Build a growth model to forecast and diagnose product growth

Stop Guessing Why You're Growing (or Not)

Most product teams can tell you their MRR and their churn rate. Very few can explain the causal chain: which acquisition channels feed which cohorts, how retention differs by segment, and what happens to the business if any single lever moves 10%. Without a growth model, product strategy is intuition dressed up as a plan.

According to Reforge's 2024 growth benchmarks, companies with a quantitative growth model are 2.5x more likely to hit their annual growth targets than those relying on top-line metric tracking alone. The model doesn't need to be perfect — it needs to decompose growth into components your team can actually influence.

Why Dashboards Aren't Enough

A dashboard shows you what happened. A growth model shows you why it happened and what will happen next. When MRR drops, a dashboard shows the dip. A growth model tells you whether it's because new user acquisition decreased (acquisition problem), activation rate declined (product problem), or churn spiked in a specific cohort (retention problem). Each diagnosis leads to a completely different intervention.

Dan Hockenmaier, who built growth models at Thumbtack and Faire, describes the process as "building a map of your business." Once you see the map, you can identify which roads are congested (bottlenecks) and which routes are underutilized (opportunities). Without the map, you're driving blind.

How the Growth Model Prompt Works

This prompt builds a model in six steps. Growth equation decomposition breaks revenue or active users into additive and subtractive components. Acquisition funnel mapping quantifies each channel's volume, conversion, and cost. Retention curve analysis plots cohort behavior and identifies the terminal retention rate. The model itself combines these inputs into a 12-month forecast under three scenarios. Sensitivity analysis reveals which lever matters most. Finally, a growth diagnosis translates model outputs into strategic priorities.

The sensitivity analysis in Step 5 often produces surprising results. Teams that assume "we need more users" frequently discover that a modest retention improvement would have 3x the impact of doubling acquisition spend.

When to Use It

• You're creating an annual plan and need defensible growth projections
• Growth has slowed and your team can't agree on the cause or the fix
• You're presenting to investors and need to show you understand your growth engine
• You've been optimizing one lever (e.g., acquisition) without checking if it's the right one
• A new competitor has entered the market and you need to model different response scenarios

Common Pitfalls

Over-engineering the model. A growth model with 50 inputs is as useless as no model. Focus on the 5-8 variables that explain 80% of growth variance. You can always add complexity later.

Using averages when you need cohorts. Blended retention rates hide critical information. A product with 50% average retention might have 80% retention for users acquired through organic channels and 20% for paid — leading to completely different strategic conclusions.

Treating the model as truth. A growth model is a thinking tool, not a prediction engine. Its value comes from the conversations it forces: "What assumption would need to be wrong for this forecast to break?"

Sources

• Building a Growth Model — Reforge on bottom-up growth modeling
• Companies With Growth Models Are 2.5x More Likely to Hit Targets — Reforge growth benchmarks
• The Growth Handbook — Intercom's practical guide to modeling and executing growth