Price testing for SaaS: find the plan price that earns the most per visitor

Price testing for SaaS: what to test first on your plans, how Leverprice moves traffic to the price that earns the most per visitor, and which plan fits.

Short answer

Leverprice tests your SaaS plan price on real visitors. You set a range, such as $29 to $49 a month, each visitor sees one price from it, and traffic moves toward the price with the highest revenue per visitor. Your server computes the price and your own checkout charges it. Free runs a manual split; Pro automates it.

  • Revenue per visitor picks the winner: $39 at 2.0% conversion earns $0.78 per visitor, more than $29 at 2.5% ($0.725).
  • Raising a $29 plan to $39 loses money only if more than 25.6% of buyers walk away.
  • If Leverprice is slow or down, your server gets your own fallback price within 500 ms and checkout keeps working.
  • Each price is judged on its first payment; renewals and churn stay outside the test.

Example: your Pro plan costs $29 a month and 2.5% of pricing-page visitors buy it, so each visitor is worth $0.725. If $39 converted 2.0% of them, each visitor would be worth $0.78, about 7.6% more. Nobody emails a founder to say the price is too low, so that gap stays invisible.

Leverprice checks your price against the people who actually visit. You set a range you'd accept, each visitor sees one price from it, and traffic moves toward the price with the highest revenue per visitor.

Why is SaaS pricing hard to test?

Because the metric everyone watches points the wrong way, and SaaS traffic is thin. A cheaper plan almost always wins on conversion rate; the question is whether it wins on money per visitor, a smaller and noisier difference.

A classic 50/50 test also parks half your visitors on the losing price until it ends, which at SaaS volumes takes months. Leverprice runs a multi-armed bandit instead: it recomputes traffic weights every few minutes, sends more visitors to the prices that are ahead and keeps at least 5% on each. It calls a winner when one price is at least 95% likely to be best, or when the rest are practically equivalent.

The price also has to hold from page to checkout: a signed first-party cookie keeps each visitor on one price, and the user id you pass (hashed on your server) keeps signed-in users on it across devices.

Illustration: raising $29 to $39 loses money only if more than 25.6% of buyers walk away, because 1 - 29/39 = 0.256 (before fees). The price increase calculator runs that break-even for your own prices.

What should a SaaS company price test first?

Start with the plan most new customers buy, then the annual price, then the next plan you launch. Test one at a time and keep the prices far apart. In our simulation of the production engine, a price that earned 5% more per visitor took a median of 195 days to find at 100 sales a month, and one that earned 20% more took 68 days (full simulation).

1. The monthly price of your main plan

Example: Pro is $29 today. Test $29, $39 and $49 with fallbackPrice: 2900, minPrice: 2900 and maxPrice: 4900, in minor units. The SDK never serves a price outside that range.

2. The annual price

Example: twelve months of a $29 plan cost $348. Test $249, $290 and $319 a year, which are 28%, 17% and 8% off; the annual discount calculator does this math for your prices. The test counts annual purchases only, so check that monthly sign-ups held steady while it ran.

3. The next plan you launch

A new tier has no price history and no customers to grandfather. Example: a Team plan tested at $79, $99 and $129 a month from the day it ships.

How does Leverprice fit a SaaS stack?

It runs in your server code, next to the checkout you already have: your server asks the SDK for a price where it renders the page and again where it creates the checkout. No Leverprice JavaScript runs in the browser.

The SDKs are open source with zero dependencies: TypeScript (Node 20+, Bun, Deno, Cloudflare Workers, Vercel and Netlify functions, AWS Lambda), Python 3.9+, PHP 8.1+, Ruby 3.1+ and Go 1.22+. They pass the same conformance vectors, so a Next.js front end and a Django checkout agree on a visitor's price.

A marketing site with no backend can use no-code checkout links instead. If you sell worldwide, regional prices give each country its own price, set by hand or from 5 tiers by World Bank price level; in a test, each visitor keeps their variant at that variant's price for their country.

If Leverprice is slow (no signed config within 500 ms), down or switched off with LEVERPRICE_DISABLED=1, getPrice() returns your fallbackPrice and doesn't throw, so checkout keeps selling at your normal price.

Revenue comes from Stripe, Polar, Lemon Squeezy or Dodo Payments through a webhook signing secret, which can confirm events but cannot move money. Leverprice never sees API keys or card data. Lemon Squeezy stops processing payments on March 31, 2027 (announcement); after you move, connect the new provider.

Coding agents get a ready prompt on the dashboard's Install page for Claude Code, Codex or Cursor, and npx leverprice setup connects them to Leverprice's MCP server to read results and lock in winners.

How do you set up a SaaS price test?

  1. Create a project and put its test key in LEVERPRICE_SECRET_KEY on your server.
  2. Install the SDK and add the plan with its fallback price and range, as above.
  3. Call getPrice() where the price is shown and again where the checkout is created. Charge quote.price in quote.currency, pass quote.metadata whole and append every quote.setCookie value; the docs have the code for each language.
  4. In Payments, add the webhook URL to your provider and paste back the signing secret.
  5. Choose a goal (revenue, sales or profit) and your prices, run it in test mode, then switch to the live key. When the engine calls a winner, lock it in; it's live for everyone in about a second.

Which Leverprice plan fits a SaaS company?

PlanPriceWhat you get
Free$0Manual A/B split, 1 product, 1 experiment, up to $500/month in tested revenue
Pro$29/month, or $199/year (about $17/month)Automatic optimization, unlimited tested revenue, up to 20 sites
Lifetime$299 once, 50 spotsEverything in Pro, forever

Pro pays for itself when the better price adds more than $29 a month: if the plan you test brings in $2,000 a month from new customers, a 1.5% lift covers it. Paid periods are non-refundable, so start on Free or monthly.

Frequently asked questions

Does Leverprice work with subscriptions?

Yes for the price itself: your server creates the subscription checkout with the tested amount. Polar verifies a subscription's first payment through its signed webhook. Stripe verification covers one-time Checkout Sessions only, so with Stripe subscriptions you report the first payment with trackPurchase() and it shows as unverified. Set the project's Learn from setting to Reported so the engine learns from it. Dodo Payments verifies one-time payments only.

Does the test account for churn?

No. Each price is judged on its first payment, minus refunds; renewals don't count. After you lock in a higher price, compare churn for the new cohort against the old one for two or three billing cycles.

Will existing customers see a different price?

Their subscriptions keep billing at whatever your provider has on file. On the pricing page they'd see a test price like any visitor, so skip the getPrice() call for signed-in customers to keep them on the standard price.

How long does a SaaS price test take?

It depends on sales volume and how far apart the prices are. In our simulation of the production engine, a product making 300 sales a month found a price that earned 20% more per visitor in a median of 23 days, and one making 1,000 sales a month took 8 days (full results).