A growth model is a working equation that ties the levers you can pull (new users, activation rate, ARPU, retention) to the output you care about (MRR or revenue). You do not need a 40-tab spreadsheet or a finance degree. You need 90 focused minutes, a blank Google Sheet, and a willingness to write down assumptions you can defend.
This guide walks you through the exact 90-minute workshop we run with founders and growth teams: define the equation, back into inputs, name the levers, and stress-test. Three worked examples included. Free Google Sheet template at the bottom.
What is a growth model?
A growth model is a quantitative representation of how a business produces its core output, expressed as a multiplicative equation of inputs you can influence. For most SaaS companies, the canonical form is:
New Users x Activation Rate x ARPU x Retention Rate = MRR
It is not a forecast. It is a map of the mechanics. The point is to make the engine legible so you can argue about which input matters most this quarter.
Brian Balfour, founder of Reforge, frames a company's growth as a Racecar with four parts: an engine (the loop driving most growth), fuel (inputs the engine needs), lubricants (optimisations) and turbo boosts (one-off events). Your growth model is the math underneath the engine.
A good growth model does three things:
- Names every input that turns into the output (no hidden variables)
- Quantifies each input with a current number and a target number
- Identifies 1-2 levers per input (the actions that move that number)
If your model has a variable nobody on the team can define or measure, that is the first thing to fix.
What is the difference between a growth model and a financial model?
A growth model describes how the business produces customers and revenue. A financial model describes the full P&L, cash flow, and balance sheet that result. Growth models are upstream of financial models.
Think of it this way:
| Growth model | Financial model | |
|---|---|---|
| Question it answers | How does the business actually grow? | What does the P&L look like in 18 months? |
| Primary output | MRR, customers, activations | Net income, runway, burn |
| Time horizon | Monthly, leading indicators | Quarterly / annual, lagging |
| Audience | Founders, growth, product | Investors, finance, board |
| Granularity | Per-channel, per-loop | Per-line-item |
According to Kruze Consulting, a financial model is the foundation of a startup's strategy and fundraising story. But the financial model is only as honest as the growth model feeding it. Plug an unrealistic CAC or retention rate into the financial model and the runway projection is fiction.
Build the growth model first. Then layer costs on top to get a financial model.
How do you build a SaaS growth model in 90 minutes?
You build a SaaS growth model in 90 minutes by working through four time-boxed blocks: define the equation (15 min), back into inputs (30 min), name 1-2 levers per input (25 min), and stress-test assumptions in a spreadsheet (20 min). Use a single Google Sheet. Do not open a CRM or a BI tool until you finish.
Before you start, gather three things:
- Last 30 days of paid sign-up volume
- A rough activation rate (whatever your team currently calls 'activated')
- Average revenue per user (ARPU) and monthly logo churn
If you do not know one of these numbers, write down your best guess. The point of the workshop is to find out which guesses break the model.
Block 1: Write the core equation (15 minutes)
Open a blank doc. Write the equation that produces your single most important output. For most SaaS companies that is MRR or new ARR. The default form:
MRR = New Users x Activation Rate x ARPU x (1 / Monthly Churn Rate)
The last term is the customer lifetime in months, which determines how much MRR each cohort eventually represents. If you sell annual contracts, swap MRR for new ARR and use logo retention rate instead of churn.
If your business has a marketplace, two-sided, or usage-based motion, the equation looks different (see the worked examples below). The rule: every variable on the right side must be something a team owns.
Block 2: Back into the inputs (30 minutes)
Decompose each top-level variable into the components that produce it. For New Users, that is usually channel mix:
New Users = Organic + Paid + Referral + Outbound + Partner
For Activation Rate, it is the conversion through the onboarding steps that lead to your activation event. For ARPU, it is plan mix x average plan price plus expansion revenue. For Retention, it is the percent of customers still paying at month 1, 3, 6, 12.
For each leaf node, fill in two cells: current value and target value at end of next quarter. This is where most teams discover they have been arguing about strategy without agreeing on the numbers.
