Financial Forecasting for a Small SaaS

A forecast that stays usable has a clear structure, with assumptions separated from calculations. A practical layout has five tabs:

Chargebee recommends keeping inputs in one clearly labeled tab so every downstream formula points back to it. That single rule prevents the most common modeling error: a number typed directly into a formula and forgotten.

This is a driver-based forecast. The Corporate Finance Institute describes driver-based planning as linking budgeting and forecasting to key business drivers that directly impact a company's financial performance. Instead of forecasting revenue as "last month plus 8%", you forecast the things that cause revenue, such as trials, conversion and churn, and let revenue follow. FP&A Trends notes that driver-based forecasting is rarely one model but a connected set that must remain understandable and governable. For a small company, that means a short list of drivers you can explain in a sentence each.

Wall Street Prep describes bottom-up forecasting as breaking a business apart into the underlying components that drive its revenue, relying on granular historical and operating data. For SaaS, the core is the MRR waterfall. Chargebee states the identity: beginning MRR, plus new MRR, plus expansion MRR, minus contraction MRR, minus churned MRR, equals ending MRR.

For each month:
  new customers      = trials started x trial-to-paid conversion
  new MRR            = new customers x average starting price
  expansion MRR      = beginning MRR x monthly expansion rate
  contraction MRR    = beginning MRR x monthly contraction rate
  churned MRR        = beginning MRR x monthly revenue churn rate
  ending MRR         = beginning MRR + new + expansion - contraction - churned

David Skok's SaaS metrics guide calls the net of these movements Net New ARR, the sum of new ARR plus expansion ARR minus churned ARR, and argues it is the most important number to run a SaaS business by month to month. Your forecast should show each component separately, because each has a different owner: marketing and sales drive new, customer success drives expansion and churn.

Three practical rules for the revenue drivers:

1. Base drivers on your own history. Use the last three to six months of conversion and churn, not industry averages. If you have no history yet, use conservative guesses and replace them as data arrives.
2. Forecast top-of-funnel separately by channel. Paid, organic and outbound behave differently, and each has its own cost.
3. Model usage revenue as units times rate. Chargebee warns that a flat MRR assumption breaks for usage pricing, and recommends treating usage as units consumed times rate, projected from expected adoption rather than a single monthly number.

Keep bookings, billings and revenue distinct. Skok's example: a one-year contract at a fixed monthly fee, paid up front, creates bookings equal to the entire contract value, while revenue is recognized month by month and the cash arrives at signing. A model that mixes them will overstate either revenue or cash.

In a small software company, most spending is people. Graphite recommends building anticipated hires into the model by role and expected start date, including compensation and other employment costs, and lists treating payroll as a generic percentage of revenue as a common mistake.

A simple headcount plan has one row per person or planned hire:

Employer taxes and benefits vary by country and state, so get the rate from your payroll provider or accountant rather than guessing.

Then add the non-people costs in two groups:

• Cost of revenue, which scales with customers: hosting, third-party APIs and AI model usage, payment processing fees and customer support tools. Forecast these per customer or per unit of usage, so margin stays honest as you grow.
• Operating expenses, which are mostly fixed or chosen: software subscriptions, marketing programs, legal, accounting, insurance and rent.

Separating the two lets you calculate gross margin, which tells you what each new dollar of revenue is really worth. For AI products in particular, model the per-unit cost explicitly. A usage spike that grows revenue can also grow vendor bills faster.

Profit and cash diverge in SaaS mostly because of billing terms. Graphite notes that a customer paying annually in advance can create a very different cash pattern from one paying monthly, even when recurring revenue is identical. Your cash tab should therefore model collections by billing frequency, payment processor payout delays, and the timing of large annual costs.

From the cash balance, calculate burn and runway. Wall Street Prep defines gross burn as monthly cash expenses and net burn as monthly cash sales minus monthly cash expenses, and gives the runway formula:

Runway (months) = cash on hand / monthly net burn

Calculate runway from the forecast, not only from last month. A company that plans to hire three people next quarter has a shorter runway than today's burn suggests. Graphite describes the goal as showing how the company's cash position changes over time, so leadership can see when the plan creates a financing need.

J.P. Morgan stresses that founders need the most up-to-date information, which requires regularly updating and recalculating the cash runway. Fundraising takes months, so the useful number is not "when do we run out" but "when must we start raising, or cutting, to avoid running out".

A single forecast is a guess presented as a fact. Scenarios make the uncertainty visible. Chargebee recommends duplicating the driver assumptions into base, best and worst cases, plus a fundraising case, and confirming the outputs shift when you flex an input.

Graphite suggests asking pointed questions of the model: what happens to runway if new ARR comes in 20% below plan, what if retention improves, can hiring continue if fundraising takes three months longer?

A practical way to run this for a small team:

Pick the downside numbers so that they genuinely scare you. Then write down the trigger that moves you from the base plan to the adjusted plan, such as "if net new MRR misses plan for two consecutive months, delay the next two hires". A pre-agreed trigger is worth more than any number of scenarios, because it turns a forecast into a decision you have already made.

A forecast decays the day you finish it. The habit that keeps it useful is a monthly comparison of actuals against plan. Graphite recommends that you compare actual performance with the model regularly and investigate meaningful variances: did acquisition miss, was churn higher, did hiring happen later, did cash collection slow?

A monthly routine that takes about an hour once the model exists:

1. Close the books for the month so the actuals are reliable.
2. Load actuals into the actuals tab: MRR components, customers, headcount, expenses and cash.
3. Compare each driver with its forecast and note the biggest misses.
4. Ask why for every material variance, and decide whether it is noise or a change in the business.
5. Update the drivers where the business has changed, and roll the forecast forward one month.
6. Recalculate runway under base and downside scenarios.
7. Check your triggers and act on any that have fired.

Common mistakes to avoid:

• Top-down revenue. "We will capture a small share of a huge market" is not a forecast. Build from funnel drivers.
• Hard-coded numbers. Every assumption belongs on the drivers tab.
• Ignoring churn early. Small monthly churn compounds into large annual losses.
• Confusing bookings, revenue and cash. Keep three separate lines.
• Forgetting cost of revenue. Usage-heavy products can grow revenue and lose margin at the same time.

Expert agents and playbooks, such as those on Quanta Intelligence, can help you draft a first model and pressure-test assumptions, but the drivers must come from your own data. Tax, payroll and revenue recognition assumptions should be confirmed with an accountant before you rely on them for a fundraise or a hiring decision.