Software monetization is how a software company turns the value its product creates into revenue: what it charges for, how it packages that, and how it bills for it.
In every pricing project I run at Potio, the team arrives thinking pricing means (mostly) price points. It doesn’t. The structure underneath (what you charge for, how you package it, which model you bill on) is about 80% of the game. Get the structure right and the price points are easy to set. Try to set price points without it and you can’t.
This guide covers the six layers of a monetization strategy, the main models and how to pick one, what happened when Adobe, Unity, Salesforce, BMW and others changed how they charge, what AI changed in 2025 and 2026, how open source companies make money and how to switch models without losing customers.
Software monetization is the system a software company uses to get paid for its product. It has two halves:
The term itself came from the infrastructure side. In the early 2010s, licensing vendors SafeNet (now Thales) and Flexera (now Revenera) used “software monetization” as the umbrella label for copy protection, licence keys and compliance. More recently, billing companies like Paddle, Orb and Schematic have pulled the term toward subscription billing and usage metering, which is still infrastructure.
Why pricing architecture deserves more attention: in a Price Intelligently study of 512 SaaS companies, a 1% improvement in monetization lifted revenue by 12.7%. The same 1% improvement in retention lifted it by 6.7%, and in acquisition by 3.3%.
A software monetization strategy is a stack of decisions, and each layer limits the one above it. I build every pricing architecture from the bottom up:
Layers one to four are the pricing structure. That’s the 80%. Most teams start at layer six because it’s the only layer they can see on a competitor’s pricing page.
There are about a dozen software monetization models, and most software companies in 2026 combine two or three of them.
| Model | You charge for | Fits when | Breaks when | Example |
|---|---|---|---|---|
| Perpetual licence | The right to use a version, often plus yearly maintenance | Buyers want to own the software, run it offline or on-prem | You need recurring revenue to fund continuous updates | Adobe Creative Suite until CS6 |
| Flat subscription | Access for a period, one price | One product, customers of similar size and use | Customers vary a lot in size or usage | Setapp’s Mac app bundle |
| Per seat | Each user | Each user gets their own value from the product | AI does the work and customers need fewer people | Slack |
| Tiered | Plans with different features and limits | You serve light and heavy users | Tiers are split by features nobody pays more for | HubSpot Starter, Professional, Enterprise |
| Usage-based | Units consumed, such as API calls, messages or compute | Usage tracks value closely and buyers accept a variable bill | Customers need a predictable budget | Twilio, Snowflake |
| Hybrid | A platform fee or seats, plus usage | Cost to serve varies a lot between customers | The fee and the usage part don’t relate to each other | Salesforce Core with bundled Flex Credits |
| Credits | Prepaid units drawn down across features | Many features with different costs to serve, common in AI | Customers can’t tell what a credit buys | HubSpot Credits |
| Output-based | A completed result, such as a resolved ticket | The result is countable and attributable | Vendor and customer disagree on what counts | Intercom Fin, $0.99 per resolution |
| Outcome-based | A business result, such as revenue gained | Almost never in practice | The outcome is too far down the value chain to measure and attribute | Rare |
| Ad-supported | Users’ attention, paid for by advertisers | Consumer apps with very large audiences | B2B, or small audiences | Free consumer apps |
| OEM and embedded | A licence for another vendor to ship your software inside theirs | Components, SDKs, engines | The partner owns the customer relationship and the price | Embedded SDKs |
| Open core and managed cloud | Enterprise features or hosting on top of free open source | Open source projects with real adoption | Cloud providers can host the free version themselves | MongoDB Atlas |
What companies use depends on who you ask and what they count:
Per-seat pricing is shrinking but nowhere near dead: a third of signed contracts are still priced per user.
I go through each of these models with current examples in my guide to SaaS pricing models.
Find the value metric first. It shows you the model, not the other way round.
The value metric is the unit you bill on: a seat, an API call, a credit, a resolved ticket. A good one passes three filters:
I go deeper on running those filters in how to find your SaaS value metric. Once you have a metric, these questions narrow the model down.
Usage-based pricing generally works better in B2B than B2C. A business can come out ahead when a bigger bill means more value, and a seasonal business likes paying less in quiet months. Consumers mostly want predictability. The exception is the laundromat: you happily pay per wash because you expect to, and because you can see the running cost of water, electricity and machine time.
The more vertical the product, the closer to value you can price. A product built for one industry can charge on a metric that industry cares about. A horizontal product has to pick a metric that satisfies a very wide set of customers, which pushes it back toward infrastructure pricing: tokens, compute, API calls. Eventually that’s electricity.
Bursty usage (a solo user, a one-off project) suits top-ups, so you monetize flexibility. Steady usage (a team that works in the product every day) should push customers to a higher tier, so you monetize predictability. Selling top-ups to steady users trades good recurring revenue for one-off cash.
