Free trial vs book a demo: the choice that sets your ad budget
Founders ask us this before almost anything else about ads: should the button say "start free trial" or "book a demo"? We checked what 18 of the best-known YC B2B companies chose, and then did the math on what each answer costs you on Meta. The short version: the CTA is a budget decision wearing a design decision's clothes.
The question sounds like a copywriting choice. It isn't. The CTA you pick decides your entire acquisition motion: who touches the product first, how long your sales cycle runs, and, the part almost nobody prices in, how much you'll have to spend on ads for the campaigns to work at all. Meta's delivery system needs a steady diet of conversion events to find your buyers, and a demo request costs three to five times what a trial signup does. Pick the expensive event with a seed-stage budget and your campaigns starve before they learn anything.
So before the theory, we did what we did for our YC ads census: we went and looked. This week we pulled up the homepages of 18 of the best-known B2B companies to come out of Y Combinator and recorded what the button actually says.
The short version
- Ten of the 18 YC winners lead with self-serve ("Start free", "Create free account"), five lead with a demo, three run both side by side. The split maps almost perfectly to deal size.
- The rough rule: self-serve below ~$10k ACV, hybrid to ~$50k, demo-first above that. None of the 18 breaks it by much.
- On Meta, a trial signup runs $30–80 and a demo request $80–250. A qualified, booked demo runs $200–600.
- Meta needs ~50 conversion events per ad set per week to learn. With a demo CTA that means $4,000–12,500 a week, and $10,000–30,000 if you optimize on booked demos. Without that budget, optimize higher in the funnel.
- Cheap signups are worth less: a credit-card trial produces roughly 3× more paying customers per visitor than a no-card trial, on a third fewer signups.
What 18 famous YC companies chose
Here's what the primary homepage CTA says today at 18 YC B2B companies you'd recognize, grouped by the motion they lead with.
| Company (batch) | Primary CTA today | Backup path |
|---|---|---|
| Self-serve first | ||
| Stripe (S09) | "Start now" | Contact sales |
| Mixpanel (S09) | "Get Started Free" | "Book a Demo" |
| PagerDuty (S10) | "Start for free", no credit card required | "Request a demo" |
| Zapier (S12) | "Start free" | Talk to sales |
| Gusto (W12) | "Create free account" | Self-guided demo |
| Amplitude (W12) | "Start for free" (2M events free) | Talk to sales |
| Webflow (S13) | "Get started — it's free" | Talk to sales |
| Brex (W17) | "Get started" (signup) | — |
| Retool (W17) | Explore the app builder for free | — |
| PostHog (W20) | "Get started – free" | "Talk to a human" |
| Both, side by side | ||
| Checkr (S14) | "Sign up" | "Talk to sales" |
| Front (S14) | "Start free trial" | "Request demo" |
| Algolia (W14) | "Get started" (signup) | Demo request |
| Demo first | ||
| Ironclad (S15) | "Request Demo" | Watch overview video |
| Podium (W16) | "Watch a demo" | Sales phone line |
| Rippling (W17) | "Get a demo" (on every product section) | "Create free account" email gate in the hero |
| Vanta (W18) | "Get a demo" | Interactive demo |
| Deel (W19) | "Book a demo" | — |
Look at who sits in each group and the pattern stops being mysterious. The self-serve group is payments, dev tools, analytics and SMB software: products a single person can adopt with a credit card in an afternoon. The demo group is global payroll, compliance, contract management and multi-product HR platforms: five-figure contracts where legal, IT and finance all get a say before anything goes live. Nobody in the first group is precious about sales calls, and nobody in the second group believes a lawyer will self-onboard a CLM. The CTA follows the deal, not the fashion.
One more detail worth stealing: Rippling, the most aggressive advertiser of the demo group, doesn't actually open with the demo. The hero asks for your work email to "create a free account", and the demo asks come after. They capture the cheap, high-volume event first and sell the expensive conversation second. Hold that thought for the budget section.
The ACV rule behind the pattern
The industry shorthand that matches both the benchmark data and the table above is an ACV ladder.⁵
| Your ACV | Lead with | Why |
|---|---|---|
| Under ~$10k | Self-serve trial or free plan | A sales-assisted motion costs more to run than the contract is worth. Volume and activation do the selling. |
| ~$10k–50k | Hybrid: trial plus a demo offer for high-intent accounts | Both motions pencil out. GrowthSpree's data shows demo-plus-trial hybrids beating standalone trials above roughly $35k ACV. |
| Above ~$50k | Demo | Buying committees don't self-onboard. The trial becomes a proof of concept inside the sales process, not the front door. |
If your ACV is genuinely undecided, which at seed stage it often is, the honest answer is that the CTA question is premature and both buttons are an experiment. That's fine. What you shouldn't do is copy Deel's homepage because Deel looks successful. Deel sells multi-country payroll to companies with legal exposure in twelve jurisdictions. If you're selling a $49-a-month tool to startups, their CTA is the wrong answer for you and their ad budget is too.