Block 3: Name 1-2 levers per input (25 minutes)
For every leaf-node input, write the 1-2 specific actions that move that number. A lever is something a team can ship in a quarter, not a vague theme. Examples:
- Organic sign-ups -- launch programmatic SEO for the top 200 long-tail keywords
- Paid sign-ups -- shift 30% of LinkedIn budget to YouTube pre-roll for ICP titles
- Activation rate -- ship in-product onboarding checklist with progress bar
- Monthly churn -- launch save-flow for cancellation page (offer pause + downgrade)
If you cannot name a concrete lever for an input, that input is stuck. That is a finding, not a failure.
Block 4: Stress-test in a spreadsheet (20 minutes)
Move everything into a Google Sheet. Build columns for current, target, and three scenarios: pessimistic (-25%), base, optimistic (+25%). Watch which input moves MRR most when you flex it.
Wall Street Prep calls this sensitivity analysis. Per the NCUA's framework, sensitivity analysis varies one input at a time while stress testing combines adverse inputs. You want both. Vary inputs individually to find your highest-leverage variable, then combine three pessimistic inputs to model a bad quarter.
What inputs should a growth model have?
A SaaS growth model should have between 5 and 12 leaf-node inputs. Fewer than 5, and you are hiding mechanics. More than 12, and you cannot hold the model in your head. The minimum viable input set:
- Sign-ups by channel (organic, paid, referral, outbound, partner)
- Sign-up to activated conversion rate
- Activated to paid conversion rate (or trial-to-paid for freemium)
- Average revenue per user (ARPU) by plan
- Monthly logo churn
- Net revenue retention (NRR) for expansion-heavy businesses
For product-led businesses, add a viral coefficient or invites-per-user input if growth loops contribute meaningfully. For sales-led, add SQLs, win rate, and average contract value.
A practical rule from Lenny Rachitsky's Racecar Framework: every input in your model should map to either the engine (loops/funnels driving acquisition), lubricants (optimisations like activation and retention), or fuel (inputs the engine consumes, e.g. content, capital, salespeople).
If you cannot place an input into one of those buckets, it probably belongs in the financial model, not the growth model.
What are 3 worked examples of growth models for different motions?
Different motions need different equations. Same logic, different variables. Here are three real-world templates.
Example 1: Product-led B2B SaaS (Notion-style)
Equation: MRR = (Organic Sign-ups + Invited Sign-ups) x Activation Rate x Trial-to-Paid x ARPU / Churn
In product-led motions, invited sign-ups is its own variable because it compounds. Each new paid user invites N teammates, some of whom convert.
Sample inputs:
- 12,000 organic sign-ups/month
- 1.4 invites per active user x 22% acceptance = 3,696 invited sign-ups
- 38% activation rate
- 7% trial-to-paid
- $14 ARPU
- 4% monthly churn -> 25-month average lifetime
Resulting MRR per monthly cohort: ~$57K, or $1.4M LTV per cohort.
Levers: improve invite mechanic (Slack-style), shorten time-to-activation, raise ARPU through team plan upsell. According to Stage 2 Capital's PLG playbook, trial-to-paid for PLG businesses typically lands at 15-25%, vs 5-10% for sales-led.
Example 2: Sales-led B2B SaaS
Equation: New ARR = MQLs x MQL-to-SQL x SQL-to-Win x ACV
No activation rate, because the motion is human-to-human. Retention shows up as logo retention plus net revenue retention.
Sample inputs:
- 800 MQLs/month
- 35% MQL-to-SQL
- 22% SQL-to-Won
- $24K ACV
- 92% gross retention, 115% NRR
Resulting new ARR per month: ~$1.48M.
Levers: tighten ICP filter to lift SQL conversion, shorten sales cycle, drive expansion through usage-based pricing. The model exposes whether your bottleneck is top-of-funnel volume, conversion quality, or deal size.
