Deep-dive guide
SaaS Churn & AI Cost Burn Simulator
Executive industry analysis
A subscription business is a bucket with a leak. Customers pour in through acquisition and drain out through churn, and the revenue you earn each month is the water level. Because churn takes a percentage of the whole base while acquisition adds a fixed number, the two always settle at a steady state: new customers divided by the churn rate. A product that adds forty-five customers a month and loses three and a half percent of them will level off near thirteen hundred customers, however long you wait. Seeing that ceiling early is the single most useful thing a simulator can do for a founder.
Acquisition cost is judged by how quickly it is repaid, not by how large it looks. If a customer pays eighty a month at an eighty percent margin, a cost of two hundred and eighty takes just under five months to recover, which is healthy. The same cost on a thirty-dollar product takes more than a year and quietly consumes cash while the customer is still deciding whether to stay. Payback beyond about eighteen months means growth has to be financed by investors or reserves, which is why the model marks that node red long before profit does.
AI features have changed the cost structure of software. Traditional products had near-zero marginal cost per user, so gross margin was almost entirely hosting. A product that calls a language model for every action carries a real variable cost, and heavy users can cost more than they pay. Because AI cost scales with usage rather than with price, it behaves differently from other costs: it grows when customers love the product. Routing simple requests to cheaper models, caching repeated answers and setting plan limits are now core pricing decisions, not engineering details.
The hidden layer matters because it is spread across places nobody owns. Card processing, failed payments that never recover, refunds and goodwill credits, and support time together take a meaningful slice of revenue, often more than the hosting bill that gets all the attention. The scenario comparison shows how reducing churn by a third and cutting AI cost by forty percent outperforms spending the same effort on acquisition: the base grows more slowly at first, then the leak closes and cash flow turns up. The breaking point shows how close the business is to the edge today.
Formula and methodology blueprint
The model computes one month of recurring economics from eight inputs and projects the customer base forward with the churn and acquisition dynamics below. Constants for fees and leakage are typical subscription-business assumptions and are listed so you can judge them.
- Monthly recurring revenue
- MRR = customers × revenue per customer
- The water level in the bucket.
- Variable costs
- hosting = MRR × (1 − gross margin); AI = customers × AI cost per customer
- AI is kept separate because it scales with usage, not with price.
- Hidden revenue leakage
- fees = 2.9% × MRR + 0.30 × customers; support = 6% × MRR; failed payments = 1.8% × MRR; discounts = 2% × MRR
- Costs that rarely appear on a single line in the accounts.
- Acquisition
- new customers × cost to acquire
- Charged in the month the customers are won.
- Customer base over time
- c(t) = N/churn + (c0 − N/churn) × (1 − churn)^t
- N is new customers per month; the base converges to N divided by churn.
- Net cash flow
- net = MRR − hosting − AI − leakage − acquisition − fixed costs
- The number the breaking point drives to zero.
Cascading elasticity
Cascading elasticity: churn acts on the whole base, so a one-point change compounds every month, while price changes act immediately on every customer. AI cost is multiplied by customers, so growth in users raises it even when price stays flat. That is why churn and AI cost usually top the sensitivity ranking.
Three real-world case studies
Bootstrapped starter
Bootstrapped starter
A solo founder runs a small niche tool for a few hundred customers at a low price. With almost no fixed costs and cheap acquisition through content, the product is quietly profitable while it finds its audience.
Scaling mid-market
Scaling mid-market product
Several thousand customers at a higher price and much larger fixed costs. The business is profitable, but acquisition cost and AI usage are large numbers now, so a small slip in churn moves tens of thousands a month.
Black-swan crisis
Black-swan crisis
A competitor undercuts the product while a model price change and a wave of heavy usage double AI cost. Churn jumps to nine percent. Load it to see which lever returns the business to cash-flow positive.
Industry benchmark matrix
| Metric | Danger zone | Industry average | Top 10% elite |
|---|---|---|---|
| Monthly customer churn (SMB) | Above 6% | 3–5% | Below 2% |
| Acquisition payback | Above 18 months | 9–14 months | Below 6 months |
| AI and hosting cost as % of revenue | Above 15% | 8–12% | Below 6% |
| Gross margin including AI | Below 60% | 70–78% | Above 82% |
| Net margin after all costs | Below 0% | 5–15% | Above 20% |
Variables and levers
Paying customers
Active subscribers today.
Lever: A bigger base only helps if it stays: lock in the customers you have with annual plans before buying new ones.
Revenue per customer per month
Average subscription price after discounts.
Lever: Raise price or introduce a higher tier; a small increase on the same base flows almost entirely to profit.
New customers per month
Customers you add each month.
Lever: Shift spend to channels that convert at the lowest cost per customer, and cut the ones that do not repay within a year.
Monthly churn
Share of customers who cancel each month.
Lever: Churn is the multiplier on everything: fix onboarding in the first seven days, offer annual plans and win back cancelled accounts.
Cost to acquire a customer
Sales and marketing spend divided by new customers.
Lever: Lower acquisition cost with referrals, content and product-led trials before adding paid channels.
Gross margin before AI
Revenue left after hosting and third-party services, excluding AI usage.
Lever: Review hosting and vendor contracts; reserved capacity and caching often free several margin points.
AI cost per customer per month
Model API and compute bills divided by active customers.
Lever: Route simple requests to a cheaper model, cache repeated answers, and add usage limits per plan so heavy users pay for what they cost.
Fixed costs per month
Salaries, tools and overhead.
Lever: Hold hiring until payback is under twelve months; fixed costs decide how long you can survive a bad quarter.
How to use this simulator
- 1Load the Realistic preset or enter your current subscriber numbers.
- 2Watch the five nodes to see whether acquisition, retention or AI cost turns red first.
- 3Press Find my breaking point to see how far churn or acquisition cost can move before cash flow turns negative.
- 4Compare the retention and cost-efficiency plan against the current plan on the 24-month timeline.
- 5Share the scenario link with your co-founder or investor.
Mistakes to avoid
- Celebrating new sign-ups while churn quietly cancels them. The base settles at new customers divided by churn, so a leaking product cannot grow past that ceiling however much it spends on acquisition.
- Measuring gross margin without AI usage. A single heavy user can cost more in model bills than they pay, so track AI cost per active customer separately and tie limits to each plan.
- Ignoring payment failures, refunds and support in the unit economics. Together they often take a larger share of revenue than hosting, and none of them appears on one line of the accounts.
- Hiring ahead of payback. Fixed costs set how long you survive a bad quarter, so add people only when acquisition pays back in well under eighteen months.
Frequently asked questions
Why does churn matter more than growth?
Churn removes a percentage of every customer you have, every month, so it compounds. Growth adds a fixed number. Over time the base settles at new customers divided by the churn rate, so lowering churn raises the ceiling directly.
What is a healthy payback period?
Under twelve months is comfortable for most subscription products. Beyond eighteen months you are borrowing against future customers, which only works with outside funding or large cash reserves.
How should I think about AI costs?
As a variable cost that rises with usage. Track cost per active customer, route simple tasks to cheaper models, cache repeated answers and limit heavy usage by plan so the most active users cannot make a plan unprofitable.
How accurate are the numbers?
They are a planning model with explicit assumptions rather than accounting. Use your own inputs and read the output as a guide to which lever matters most.
What do the Monte Carlo results mean?
The tool reruns the scenario 1,000 times with every input varied randomly by around fifteen percent and shows how often monthly cash flow ends at zero or below, with the range of likely outcomes.
This is an educational planning model with stated assumptions. It is not financial, tax or legal advice. Replace the inputs with your own figures and verify important decisions with a qualified professional.