What each conversion event costs on Meta
Now the money. On Meta, cost scales with funnel depth: the deeper the commitment you ask the algorithm to find, the more each event costs. Across 2025–2026 benchmarks, the all-industry average cost per lead is $27.66, and a raw B2B SaaS lead runs around $63.¹ ⁶ For the specific events this article cares about, the agency-published ranges cluster like this:⁵ ⁶
Further down the funnel it keeps escalating: a sales-qualified lead from Meta runs anywhere from $400 to $2,400 depending on ACV, and enterprise qualified leads run $350–550 each.⁵ Treat all of these as directional ranges rather than targets; they're agency-aggregated figures, not audited datasets, and two accounts in the same vertical can differ several-fold. But the shape of the ladder is not in dispute, and the shape is what decides your budget.
The 50-demos-a-week problem
Here's the constraint founders consistently discover too late. Meta's delivery system has a learning phase, and per Meta's own guidance an ad set needs roughly 50 optimization events within 7 days to exit it.⁴ Only the event you're optimizing for counts. Below that volume the ad set sits in "Learning Limited", which in practice means unstable delivery and costs that never come down, because the algorithm never accumulates enough signal to know who converts.
So if your conversion event is "book a demo", you're realistically signing up to buy about 50 demo requests a week. Multiply that by what the events cost and you get the entry ticket:
| Conversion event | Typical Meta cost | Weekly budget for ~50 events |
|---|---|---|
| Free trial signup (no card) | $30–80 | $1,500–4,000 |
| Demo request (form fill) | $80–250 | $4,000–12,500 |
| Qualified, booked demo | $200–600 | $10,000–30,000 |
That's the real cost of the demo button for an advertiser: not $150 per lead, but a $4,000-to-$12,500-a-week commitment before the campaign can even stabilize, and $10,000-plus if you want the algorithm optimizing on demos that actually get booked and attended. The quick formula for your own numbers is target cost per event × 50 ÷ 7 = minimum daily budget per ad set. A $150 demo request works out to about $1,070 a day.
If that's not your budget, and at seed stage it usually isn't, the answer is not to run the demo campaign anyway on $500 a week. It's to move your optimization event up the funnel to something you can afford 50 of: a trial signup, a completed registration, even a landing page view at the extreme. Then pass the downstream outcomes (qualified lead, booked demo, closed-won) back to Meta through the Conversions API as custom conversions, so the algorithm still learns what a real buyer looks like even though it's optimizing on a cheaper proxy.⁵ This is exactly the Rippling homepage trick in ad-account form: buy the cheap event at volume, sell the expensive conversation behind it.
Budget rule of thumb: your weekly Meta budget divided by 50 is the most expensive conversion event you can afford to optimize on. $2,500 a week buys you a $50 event, which is a trial signup, not a demo.
Why cheap signups are worth less
At this point the trial CTA looks like the obvious winner: cheaper events, faster learning, lower entry ticket. Time for the other shoe. The cheaper the event, the less each one is worth, and the spread is bigger than most founders expect.
The clearest data comes from the 2026 free-to-paid conversion report that Kyle Poyar ran with ChartMogul and ProductLed across 200 B2B software products, plus First Page Sage's benchmarks from their own client base.² ³ Trial-to-paid conversion by model:
| Trial model | First Page Sage | ChartMogul 2026 | GrowthSpree range |
|---|---|---|---|
| No credit card required | 18.2% | 8.9% | 8–22% (median 14%) |
| Credit card required | 48.8% | 31.4% | 35–55% (median 44%) |
| Freemium | 2.6% | 5.6% | 2–8% |
The studies disagree on levels because they measure different populations, but they agree on the multiple: a card-gated trial converts roughly 3 to 5 times better than an open one. And the per-visitor math is even starker. In the ChartMogul model, 1,000 visitors to a no-card trial produce 45 signups and about 3.6 paying customers. The same 1,000 visitors to a card-required trial produce only 35 signups but 10.5 paying customers.³ Fewer, more expensive events, three times the revenue.
Put the two halves together and you can price a customer from each motion. Illustratively: a $50 no-card trial signup at a 15% trial-to-paid rate is about $333 per customer. A card-gated signup, even if the friction pushes your cost to $120, converts at ~44% and lands around $273 per customer. A $400 booked demo at a 70% show rate and a 25% close rate is roughly $2,285 per customer, which is why that motion belongs to companies whose contracts have another zero on them.² ⁵
Optimize for cost per paying customer, not cost per event. The cheapest event Meta can buy you is also the one the algorithm finds easiest to fake.
There's one genuine tension to manage: the card gate that triples your conversion rate also cuts the event volume your campaigns learn from. A common resolution is to run the open trial as the paid-traffic front door for volume, keep the card gate for organic traffic, or use a reverse trial (full access that downgrades to a free plan) which benchmarks around a 24% median conversion. Whichever you choose, choose it with both halves of the math in view, and sanity-check your numbers against current platform benchmarks rather than a single agency's blog post.
