Financial model and numbers
How to build a bottom-up revenue forecast
Short answer
A bottom-up forecast builds revenue from what you can count: new customers each month, how many stay, and what each pays. Start from sales capacity or marketing spend, apply conversion rates you have observed, then apply retention and price. It is slower than typing a growth rate, and far easier to defend.
The chain
Capacity → leads → customers → retained customers → revenue
Each link has a number you can observe.
Step by step
- Start from what produces customers. Sales reps, marketing budget or partner count.
- Apply productivity. Meetings per rep, leads per rupee, conversions per partner.
- Apply conversion. The share that become paying customers, and how long that takes.
- Apply retention. How many customers remain each month.
- Apply price. Revenue per customer, with any planned increases.
Example
This example is illustrative.
| Line | Month 1 | Month 6 | Month 12 |
|---|---|---|---|
| Sales reps | 4 | 6 | 10 |
| New customers per rep a month | 8 | 9 | 10 |
| New customers | 32 | 54 | 100 |
| Customers at start | 200 | 395 | 760 |
| Lost at 2.5% a month | 5 | 10 | 19 |
| Customers at end | 227 | 439 | 841 |
| Revenue per customer a month | ₹2,000 | ₹2,000 | ₹2,200 |
| Monthly revenue | ₹4.5 lakh | ₹8.8 lakh | ₹18.5 lakh |
Every number can be questioned and defended: why ten reps, why eight customers each, why 2.5% churn.
By business model
- Subscription: new customers, churn, expansion, price
- D2C: new customers from spend ÷ acquisition cost, repeat rate, order value
- Marketplace: active buyers and sellers, transactions per user, value per transaction, take rate
- Enterprise: pipeline by stage × probability × contract value, with implementation time
Common errors
- New customers growing while the sales team stays the same size
- No ramp time for new hires
- Retention assumed at 100%
- Price increases with no effect on churn
- Seasonality ignored
Build yours in Model Builder, which uses this structure.
Quick answers
What is wrong with a growth percentage?
Nothing as a result, everything as an input. It hides how the growth is produced, so nobody can test it.
What if I have no history?
Use results from your first tests, however small, and label assumptions clearly.
How do I check the result?
Compare the implied growth with companies in your sector, and check that sales capacity can deliver the customer numbers.