Startup CAC: check the cost to win a customer

Agree who counts as a new customer

Ask for the first paid-plan date and the account ID. An ID is a unique label for one account. Keep free trials, leads and old buyers apart. State how refunds, new seats and linked accounts are treated.

A subscription is a plan paid for over time. For these plans, use one rule for new paid accounts. Stripe leaves out trials and leads that did not pay. [1] Check the sales model first. Not all firms use paid plans.

• Name the product and buyer group.
• State the start and end dates.
• Define new, paid and one account.
• Keep trial counts in a separate row.

Trace costs, not just ad spend

Ask for ad bills, sales pay, fees and tools. Stripe includes sales and marketing pay, tools and agency fees. [1] Fully loaded means all costs within the stated scope. It does not mean every cost in the firm.

Split shared costs using a clear rule. Separate work to win new buyers from work on renewals. Do not count the same work twice. Put direct costs to serve buyers in the payback check.

State when founder sales work has no wage in the books. Do not call that work free at scale. If you use a wage estimate, show it apart. Explain how you chose it. AI must not invent pay or a missing bill.

• Keep the reported cost total.
• Link each added cost to its source.
• Explain shared-cost splits and exclusions.
• Flag unpaid time and missing records.

Rebuild one small CAC example

This is a made-up example in US dollars. Assume one month and a short sales process. Ads cost $2,000. Sales and marketing pay costs $4,000. Tools and agency work cost $1,000. All are for new buyers.

The total cost is $7,000. There are ten new paid accounts and 90 free trials. CAC is $7,000 divided by ten: $700 per paid account. Dividing by all 100 accounts gives $70. That is not paid-customer CAC.

A long sales process can break a same-month match. Costs now may win buyers three months later. Stripe warns about this lag. [1] Ask for sales dates and the match rule. A cohort is a group tracked from the same start time.

• Cost scope: $7,000 to win new paid accounts.
• Buyer count: ten, not 100.
• Result: $700 for each new paid account.
• Timing limit: no sales lag in this example.

Check payback using gross profit

Payback means time to earn back the cost to win a buyer. Stripe uses monthly revenue less the cost to serve them. [2] Revenue means sales earned. Gross profit is what remains after direct delivery costs. It is not the firm's net profit.

Keep the made-up $700 CAC. Each buyer earns $200 a month in revenue. Direct service costs are $100. Monthly gross profit is $100. Payback is $700 divided by $100: seven months. Using revenue alone gives 3.5 months and skips service costs.

Assume each buyer stays and both monthly amounts stay fixed. The example has no refunds, growth or late payments. Seven months is not a cash forecast. Upfront payment and later cash collection can change cash needs.

If monthly gross profit is zero or less, stop. This simple payback formula has no finite positive result. Do not hide that gap with a zero-month score.

• Keep revenue and delivery costs separate.
• Check the margin for this buyer group.
• State the stay-rate and cost assumptions.
• Show cash timing in a separate plan.

Give AI a source-linked review record

Due diligence means checks before you invest. Ask AI to sort files you may use. It must not make a deal choice. Share only what is needed. Use buyer IDs when names are not needed.

For each result, save cost rows, buyer rows and dates. Add the rule, source IDs and the person who checked. Keep the founder's figure beside your rebuilt view. Explain why they differ. Do not erase the first view.

Mark a missing field as unknown, not zero. Pause when bills are missing, buyers are mixed or dates conflict. Ask the founder a short question tied to the source. A tidy AI answer is not proof.

• Reported figure and rebuilt figure.
• Currency, dates and buyer group.
• Included costs, count rule and source IDs.
• Open gaps, reviewer and next check.

Compare like with like, then decide what to check

Bessemer links payback to the type of buyer. Its guide adjusts revenue for gross margin. It also counts some costs to renew and grow sales. [3] Our new-buyer-only scope is not the same. Match the scope before you compare.

Old software benchmarks are not a global pass mark. Pre-seed firms may have too few buyers for a clear trend. Seed teams can show more groups over time. At Series A, check buyer types and sales routes. One mean can hide big gaps.

Check if buyers stay long enough to repay the cost. A low CAC alone does not prove demand or a sound deal. Use revenue, delivery cost and retention checks next. Retention means buyers stay or keep paying.

• Compare the same scope and sales model.
• Check more than one buyer group over time.
• Pause when the data is too thin.
• Use the result to choose the next question, not promise returns.