Named pricing plans vs per-server pricing: about +100% revenue
Commentary byLiftArchiveOur analysis, not theirsServer Density replaced per-server pricing with a short grid of named plans starting at $99 a month, and new monthly revenue per visitor roughly doubled, from about $0.19 to $0.41. Trial sign-ups fell 25% at the same time.
Server Density ran the test in VWO after taking pricing advice from Patrick McKenzie, and the results were published on the Kalzumeus blog.
- Monthly revenue added: $420 to $876 per roughly 2,160 visitors per arm, about +100%, or $0.19 to $0.41 per visitor. VWO showed 93% confidence, below the usual 95% bar.
- Free trial sign-ups: 150 to 113, down 25%. The drop was significant at over 99% confidence. This is a negative result on its own.
- Credit cards added: 23 to 18, a fairly similar rate, so fewer but higher-paying customers drove the revenue gain.
- Server Density reported that support requests on the new plans came mostly from paid experts at larger companies rather than hobbyists. This was an early impression, not a measured result.
The change
A per-server price calculator replaced by a grid of named plans starting at $99
Simplified mockup drawn by LiftArchive from the shared screenshots.
Control
The pricing page charged about $11 per server plus $2 per website, with lower per-server rates above 10 and 50 servers. Buyers had to work out their own price.
Variant
Three to four named plans with a $99 per month minimum, split mainly by number of users and support level, plus a custom quote for over 100 servers.
Why it likely won
Our reading of the result, not the practitioner's claim.
- The old price grew in a straight line with server count, while the value of monitoring grows much faster for serious businesses. The plans priced closer to value.
- A $99 floor filtered out hobby users who paid little and tended to need a lot of help.
- Charging by users and support level raised the price for larger teams no matter how many servers they ran.
- One plan choice is easier to buy than a price that depends on two numbers a non-technical buyer may not know.
Caveats: Revenue confidence was 93%, the test length was not shared, and only the first cohort's monthly revenue was measured, not long-term retention or upgrades. GBP to USD conversion distorts the exact dollar amounts.
Who ran it
Server Density ran this test and published the results (Kalzumeus (patio11) blog). The numbers are theirs; the summary and commentary on this page are ours.
Read the original case study on Server Density →
How to run this test on your site
- Hypothesis
- If we replace usage-based per-unit pricing with a few named plans and a higher entry price, then revenue per visitor will rise because pricing will track customer value and screen out low-value accounts.
- Primary metric
- New monthly recurring revenue per visitor
- Guardrails
- Trial sign-up rate
- Trial-to-paid conversion
- Early churn and downgrades
- Who should test it
- B2B SaaS with complex usage-based pricing and a core customer who gets far more value than they pay for. It is riskier for products that depend on a large self-serve base of small users, or with too little traffic to read revenue per visitor.
Questions
Is tiered pricing better than usage-based pricing for SaaS?
It can be when customer value grows faster than usage.
In this test named plans cut sign-ups by 25% but roughly doubled new monthly revenue per visitor.
Is it legal to A/B test prices?
Price testing is generally legal. Server Density offered the old pricing to the few new customers who asked for it, and reported no lasting complaints.
What metric should a pricing test use?
Use revenue per visitor or per cohort, not sign-ups.
A higher price often lowers sign-ups while raising revenue, as it did here.