrong constant then propagates into the forecast, into headcount, into the board number. The reference version of that math lives in the forecasting topic; the point here is only that a copied number carries a hidden assumption, and recurring revenue keeps breaking the assumption.
The number is three numbers
The factory chapter established that a subscription business produces a stream, not a sale. First principles tell you the stream has three sources, and Revenue Architecture writes them as a sum:
ARR = Acquisition + Retention + Expansion. New logos, revenue kept from the existing base, and revenue added inside that base. Three engines, three separate growth curves, one headline number that hides which engine is actually carrying the load. The load moves as the company grows. Acquisition dominates the first few years. Somewhere around $10M ARR retention overtakes it. Around $100M, expansion overtakes acquisition too. None of that is a strategy choice; it is what the arithmetic does to a compounding base, and re-pointing an operation onto the engine that now drives the next dollar takes roughly two years, so you have to see it coming before the number makes it obvious.
View as table
| Stage | Value |
|---|---|
| Y1 | 2 |
| Y2 | 3 |
| Y3 | 4 |
| Y4 | 5 |
| Y5 | 5.5 |
| Y6 | 5.5 |
| Y7 | 5 |
| Y8 | 4.5 |
Watch acquisition flatten while retention keeps bending upward. That is the whole game in one picture. A leadership team still running the acquisition-first playbook that worked at $5M is optimizing the smallest of its three engines at $80M and calling the shortfall a sales problem. It is not a sales problem. It is a decomposition problem, and the scoreboard that only counts new ARR cannot even see it.
The five first principles
Revenue Architecture reduces the recurring-revenue business to five statements that stay true regardless of your product, market, or motion. Reason from these and you stop importing playbooks that quietly expired.
One: growth rate naturally decreases. Add the same $2M in new ARR every year and, even at perfect retention, the growth rate falls: 100%, then 50%, then 33%, then 25%. Nothing broke. The denominator got bigger. Operators who never internalize this read ordinary maturation as failure and reach for the people lever to fix an arithmetic fact.
Two: retention and expansion power growth past $10M, then $100M. Acquisition leads for the first few years. Retention passes it near $10M ARR; expansion passes it near $100M, often about eight years after the first million. Each stream rides its own curve, which is why you have to manage them as three engines and not one.
Three: the effects of recurring revenue take time. Profit compounds over years, and the costs are wildly uneven. The book models acquisition at roughly 100% of first-year revenue (year one nets nothing), retention near 10% of what you keep, and expansion near 20% of what you add. Those numbers are illustrative, but the ratio is the lesson: keeping and growing a customer is cheap, winning one is not.
Four: recurring revenue shifts risk from the buyer to the seller. In the ownership world the buyer bought the thing and owned the risk of getting value from it. Now the customer can leave the moment the impact stops, so the seller carries the delivery risk. That one shift is what pushed the market toward quality-over-quantity outreach and toward product-led motions, where the product proves itself before anyone signs.
Five: recurring revenue is the result of recurring impact. Customers keep paying only while they keep receiving what you promised. The factory chapter planted the distinction: value is the promise that wins the deal, impact is the fulfillment that renews it. No impact, no renewal, and because principle two makes the base most of the business at scale, delivered customer outcomes are an operations concern, not a courtesy the CS team extends.
Build the decomposition trendline
The five principles are only useful if you can see them in your own numbers, and you can. Revenue Architecture gives a specific method for it, and it is the single most useful reporting habit in this whole module. Build it once and every conversation about “why is growth slowing” gets shorter.
- 1
Split every period into three buckets
For each period, break new ARR into Acquisition (new logos), Retention (recurring revenue held from the existing base), and Expansion (upsell plus cross-sell into that base). One row per period, three columns.
- 2
Pull 8 to 10 periods
Fewer and the shape is noise. A trend needs enough points to be a trend and not two dots and a hope. Quarters or years both work, as long as the motion mix is consistent across them.
- 3
Plot side by side, never stacked
Stacking hides which engine is moving. Side-by-side columns let each stream show its own trajectory, so the crossover is visible instead of buried inside a total.
- 4
Fit a polynomial trendline and read the R²
Fit a second- or third-degree polynomial to each stream and check the fit. Aim for R² near 0.98. A clean fit means the stream behaves like a system you can forecast; a messy one means you are still mixing motions or measuring noise.
- 5
Read the crossovers as operating signals
Where retention passes acquisition, the business has already changed shape. Your coverage model, comp plan, and where you spend the next dollar should change with it, not two years later when the board deck notices.
The decomposition is not a slide, it is instrumentation. It wants three fields wired cleanly at the source (new-logo ARR, retained ARR, expansion ARR), which is a data-model job the GTM Engineer Academy covers on the build side and that Module 2 of this course sets up when we draw the Bowtie. Once those fields are trustworthy, the trendline updates itself and the R² tells you whether to believe it.
None of this is exotic math. It is the refusal to run a compounding, three-engine business on a single gross number and a copied ratio. Hold onto the decomposition and the five principles, because everything downstream is built on them. Next, in Models, Systems, and Loops, we make the toolkit explicit: how operators use models instead of raw data, and how to tell an open loop that runs away from a closed loop that stabilizes itself, which is exactly the difference between the left and right sides of the number you just learned to split.