AI Product ROI Monte Carlo
?Community
00
12-month AI product ROI model (adoption, ARPA, churn, CAC, token economics, cloud, headcount, risk) → revenue, cost, profit.
NOTEModel scope
Lean 12‑month Monte Carlo for an AI product: adoption, ARPA, churn, CAC, token usage & cost, fixed cloud, headcount, and one‑off risk costs → revenue, cost, profit.
NOTECustomize
Swap in your assumptions: TAM & penetration, ARPA net of discounts, churn, CAC, token usage/customer, and infra & staffing. Extend to multi‑year growth or enterprise deal tiers if needed.
VARIABLEtam_customers
triangular(2000, 5000, 12000)
Addressable potential paying customers in segment.
tam_customers
VARIABLEpenetration_year1
beta(6, 114)
Year‑1 penetration of TAM to paying customers (mean ~5%).
penetration_year1
VARIABLEannual_arpa_net
lognormal(11000, 4500)
Net ARPA after discounts/credits.
annual_arpa_net
VARIABLEannual_churn
beta(9, 51)
Annual logo churn fraction (mean ~15%).
annual_churn
VARIABLEcac_per_customer
lognormal(2500, 1200)
Customer acquisition cost per new customer.
cac_per_customer
VARIABLEtokens_per_customer_month
lognormal(1800000, 900000)
Avg inference tokens per active customer per month.
tokens_per_customer_month
VARIABLEcost_per_1k_tokens
triangular(0.0008, 0.002, 0.006)
Blended marginal cost per 1K tokens.
cost_per_1k_tokens
VARIABLEfixed_cloud_month
triangular(2500, 6000, 16000)
Fixed monthly infra (DB, networking, observability).
fixed_cloud_month
VARIABLEfte_count
triangular(4, 7, 12)
Average FTEs allocated to this product.
fte_count
VARIABLErisk_cost_oneoff
triangular(0, 20000, 250000)
One‑off costs (outage credits, legal/regulatory, security incident).
risk_cost_oneoff
CONSTANTmonths
12
Horizon used for annualizing usage and revenue.
months
CONSTANTloaded_salary_per_fte
210000
Fully loaded cost per FTE.
loaded_salary_per_fte
CONSTANTother_fixed_overhead
60000
Other fixed overhead not modeled elsewhere.
other_fixed_overhead
FORMULAcustomers_year1
tam_customers * penetration_year1
Paying customers acquired in year 1.
tam_customers
penetration_year1
customers_year1
FORMULAavg_active_customers
customers_year1 * (1 - (annual_churn / 2))
Approx avg active customers across year (linear churn).
customers_year1
annual_churn
avg_active_customers
FORMULAgross_revenue
avg_active_customers * annual_arpa_net
Net revenue for year 1.
avg_active_customers
annual_arpa_net
gross_revenue
FORMULAinference_cost
(avg_active_customers * tokens_per_customer_month * months / 1000) * cost_per_1k_tokens
Variable inference spend.
avg_active_customers
tokens_per_customer_month
months
cost_per_1k_tokens
inference_cost
FORMULAtotal_cost
inference_cost + (fixed_cloud_month * months) + (fte_count * loaded_salary_per_fte) + (customers_year1 * cac_per_customer) + other_fixed_overhead + risk_cost_oneoff
All‑in annual cost.
inference_cost
fixed_cloud_month
months
fte_count
loaded_salary_per_fte
customers_year1
cac_per_customer
other_fixed_overhead
risk_cost_oneoff
total_cost
FORMULAprofit
gross_revenue - total_cost
Profit (negative = burn).
gross_revenue
total_cost
profit
FORMULAprob_profitable
if(profit > 0, 1, 0)
1 if profitable in year 1.
profit
prob_profitable
OUTPUTPaying customers (year 1)
Paying customers (year 1)
Distribution of year‑1 paying customers.
OUTPUTRevenue (year 1)
Revenue (year 1)
Net revenue distribution.
OUTPUTTotal cost (year 1)
Total cost (year 1)
All‑in cost distribution.
OUTPUTProfit / burn (year 1)
Profit / burn (year 1)
Profit distribution (negative = burn).
OUTPUTProbability of profitability
Probability of profitability
Mean = probability profit > 0.
What is Carlo?
Carlo is a visual tool for Monte Carlo simulation. Model uncertainty by dragging probability distributions, connecting them visually, and running thousands of scenarios instantly.