# How to Build a Growth Model: A Step-by-Step SaaS Guide

> Learn how to build a growth model for your SaaS business: define inputs, build the spreadsheet formula, and use it to pick your next growth bet.

- URL: https://missiongrowth.io/blog/growth-model
- Published: 2026-08-25 · Updated: 2026-09-24
- Author: Ömer Furkan Aktaş, Founder, Mission Growth
- Publisher: Mission Growth. Company facts: https://missiongrowth.io/llms.txt

A growth model earns its keep the moment it tells you which lever to pull first, not the moment it looks good in a slide deck.

Here's how to build a growth model: pick one output metric, map the qualitative version, list 5-8 real inputs, connect them with a formula, then compute a full example before you trust it with a real decision.

In this guide:

- The types of growth models (visual and mathematical) and when to move from one to the other.
- The growth model formula: how to connect acquisition, activation and retention inputs into one number.
- A full saas growth model example, Atlas's, computed step by step, that doubles as a growth model template.
- Which lever to fix first once your own numbers are in.

## What a growth model actually is (and when you don't need one yet)

A growth model is a spreadsheet system of named input variables connected by a formula that predicts how one output metric moves when you change an input.

That output is usually your [North Star Metric](https://missiongrowth.io/blog/north-star-metric), something like MRR or activated users. The model is only worth building once you have something real to calibrate it against.

A [growth loop](https://missiongrowth.io/blog/growth-loops) is one mechanism inside that system that reinforces itself, a referral loop or a content loop, for example. A growth model is bigger: it wires every loop and channel your company runs into a single formula that rolls up into that one output number.

A diagram of that system can still drive a real decision on its own. What it can't do is tell you how much a lever is worth, or let you check that call again once new numbers come in. That's the arithmetic's job, and it's why this guide builds one end to end.

Don't build the spreadsheet yet if you're missing any of these three:

- **A product-market fit signal.** Retention curves have flattened for at least one real segment. Model a moving target and you're just modeling noise.
- **At least one month of real historical funnel data.** The formula needs something to calibrate against, not a guess dressed up as a rate.
- **More than one team or channel competing for the same priority.** If there's only one obvious next move, you don't need a spreadsheet to find it.

Check all three, and pick the one output metric before you touch a single input.

## Map the model before you build the spreadsheet

Before opening a spreadsheet, map the qualitative version first: a simple diagram of how acquisition, product and monetization loops connect to the output metric you picked.

Sketching it costs an afternoon, and it forces you to name the pieces before you start attaching numbers to them.

A visual model and a mathematical model do different jobs:

- **Visual model.** A diagram of how the pieces connect, no numbers attached. Good for aligning a team on the shape of the system.
- **Mathematical model.** The same relationships rebuilt in a spreadsheet with real conversion rates, costs and a formula linking them. Good for sizing a lever and defending a bet.

Move on readiness, not the calendar: once you can name each loop's two input variables and point to where their historical data already lives, add the numbers. If you can't answer both, go find the data first.

One option when you sketch the diagram is to label its stages with the [AARRR framework](https://missiongrowth.io/blog/pirate-metrics) (acquisition, activation, retention, referral, revenue). It's an optional naming convention.

A diagram alone can carry real weight. On Reforge's growth-model gallery, one of ten featured operator write-ups, David Shein's, stays a purely visual diagram, and it's credited with helping the team decide where to prioritize investment. But it's the exception: Quizlet's US and international models, Braid, Lyft, Graphite, Edvoy, AdRoll and Bonsai all end up as numeric spreadsheet models once the team needs to defend a specific bet.

::figure{src="/blog/figures/growth-model-1.svg" alt="A matrix table classifying Reforge's 10 featured operator examples as quantitative or qualitative, showing 9 quantitative and 1 qualitative." caption="Most real operator examples end up quantitative: 9 of Reforge's 10 featured write-ups are numeric spreadsheets, only 1 stops at a visual diagram." width="720" height="541"}

## List the inputs and connect them with a formula

A growth model's spreadsheet needs one line per major acquisition, activation or retention stage, each carrying a rate or a cost, multiplied and added together into the output metric.

Group the lines into categories:

- **Acquisition.** Traffic by channel, cost per lead, channel mix.
- **Activation.** Signup-to-trial rate, trial-to-paid rate.
- **Retention and expansion.** Monthly churn, net revenue retention.
- **Cost.** CAC by channel, payback period.

If one of your acquisition lines is a referral or viral loop, its own multiplier is the [K-factor](https://missiongrowth.io/blog/viral-coefficient), just one input inside that acquisition term.

Before an acquisition channel like organic search has a full month of real conversion data to plug in, forecast its likely contribution with an [SEO ROI](https://missiongrowth.io/blog/seo-roi) calculation, then swap in the real number once it lands.

Some teams build one company-wide model. Others build smaller team-level "mini models" for a single channel and roll the outputs up into the company number. Paper, Miro, a spreadsheet or a custom app all work for the format; what matters is whether the inputs are wired to a formula, a list on a slide doesn't count.

Keep a first version to 5-8 inputs: one line per stage, plus at most two quality levers. Every extra line is a line someone has to keep updating every month, and most of them won't move the output enough to earn the upkeep.

That range is our own heuristic, not a number any single source states, but it's informed by the same discipline real operators use. One operator's write-up on Reforge, published in 2023, describes Quizlet's US growth model as built with "the fewest possible cuts to avoid complexity."

## How to build a growth model: Atlas's worked example, start to finish

Atlas, an illustrative small B2B SaaS company, turns 10,000 monthly visitors into a computed 18% month-over-month MRR growth rate once its five inputs are connected in a spreadsheet.