Example 3: Marketplace / two-sided
Equation: GMV = Active Buyers x Transactions per Buyer x Average Order Value and Take Rate x GMV = Revenue
Marketplaces have a supply input you cannot ignore: if active sellers do not match active buyers, the equation breaks before you get to revenue. Always model both sides.
Sample inputs:
- 45,000 active buyers/month
- 1.8 transactions per buyer
- $62 average order value
- 12% take rate
Resulting monthly revenue: ~$602K.
Levers: improve search quality (drives transactions per buyer), seed supply in cold geographies, raise take rate through value-add services.
How do you stress-test a growth model?
You stress-test a growth model by flexing each input independently by +/- 25%, recording how much MRR changes, then combining the three worst-case inputs into a single pessimistic scenario. The input that moves MRR most when flexed is your highest-leverage variable. The input that breaks the model when combined is your biggest risk.
A structured stress test takes 20 minutes:
- Build a sensitivity column. For each input, calculate MRR at -25%, base, +25%. Hold all other inputs constant.
- Rank inputs by output sensitivity. The input causing the biggest MRR swing is where you should focus next quarter.
- Build a combined pessimistic scenario. Set the three weakest-controlled inputs to -25% simultaneously. Does the company still hit plan?
- Pressure-test assumptions with the team. For each base-case input, ask: 'What evidence do we have for this number? What would invalidate it?'
Per Wall Street Prep, sensitivity analysis isolates which assumptions matter most. The O'Reilly chapter on stress-testing recommends documenting the rationale behind each base-case number, not just the number itself.
The goal is not to get the model 'right'. The goal is to find out which inputs you would bet the company on and which ones you cannot defend.
What goes in the free growth model template?
We built a Google Sheet template that mirrors the 90-minute workshop. It has four tabs:
- Equation -- the top-level math, with editable cells for your output metric
- Inputs -- decomposed channels, conversion rates, ARPU, retention with current vs target columns
- Levers -- 1-2 levers per input with owner, ship date, expected lift
- Stress test -- pre-built sensitivity table with -25% / base / +25% columns and combined pessimistic scenario
Make a copy, fill it in during the workshop, and ship it to your team as the operating model for the quarter.
Do not skip the levers tab. A growth model without owners is a forecast. A growth model with owners and ship dates is an operating plan. The difference shows up in the next board meeting.
What are common mistakes when mapping a growth model?
Most growth models break for the same five reasons. Watch for these as you build yours:
- Modeling outcomes instead of inputs. 'We need $5M ARR' is a goal, not a model. The model is what you would multiply to get there.
- Inputs nobody owns. If 'activation rate' has no owner, it will not move. Assign every leaf node to a single name.
- Over-precise numbers. A 73.4% conversion rate from a sample of 40 is theatre. Round to 70% and move on.
- Skipping the stress test. A model you have not flexed is a model you do not understand.
- Treating the model as static. Refresh inputs monthly. The numbers drift, the levers change, and stale models give false confidence.
Reforge's writing on north star metrics makes a related point: the metric you optimise can deceive you if it does not capture the actual value being delivered. Same risk applies to growth models. If your equation outputs the wrong target, every lever points the wrong way.
| Motion | Core equation | Critical inputs | Highest-leverage lever |
|---|---|---|---|
| Product-led SaaS | Sign-ups x Activation x Trial-to-Paid x ARPU / Churn | Invites per user, activation rate, trial-to-paid | Onboarding to first value moment |
| Sales-led SaaS | MQLs x MQL-to-SQL x SQL-to-Won x ACV | ICP fit, SQL conversion, ACV | Tighter ICP filter on inbound |
| Marketplace | Active Buyers x Transactions x AOV x Take Rate | Buyer/seller liquidity, AOV | Search quality and matching |
| Usage-based | New Accounts x Activation x Usage Growth Rate x Unit Price | Time to first usage, expansion rate | In-product usage triggers |
| Hybrid PLG + Sales | Free Sign-ups x PQL Conversion x Sales Win x ACV | PQL definition, sales-assist trigger | PQL scoring model |