Pricing Friendsby hy, homeTalk to hy

Pricing strategy · 7 min read

Software pricing models: which ones exist and when each fits

Partner and Co-CEO, hy

Published
Updated

Key points

  • A software pricing model involves two decisions: how the customer pays, such as by subscription or consumption, and what they pay for, which is the pricing metric. The metric comes first.
  • The right pricing model for your software depends on whether costs and customer value grow with usage and how much predictability customers need.
  • A hybrid pricing model of base fee plus consumption splits the usage risk. 63% of the SaaS and AI vendors surveyed expect it to lead in two years.
  • Per-user pricing stays common in software because it gives budget certainty. When AI drives ongoing costs, a second, usage-based metric is often added.

If you sell software, your pricing model comes down to two decisions: how the customer pays and what they pay for. For the how, there are four basic forms: one-time purchase, subscription, hybrid and pure usage-based. For the what, you choose the pricing metric, such as per user, per transaction, per credit or per outcome. Which combination fits depends on whether costs and customer value grow with usage and how much predictability your customers need. This article is for founders, CEOs and product leaders who want to choose or review their model.

Pricing model and pricing metric are two decisions

The pricing model describes how you bill: once, as a subscription, by consumption or as a mix. The pricing metric is the unit the customer pays for. It defines what gets paid for; the amount comes later: per user, per message, per processed document or per completed task. The metric decides two things the price alone can't control: whether revenue grows with customer value and whether it grows with variable costs.

The order matters. First the unit, then the number. I see three basic forms of metrics. Here's how I put it on the podcast: “The big variants are still user-based, transactional, meaning usage-based, or outcome-based. I think those are the three basics, and then there are variations of them.”

The four billing models compared

Model How the customer pays Predictability for vendor / customer Cost risk for vendor / customer Main model 2026 Expected in two years
Purchase One-time payment, regardless of usage low / high low / low 22% 3%
Subscription Fixed fee per period high / high high for heavy users / low 40% 22%
Hybrid Base fee plus consumption medium / medium medium / medium 34% 63%
Usage-based Consumption only low / low low / high 4% 13%

Source: SaaS & AI Pricing Report 2027, p. 29 (assessment) and p. 30 (survey of 153 companies).

The table shows the core of the choice: Someone carries the risk of fluctuating usage. With a subscription, the vendor carries it when individual customers use a lot. With a pure usage-based model, the customer carries it and has a hard time budgeting costs. The hybrid model splits the risk. That's why 63% of the 153 SaaS and AI vendors in our survey expect it to be their most important model in two years.

Bar chart from the report: subscription as the main model falls from 46% (2025) to 40% (2026), hybrid models rise from 27% to 34%, and 63% expect hybrid models to lead in two years
From the SaaS & AI Pricing Report 2027 by hy, p. 30. Download the report for free

Common pricing metrics

For the SaaS & AI Pricing Report 2026, we asked 180 software companies how they bill. Multiple answers were allowed.

Pricing metric Share of vendors
Per user (seat) 54%
Per feature package 41%
Flat fee 32%
Per transaction 19%
Per API call 8%
Per compute unit 4%
Several metrics combined 59%

Source: SaaS & AI Pricing Report 2026, p. 19.

In our SaaS & AI Pricing Report 2027, respondents see a clear shift ahead. As the main metric, the seat loses 31 percentage points. Outcomes gain 20, credits or tokens 14 and API consumption 9 percentage points. 25% haven't set their future metric at all yet.

Chart from the report: change in the future main metric in percentage points, outcome plus 20, credits or tokens plus 14, API consumption plus 9, workflow plus 3, seat minus 31; 25% have no target metric yet
From the SaaS & AI Pricing Report 2027 by hy, p. 32. Download the report for free

Which model fits when

If … the better fit tends to be … Example or evidence
value grows with the number of people working in the tool, and each extra license costs you almost nothing Subscription per user CRM, a case where per user makes sense, according to Timo Müller
customers want budget certainty but usage fluctuates Hybrid: base fee plus consumption 63% expect this to be the most important model
usage and cost per customer fluctuate strongly, such as with compute-heavy AI tasks Usage-based or credits Video creation, data analysis, occasional batch processes
the outcome is clearly measurable and can be attributed unambiguously to your product Price per outcome Three conditions must be met
you deliver once and customers run it on-premises Purchase with maintenance Only 3% still expect this to be the most important model

Why per user stays so common even though it often reflects value poorly: It gives budget certainty. Buyers don't have to ask three times whether they'll get more money. I once said on the podcast: “There are so many reasons why per-user pricing works and why fixed pricing works too.”

I'm cautious about price per outcome. It needs a very narrow, repeatable use case that's fully measurable. Otherwise you end up arguing about whether the outcome was achieved. Intercom shows what such a price looks like: its AI agent Fin costs $0.99 per outcome. Billing covers the conversations Fin resolves on its own. Investor Oliver Pabst gave an order of magnitude: about 5% of AI companies will adopt outcome-based pricing, meaning a price per outcome.

Why AI changes the choice

Classic SaaS products scaled at almost no additional cost. With AI, every interaction costs real money. At the same time, fewer people often work in the software, because agents take over tasks. Julius Göllner of ARRtist summed it up this way: when AI drives ongoing costs and fewer people use the software, a pure per-user model hits its limits.

I expect many vendors to solve this with a second, transactional metric next to the seat. I like to compare it with the phone plans of the past: “You pay €25 a month and 3 cents per minute of calls, roughly.” Later I went further: “I have a strong hypothesis that AI pricing, meaning token pricing, will develop exactly the way telecom pricing has over the last 30 years.” Whether the switch pays off for you is shown in the comparison of seat and usage pricing. How you combine a base fee and usage is the second question.

What you can check on Monday

  1. Write your current metric in one sentence: “Our customers pay per …”
  2. Put revenue and variable costs per customer side by side. Do both grow with usage, or does margin drop with your most active customers?
  3. Ask five existing customers how they measure the value of your product. Does that match your metric?
  4. Check whether your customers can budget for a variable bill at all.

At hy, billing model and pricing metric are two separate steps in our approach to pricing models, before the price level.

Bottom line: First the metric, then the model, the price last

There are four billing models and a handful of common metrics. The right model follows from how costs and value grow with usage for your customers and how much predictability they need. The market is moving toward the hybrid model: 34% use it as their main model today, and 63% expect it to lead in two years.

The free pricing audit on pricing-friends.de shows where your own model stands today in about a minute. It reads your pricing page live, checks pricing model, packaging and price points, and gives a score from 0 to 100 with findings. It checks against the approach hy Consulting Group developed in more than 600 pricing projects.

Sources

Quotes are translated from German. The podcast episodes are in German.