Most clients come to me already wanting credits or usage-based pricing because that’s what everyone is talking about. For some of them, seats are still the right model. At Toggl we priced by seat because every user got their own value from tracking their own time, and that logic still holds for plenty of products. A model someone else uses was chosen for their customers, not yours.
Yes. One client, a North American automotive software company, sold API access on usage. We added a higher tier where the same API call returns enhanced data that is hard to get elsewhere, priced 50% higher per call. Over 30% of new customers chose it. The model stayed usage-based, but packaging gave customers a reason to pay more per unit.
And expect to change your model. Between zero and $1M in annual revenue, most companies go through at least three pricing architectures: the launch version, a better one around $100k based on what they learned, and another at around $1M.
The pattern across public cases is fairly consistent. Changes that move the price closer to the value customers get, and that respect what customers already paid for, tend to work after a rough patch. Changes that alter the deal retroactively, or charge customers for something they believe they already own, tend to get reversed.
Adobe is the textbook case. In 2013 it stopped shipping new versions of Creative Suite and moved everyone to Creative Cloud subscriptions. Adobe warned investors that revenue and profit would drop while upfront licence sales turned into monthly payments, and they did. But in that first year paid subscriptions grew from 300,000 to 1.4 million, and recurring revenue from its creative products went from $155 million to $768 million. Two years later total revenue was above where it started. Seven years later it had roughly tripled, to $12.9 billion.
The later chapter is less flattering. In 2024 the US Federal Trade Commission sued Adobe over hidden early termination fees and how hard it was to cancel. Subscription revenue held in place by a painful cancellation flow doesn’t last.
Autodesk stopped selling new perpetual licences for most of its products in 2016 and moved to subscription only. Atlassian ended support for its self-hosted Server products in 2024, pushing self-hosted customers to Cloud or Data Center subscriptions, to plenty of grumbling from sysadmins who liked owning their install.
The lesson from all three: expect a dip in reported revenue while upfront licence sales turn into monthly subscription revenue, and have the cash to survive it.
In 2023 Unity announced a Runtime Fee that charged studios per game install above certain revenue and install thresholds. It applied to games already built on Unity, which changed the economics of projects that had been in development for years. Developers had alternatives in Godot and Unreal, and plenty of them said so loudly.
Unity revised the terms within weeks, and CEO John Riccitiello retired a month later. A year after the announcement, the new CEO, Matt Bromberg, cancelled the Runtime Fee entirely and went back to seat subscriptions with a price increase.
Per-install pricing wasn’t absurd as a value metric. Applying it retroactively to work customers had already committed to was the mistake.
In 2022 BMW started selling heated seats as a monthly subscription in some markets, on cars that already had the heating hardware installed. Customers saw it as paying twice for something sitting in their car. BMW dropped the idea the following year.
Tesla went the other way with a feature people already understood as software. It had offered Full Self-Driving both as a one-time purchase and as a $99 monthly subscription, then stopped selling the one-time option in the US in early 2026. Mercedes sells a yearly subscription that unlocks extra acceleration on some electric models.
The difference is perception. Paying monthly for software that keeps improving feels reasonable. Paying monthly for a heater bolted into your seat doesn’t.
Salesforce went through three Agentforce pricing models in under two years:
So Salesforce kept seats and put a usage pool on top of them: a hybrid model.
HubSpot moved its Customer Agent from $1.00 per conversation, charged whether or not the agent solved anything, to $0.50 per resolved conversation in 2026, paid in HubSpot Credits. It also added credit spend caps so customers don’t get surprised by the bill.
Cursor changed its $20 per month Pro plan in 2025 from a fixed number of requests to usage-based limits, to cover variable model costs. Users were caught out by unclear limits and unexpected charges, and the company publicly apologized. The change itself was defensible. The communication wasn’t.
| Company | Change | When | What happened | Walked back? |
|---|---|---|---|---|
| Adobe | Perpetual licences to Creative Cloud subscription | 2012 to 2013 | Revenue dip, then far higher recurring revenue. FTC lawsuit over cancellation in 2024 | No |
| Autodesk | Perpetual to subscription only | 2016 | Completed the switch | No |
| Atlassian | Server end of life, push to Cloud and Data Center | February 2024 | Migration with sysadmin resentment | No |
| Unity | Added a retroactive per-install Runtime Fee | September 2023 | Developer revolt, CEO retired | Yes, cancelled September 2024 |
| BMW | Heated seats as a subscription | 2022 | Backlash over paying twice for hardware | Yes, dropped 2023 |
| Tesla | Full Self-Driving to subscription only in the US | February 2026 | One-time purchase ended | No |
| Salesforce | $2 per conversation, then Flex Credits, then credit-bundled editions | 2024 to 2026 | Seats kept, AI usage pooled per org | Partly, per-conversation pricing replaced |
| HubSpot | Customer Agent from per conversation to per resolution | April 2026 | Price halved, charged only on resolution | Partly |
| Cursor | Pro plan from request quotas to usage-based limits | Mid-2025 | User backlash, public apology | Clarified |
AI brought back marginal cost. Classic SaaS was built on the assumption that the next customer costs almost nothing to serve. Every AI request costs real money in inference, and that changed which monetization models are safe.