If you pick demo, defend the demo
For those above the ACV line where the demo is the right call, the work isn't done when the form fires. Two lossy stages sit between a $150 demo request and revenue, and both are fixable.
No-shows. Industry medians put demo show rates at 55–72%, meaning roughly a third of the demos you paid for never happen, and paid-social bookings show at the bottom of that range.⁵ The fixes are boring and effective: same-day or next-day slots instead of next week, calendar booking embedded in the form instead of "we'll reach out", reminders on more than one channel, and a five-minute recovery call the moment someone no-shows.
Unqualified bookings. Add two or three qualifying questions to the form (company size, role, timeline). Volume drops 20–30% and it's mostly the junk leaving; your cost per qualified demo usually improves even as cost per form fill gets worse.
And notice what the best demo-led companies in our table do for everyone who won't book a call: Vanta runs an interactive product tour, Podium leads with "watch a demo" instead of "talk to sales", Ironclad offers an overview video, Rippling posts an on-demand tour next to the live demo. They've all built a self-serve-shaped side door on a demo-led house, partly for the buyer's sake and partly because those low-friction events give their ad accounts something cheap to optimize on. The motions converge: trial companies add sales assist as they move upmarket, demo companies add self-serve as they scale spend.
The choice, then, was never really trial versus demo. It's which expensive thing you can afford to learn on. Pick the conversion event your budget can feed 50 times a week, instrument the funnel so the platform learns from real outcomes rather than form fills, and let the button on the homepage say what the economics already decided.
Frequently asked questions
Should we require a credit card for our free trial?
Run it as an economic experiment, not a debate. A card gate cuts signup volume hard but multiplies trial-to-paid conversion: 48.8% with a card against 18.2% without in First Page Sage's client data, and 31.4% against 8.9% in ChartMogul's 2026 survey. Measure paying customers per 1,000 visitors and effective CAC for both variants. One caveat for paid traffic: fewer signups also means fewer conversion events for Meta's algorithm to learn from, so a card gate can starve your campaigns even while it improves conversion.
How much budget do we need to run Meta ads with a book-a-demo CTA?
Enough to buy roughly 50 demo requests a week, because that's what Meta's learning phase needs per ad set. At the $80 to $250 that B2B SaaS demo requests typically cost, that's $4,000 to $12,500 a week. If you want the algorithm optimizing on qualified, booked demos at $200 to $600 each, it's $10,000 to $30,000 a week. If those numbers are out of reach, don't run a starved demo campaign; optimize on a cheaper event instead.
Can we run Meta ads on less than $4,000 a week?
Yes, by picking a cheaper conversion event. A free trial signup at $30 to $80 needs roughly $1,500 to $4,000 a week to feed the algorithm 50 events. Below that, optimize on an even higher-funnel event like a completed registration or a landing page view, and pass the deeper outcomes (qualified lead, booked demo, closed-won) back to Meta through the Conversions API so the system still learns who your real buyers are.
Our ACV is high and sales-led. Should we ever offer a trial?
Above roughly $50k ACV the demo stays the front door, but the best demo-led companies still offer something self-serve to catch visitors who won't book a call: Vanta runs an interactive product tour, Podium leads with a watch-a-demo video, Rippling collects an email for a free account before selling the live demo. These low-friction options also give your ad account a cheaper, higher-volume conversion event to learn on.
Do YC companies actually advertise on Meta?
Mostly no. In our census of all 6,055 YC companies, only 8.3% had ads in Meta's ad library. The ones that do skew exactly the way the budget math predicts: established companies with sales teams and real budgets, like Deel, Vanta, Rippling and Podium, all of which run demo-led motions and can afford to feed the algorithm. Early-stage companies mostly start with Google search, where intent arrives pre-qualified and there's no 50-events-a-week learning requirement to fund.
Sources
- Meta CPL and CPC benchmarks across industries, 2025 — WordStream, Facebook Ads Benchmarks 2025
- Trial-to-paid conversion by model, opt-in vs opt-out — First Page Sage, SaaS Free Trial Conversion Rate Benchmarks
- The 1,000-visitors model and conversion by trial type, 200 B2B products — Kyle Poyar's Growth Unhinged with ChartMogul & ProductLed, The 2026 Free-to-Paid Conversion Report
- The ~50 optimization events per week requirement — Meta Business Help Center, About the Learning Phase
- B2B SaaS Meta cost per event, cost per SQL by ACV, demo show rates and hybrid-motion data — GrowthSpree, Meta Ads Benchmarks 2026 for B2B SaaS
- Facebook ads cost ranges for SaaS by funnel stage — Aimers, Facebook Ads Cost for SaaS 2026
Homepage CTAs recorded July 23, 2026 by visiting each company's site. Cost figures are directional ranges from agency-published benchmarks, not audited datasets; your account will differ.