Every growth team needs a model, but most growth models are wrong in unhelpful ways. They're either too simple to be useful (linear extrapolation) or too complex to be understood (200-tab spreadsheet nightmares). Here's how to build one that sits in the sweet spot.

Start with the Loop, Not the Funnel

Traditional growth models are funnels: traffic → signups → activation → conversion → revenue. This works for paid acquisition but completely misses the compounding dynamics that make PLG powerful.

Instead, start by mapping your growth loops. Every growth engine has at least one reinforcing loop:

• Viral loop: Users invite other users
• Content loop: Usage generates content that attracts new users via SEO
• Paid loop: Revenue funds acquisition that generates more revenue

The Three-Layer Model

Layer 1: Acquisition

Model each acquisition channel separately. For each channel, define:

• Input (spend, content published, users who can invite)
• Conversion rate to signup
• Any natural growth rate or decay

Layer 2: Activation & Monetization

From signups, model:

• Activation rate (% who reach the aha moment)
• Time-to-activation distribution
• Conversion rate to paid
• Average starting contract value

Layer 3: Retention & Expansion

For active customers:

• Monthly/annual retention curves (cohorted)
• Expansion rate (usage-based growth, upsells)
• Contraction and churn rates

Connecting the Loops

The magic is in how you connect layers. For example:

• Layer 3 retention → feeds back into Layer 1 as referral source
• Layer 2 activation improvements → compounds through Layer 3 retention

Making It Decision-Useful

A model that just forecasts revenue isn't useful. A useful growth model answers questions like:

• "If we improve activation rate by 5 points, what's the revenue impact over 12 months?"
• "Should we invest in reducing churn or increasing top-of-funnel?"
• "At what CAC does our paid channel become unprofitable?"

Build sensitivity analysis into your model. Highlight the 2-3 levers with the highest impact.

Keep It Honest

Update your model monthly with actuals. Track where your assumptions were wrong and why. A model that's never wrong is a model that's never useful — the value is in the delta between prediction and reality.

---

The best growth model I've ever seen was a single Google Sheet with 30 rows. It modeled one acquisition loop, one conversion funnel, and one retention curve. It was updated weekly and used in every planning meeting. Complexity is not sophistication.