None of these numbers are real; they exist to show the arithmetic end to end.

| Input | Value |
|---|---|
| Monthly visitors | 10,000 |
| Visitor-to-trial rate | 4% (400 trials) |
| Trial-to-paid rate | 20% (80 new customers) |
| Existing base | 800 customers x $120/mo = $96,000 starting MRR |
| Net revenue retention | 108% |

Here's how the five lines turn into one number:

1. **Visitors.** Atlas gets 10,000 monthly website visitors, one of its acquisition inputs is organic traffic, the channel a dedicated [SaaS SEO](https://missiongrowth.io/blog/saas-seo) push is meant to grow.
2. **Trial signups.** A 4% visitor-to-trial rate turns those visitors into 400 trials.
3. **New paying customers.** A 20% trial-to-paid rate turns those trials into 80 new paying customers.
4. **Existing base.** Atlas already has 800 paying customers at $120 a month, a $96,000 starting MRR base.
5. **Retention plus new revenue.** At 108% net revenue retention, that base grows to $103,680 on its own. Add the $9,600 in new MRR from the 80 new customers (80 x $120), and Atlas ends the month at $113,280.

That's a move from $96,000 to $113,280, or ($113,280 - $96,000) / $96,000, an 18% month-over-month growth rate. The model's job was never the diagram. It's that number, and the number only exists once every input has a real value attached to it.

::figure{src="/blog/figures/growth-model-2.svg" alt="Atlas's growth model funnel: 10,000 visitors, a 4% trial rate, a 20% trial-to-paid rate, ending in 18% month-over-month MRR growth." caption="Atlas's growth model turns 10,000 monthly visitors into 18% month-over-month MRR growth once every input has a real number attached." width="720" height="244"}

## Use the model to decide which lever to pull first

In a multiplicative growth model, fixing the input with the lowest current rate produces the largest lift for the same size of effort.

A fixed percentage-point gain moves a small base rate further in relative terms than the same gain on a rate that's already high. Multiplying compounds a small gain harder on a low base than the same gain adds on a high one.

Run that rule on Atlas's own numbers. Two teams each get the same two-percentage-point improvement to work with:

| Scenario | Change | Result | Lift vs baseline |
|---|---|---|---|
| Fix the 4% visitor-to-trial rate | 4% to 6% | 10,000 x 6% x 20% = 120 paying customers | +50% |
| Fix the 20% trial-to-paid rate | 20% to 22% | 400 trials x 22% = 88 paying customers | +10% |

Both teams spent the same two points of effort. The team that fixed its lowest-rate stage still got five times the lift.

::figure{src="/blog/figures/growth-model-3.svg" alt="A 2 percentage-point gain on the 4% visitor-to-trial rate yields +50% more paying customers, the same gain on the 20% trial-to-paid rate yields only +10%." caption="Fixing Atlas's lowest-rate input, not its highest, produces five times the lift for the same two-percentage-point effort." width="720" height="249"}

Prioritize the input with the lowest rate first, all else equal. Once you've shortlisted more than one lever that looks competitive, score them with something like [RICE](https://missiongrowth.io/blog/rice-framework) (Reach, Impact, Confidence, Effort) before you commit a quarter of team time to one. Whatever wins that shortlist feeds your [growth experiment cadence](https://missiongrowth.io/blog/growth-experiment-cadence)'s weekly backlog.

## Keep the model alive after you build it

A growth model needs a monthly refresh, plugging in the latest actuals on the same structure, and a rebuild only when a new channel, loop or lever gets added and the structure itself has to change. Refresh is a data swap. Rebuild is a new formula.

The day you add a genuinely new channel, like a dedicated [startup SEO](https://missiongrowth.io/blog/startup-seo) push, is the day the structure itself needs to change, the numbers alone won't cover it. Ship the new line, wire it into the formula, and only then start refreshing it monthly like the rest.

Most models die from never being reopened, not from being wrong on day one. A growth model's edge over a plain diagram was never that it hands you a decision on its own; even a purely visual one can do that. The edge is that its arithmetic tells you how much each lever is worth and lets you re-check that call as real numbers come in.

Build yours, compute it end to end the way Atlas's is above, and use this month's actuals to name the one input you're fixing next.

## FAQ

### How many input variables should a first growth model have?

Keep a first version to 5-8 inputs: one per major acquisition, activation or retention stage, plus at most two quality levers. That's our own heuristic, informed by the same minimalism Quizlet's own growth model uses, not a number any single source states outright.

### How often should you update a growth model?

Refresh it monthly by plugging in the latest actuals on the same structure. Only rebuild it, changing the formula itself, when a genuinely new channel, loop or lever gets added and the old structure can't account for it.

### What's the difference between a growth model and a growth loop?

A growth loop is one self-reinforcing mechanism, like a referral cycle or a content cycle. A growth model is the larger system: every loop and channel a company runs, wired into a single formula that rolls up into one output metric.

### Do you need a growth model if you're pre-product-market-fit?

Not yet. Build one once you have a product-market fit signal, at least a month of real historical funnel data to calibrate against, and more than one team or channel competing for the same priority. Missing any of those, a diagram is enough for now.

### What's the difference between a visual and a mathematical growth model?

A visual model is a diagram of how the pieces connect, with no numbers attached; it's good for getting a team aligned on the shape of the system. A mathematical model rebuilds those same relationships in a spreadsheet with real rates and costs, which is what lets you size a lever instead of just naming it.