Tom recommends writing down the savings, time reductions, or revenue gains a product could deliver, then asking the customer champion to challenge the assumptions. For a company with 100 support agents costing $100,000 each per year, a tool that cuts queries by 20% could save $2 million. He suggests charging roughly 25% to 50% of the value, and using the same value equation to define pilot success metrics.
How To Price For B2B
Speaker
Tom, a Y Combinator partner, explains how startup founders can set and test B2B software prices; no specific event or co-speakers are stated.
Tom says to calculate costs after estimating customer value, since cost-plus pricing can leave software underpriced. He recommends targeting 80% to 90% gross margins and counting cloud or AI credits as real costs. If a competitor undercuts the price, he advises improving differentiation through functionality, integrations, or industry focus instead of starting a price war.
Tom advises asking how customers pay for similar software and keeping the pricing structure simple; uncapped usage fees can make buyers wary. He favors committed monthly or annual revenue where possible, and suggests converting observed usage into a discounted monthly commitment. Pricing also shapes the sales model: as a rule of thumb, a salesperson earning $100,000 in total compensation might be expected to close $500,000 in new ARR each year.
Tom recommends keeping pilots to a few weeks and tying them to clear success criteria. For confident sellers, he suggests an annual contract with a 30- or 60-day money-back option. If pricing is uncertain, start near what customers pay for similar software, raise the price by 50% with each new prospect, and use losses to find the range where more than 25% of deals are lost solely on price.
Takeaways
- Ask the customer champion to validate the measurable value before setting a price.
- Use customer value to anchor the price, then check that delivery costs support strong margins.
- When a competitor discounts, improve product differentiation instead of matching every cut.
- Keep pilots short and agree on measurable success criteria up front.
- Treat early sales as pricing experiments and increase quotes until price alone starts losing more deals.
Mentioned in this talk
Search inside this talk
What was claimed 10
- role atTom is partner at
Y Combinator
“I'm a partner here at Y Combinator”
YouTube Unverified1 source· Talk
- metricvalue equation: In the example, $2 million in potential savings supports a $700,000 contract price.
“That's $2 million in potential cost savings”
YouTube Unverified1 source· Talk
- metricvalue equation: The speaker typically recommends pricing at 25% to 50% of delivered value, roughly one-third.
“I typically pick somewhere between 25% and 50% of the value you're delivering”
YouTube Unverified1 source· Talk
- feature
AWS: Provides credits to startups, which should be counted as cash costs when assessing margins.
“AWS, Microsoft, OpenAI, and others give out tons of credits to startups”
YouTube Unverified1 source· Talk
- feature
OpenAI: Provides credits to startups, which should be counted as cash costs when assessing margins.“OpenAI, and others give out tons of credits to startups”
YouTube Unverified1 source· Talk
- metric
Anthropic: The speaker says LLM companies have been dramatically reducing costs.
“OpenAI and Anthropic keep dramatically reducing costs”
YouTube Unverified1 source· Talk
- featureusage-based billing: The speaker recommends considering a cap because customers are wary of uncapped usage pricing.
“you might want to put a cap on it”
YouTube Unverified1 source· Talk
- metricmonthly recurring revenue: The speaker recommends aiming for MRR or ARR when possible to protect revenue predictability.
“if you can, aim for MRR or even ARR”
YouTube Unverified1 source· Talk
- metricTom: A salesperson earning $100,000 in annual compensation might be expected to close $500,000 of new ARR per year, a 5:1 ratio.
“about a 5:1 ratio between new signed ARR and total compensation”
YouTube Unverified1 source· Talk
- metricPilot: Keep pilots short, around two to four weeks, with clear success criteria.
“maybe a couple of weeks, maybe four weeks”
YouTube Unverified1 source· Talk