Public markets reacted too. In early 2026, investors sold off seat-based software stocks over fears that AI agents would mean fewer seats per customer. That risk applies to any company whose revenue scales with headcount while its product cuts the headcount needed.
For AI products specifically, I cover model choice, credit design and margin protection in my guide to AI pricing models.
Yes. Open source companies make most of their money from managed cloud hosting and open-core enterprise features. Support contracts alone rarely build a big business.
MongoDB is the clearest public example. It reported $2.46 billion in revenue for fiscal 2026, up 23%, and its managed cloud product Atlas made up 73% of that. Red Hat is the exception that built a large business on subscriptions and support for enterprise Linux, and IBM paid $34 billion for it in 2019.
The main open source monetization models:
That last model is where the drama has been. Several infrastructure companies changed their licences to stop cloud providers selling hosted versions of their software, and the community forked the projects:
| Project | Licence change | Response | Where it stands |
|---|---|---|---|
| MongoDB | AGPL to SSPL, October 2018 | AWS built its own compatible DocumentDB | Still SSPL |
| Elasticsearch | Apache 2.0 to SSPL and Elastic License, 2021 | AWS forked it as OpenSearch | Added AGPL as an option in August 2024 |
| Terraform (HashiCorp) | MPL to Business Source License, August 2023 | Community fork OpenTofu under the Linux Foundation | Still BSL. IBM completed its HashiCorp acquisition in February 2025 |
| Redis | BSD to RSAL and SSPL, March 2024 | Valkey fork under the Linux Foundation, backed by AWS and Google | Added AGPL with Redis 8 in May 2025 |
Elastic and Redis both went back to an open source licence option, and neither won back the forks. The lesson is the same as Unity’s: changing the terms on something people built on costs trust, and you don’t get it back by changing the terms again.
Trials, freemium, discounts and billing terms sit on top of the pricing model. They tune conversion and cash flow. They don’t fix a wrong model.
There are plenty of benchmark numbers for trial and freemium conversion online. Most of them come from secondary compilations with no stated method, so I don’t trust them enough to repeat. Your own funnel data beats an industry average.
New customers have no anchor. They see your new model and judge it on its own merits. Existing customers compare it to what they paid last year, however fair the new model is. That’s why changing your monetization model is mostly an existing-customer problem.
What works:
Pick the tools after the pricing model, not before. The tools enforce your pricing decisions. They shouldn’t make them for you.
The stack has four layers:
The market is consolidating fast. Between 2025 and 2026, Stripe bought Metronome, Adyen bought Orb for $335 million and Zuora was taken private.
Two rules from experience:
Software monetization is mostly a pricing decision. Licensing and billing tools matter, but they enforce a decision made earlier.
Start with the value metric, let it point you to the model, and build the packaging around it. When you change the model, protect existing customers from the shock and never rewrite a deal they already signed. Then pick the tools, and pick them carefully.
Software monetization is how a software company turns the value its product creates into revenue. It covers pricing architecture (what to charge for, how to package it, which pricing model and price points) and the billing infrastructure that enforces it: licensing, entitlements, metering and billing.
Start from the value your customer gets and find a value metric that grows with it, such as users, API calls, credits or resolved tickets. Choose a pricing model that fits that metric and your customers’ tolerance for variable bills, package features into tiers and add-ons, then set price points. Pick licensing and billing tools last.
Hybrid models that combine a subscription or seats with usage or credits are the most common direction. Vertice’s Q2 2026 contract data shows SaaS contracts split roughly into thirds: 36.5% mainly usage, 32.4% per user and 31.1% hybrid. Among AI products, ICONIQ found 57% charge a subscription or platform fee and 42% charge on consumption.
No. About a third of signed SaaS contracts are still priced per user, according to Vertice’s Q2 2026 data. Seats still work when each user gets their own value from the product. They break when AI does the work and customers need fewer people, which is why companies like Salesforce now pair seats with a pool of usage credits.
Yes, mainly through managed cloud hosting and open-core enterprise features. MongoDB made $2.46 billion in fiscal 2026, with its Atlas cloud service at 73% of revenue. Support subscriptions work too, as Red Hat shows, but few companies have built a large business on support alone.
Only if your software runs on the customer’s machine, as with desktop, on-prem or embedded products, where you need licence keys and activation. A SaaS product controls access through accounts, so it needs a billing and entitlements system rather than a licensing platform.
Potio is a pricing consultancy for SaaS and AI companies.
Get a complete pricing redesign
and a clear rollout plan in one structured sprint.
