Top-down vs bottom-up forecasting: which one to use and when

Every startup revenue forecast starts with a choice, usually made implicitly, between two fundamentally different ways of thinking about how your business will grow. Getting this choice wrong does not break your model - but it does make your model either unbelievable to investors or too conservative to be useful, depending on which direction you get it wrong.

The two approaches are top-down and bottom-up forecasting. Most founders default to one or the other. The ones with the most defensible models use both, for different parts of the same forecast.

Top-down forecasting

Top-down starts with the market and works inward. You begin with the Total Addressable Market (TAM), narrow to your Serviceable Addressable Market (SAM), narrow further to your Serviceable Obtainable Market (SOM), and project revenue as a percentage of that SOM.

The logic is: if there are 100,000 potential customers and we capture 2%, that is 2,000 customers at an average of $1,200/year, which is$2.4M ARR.

What top-down is good for: showing long-term potential and the scale of ambition investors are looking for. It is hard to build an exciting 5-year outlook purely from bottom-up assumptions, because bottom-up tends to be conservative by nature. Top-down gives you the "this is what the business can become" story.

Where top-down breaks down: it does not explain how you actually get there. "We'll capture 2% of the market" is not a plan. Investors who push on top-down forecasts quickly find that the market share assumption is often arbitrary, and there is no mechanism connecting the number to anything operational. A founder who says their business will reach $10M ARR because the market is$500M and they only need 2% is, to use the polite phrasing, presenting optimism rather than a forecast.

Bottom-up forecasting

Bottom-up starts with the specific mechanisms of your business and builds up from there. For a SaaS business: how many visitors does each acquisition channel bring per month, what is the conversion rate from visitor to trial to paid, what is average contract value, and what does churn do to the retained base over time?

If you are running LinkedIn ads and they bring 500 visitors per month, with a 3% trial conversion and 30% trial-to-paid conversion, you are adding about 4.5 customers per month from that channel. At $150/month ACV, that is$675 in new MRR per month from LinkedIn. Now add your other channels, subtract churn, add expansion, and you have a bottom-up revenue model.

What bottom-up is good for: making your short-term forecast defensible. Every number connects to something real - a conversion rate you have measured, a channel you are actively running, an ACV you are actually charging. When an investor asks "how did you get to $200k MRR by month 12?", you can walk them through exactly how.

Where bottom-up breaks down: it is hard to show scale. If you model every channel at realistic conversion rates with realistic budgets, you often end up with a forecast that is honest but not exciting. Bottom-up models also struggle to capture virality, word of mouth, and the non-linear effects that happen when a business finds real product-market fit.

The case for combining both

The most defensible approach is to use bottom-up for the first 12-24 months and top-down for the longer-term view. This way, your near-term targets are specific and testable (here is exactly how we get to $500k ARR), while your long-term ambition is contextualized by market size (here is the ceiling we are building toward).

Presenting both together to an investor signals that you are rigorous enough to work from specifics and ambitious enough to think about scale. More importantly, it makes the assumptions in both time horizons explicit enough to challenge - which is exactly what a useful startup financial model should do.

How to build the bottom-up layer for SaaS

Start by listing every product or service you are selling. For each, determine the unit you are measuring - customers, seats, contracts. Then:

1. List your acquisition channels (organic, paid, referral, sales-led)
2. For each channel, estimate monthly reach (visitors, contacts, leads)
3. Apply realistic conversion rates at each stage (visitor to trial, trial to paid)
4. Set pricing per unit per billing cycle
5. Apply churn to the retained base each month
6. Add expansion revenue if applicable

The result is a month-by-month customer waterfall that connects to MRR, ARR, and cash. This is exactly what MRR, ARR, and churn tracking requires - not just the output metrics but the underlying drivers.

Assumptions are what make or break both approaches

No matter which method you use, a forecast stands or falls on the quality of its assumptions. This is not just an analytical point - it is a practical one. When an investor asks where your conversion rate comes from, "industry average" is a much weaker answer than "we ran 300 trials last quarter and 28% converted." When they ask about your churn assumption, "we checked a few benchmarks" is weaker than "our first cohort of 40 customers is retaining at 95% after 90 days."

Assumptions that can be backed by evidence - market research, contracts with early customers, actual conversion data, competitor pricing - turn a model from a spreadsheet into a credible plan. Building that evidence library is part of building the model.

What investors see in your forecast approach

When a founder uses only top-down, investors often read it as "they have not done the operational work." When a founder uses only bottom-up, it sometimes reads as "they are not thinking big enough, or they do not understand market dynamics." The combination - bottom-up near-term grounded in actual channel mechanics, top-down long-term grounded in market sizing - is what investor-ready financials actually look like.

Founder checklist

• Your 12-month forecast is built from actual channel assumptions, not percentage growth
• Every conversion rate in the model can be connected to a source or an experiment
• Your 3-5 year view is connected to market size, not just extrapolated from near-term trends
• Changing a single channel assumption updates the rest of the model automatically
• You can explain why the forecast is bottom-up for near-term and what the top-down ceiling is

How Binokl helps