GrowthPad

Most companies still treat pricing like a finance decision.

Pick a number. Create three plans. Put the middle one in a different colour. Add “Most Popular” above it.

But pricing is not just the amount customers pay. It shapes how they use the product, when they experience value, what behaviours they repeat, and whether growing with the product feels rewarding or painful.

That makes pricing part of the product experience.

In a recent Growth Roundup conversation, I spoke with Andrew Garvin, co-founder of Metronome, about how the best software companies are rethinking pricing for a world of AI agents, credits, usage-based billing, and increasingly unpredictable customer behaviour.

The big takeaway was simple:

Great pricing does more than capture value. It helps customers create more of it.

Watch the full talk here.

Pricing is now a product capability

The idea behind Metronome came from a problem that looked like billing infrastructure but was really a growth problem.

At Dropbox, the product team could design a pricing experiment quickly. The difficult part was implementing it inside the billing system and getting usable data back.

That process could take three to nine months, despite Dropbox having a large billing engineering team. 

That gap matters.

A company may have a dozen strong ideas for improving packaging, introducing credits, testing regional pricing, or creating a new usage-based plan. But those ideas are worthless if the billing system cannot support them.

Pricing flexibility has become a competitive advantage.

This is even more important for AI products. One customer might use a product five times a month. Another might connect an agent that performs thousands of actions.

The product can no longer assume that every user creates similar value or carries a similar cost.

Start with cost, comparisons, and value

Pricing discussions often become messy because every team is solving for something different.

Finance wants healthy margins. Sales wants a model that is easy to explain. Product wants fewer barriers to adoption. Customers want predictability.

Andrew breaks the problem into three parts.

First, understand your cost.

For an AI company, that could include inference, model providers, storage, data processing, support, and infrastructure. Cost gives you a floor, but it should not automatically become your customer-facing pricing model.

Tokens may matter internally. That does not mean customers want to buy tokens.

Second, understand what customers compare you against.

The comparison might be a competitor, an employee, an internal engineering team, or another category of software entirely.

Metronome, for example, could be compared with both billing infrastructure and a data platform. The most useful comparison was not necessarily another company with the same category label. It was the alternative living in the customer’s head.

Third, understand the value being created.

Does the product replace engineering work? Resolve support conversations? Generate qualified leads? Complete tasks that previously required an employee?

The strongest pricing models tend to move closer to that value.

Andrew used these three lenses to turn pricing from an abstract debate into a manageable decision. 

Do not make customers understand your infrastructure

One of the easiest ways to overcomplicate pricing is to expose too much of the underlying system.

AI products can have complicated cost structures. Different models cost different amounts. A single customer action might trigger several model calls, tools, or workflows.

That complexity is real, but it is usually the company’s problem.

Andrew shared the example of a vertical SaaS company serving golf clubs. The team wanted to model its pricing after Anthropic.

That might make sense for a developer buying access to a model. It makes far less sense for a golf professional trying to operate a club.

The customer should not need to watch a token counter to understand whether they can keep using the product.

A useful rule is:

The pricing model should reflect how the customer experiences value, not how your infrastructure produces cost.

Credits can help companies hide some of that complexity. They create a layer between changing infrastructure costs and the customer experience.

But even credits need to be simple. If customers cannot predict what an action will cost or why their balance changed, the abstraction has failed.

Gamification is really incentive design

Gamification often gets reduced to badges, streaks, points, or leaderboards.

That is not the interesting part.

The useful part of gamification is designing incentives that encourage customers to repeat valuable behaviours.

Pricing can do the same thing.

It can encourage users to return, try a feature, complete a project, invite teammates, increase usage, or move from experimentation into production.

The goal is not to manipulate customers into spending more.

The goal is to align the behaviour that creates more value for the customer with the behaviour that creates more value for the business.

That is why Lovable is such an interesting example.

What Lovable gets right

Lovable gives users recurring free credits, including people already on paid plans.

At first, that sounds strange. Why give free usage to someone who is already paying?

Because the credits are not only a pricing mechanic. They are part of the engagement loop.

A recurring credit refresh gives users a reason to come back. Each return creates another opportunity for the user to make progress on a project.

Over time, Lovable can identify the people who move from casually experimenting to seriously building and hosting applications.

Those customers are far more valuable than the average user.

But instead of immediately punishing them for increased usage, Lovable can give them additional credits at key moments. The company rewards the behaviour it wants to see.

Andrew described this as borrowing from the playbook of mobile games, where a small group of highly engaged users often drives a large share of revenue. 

The lesson is not that every SaaS product needs daily credits.

The lesson is that pricing can be used to support the habits that make customers successful.

Stop punishing customers for using the product

Many usage-based products create the wrong emotional experience.

A customer starts getting more value, increases usage, and immediately sees a warning, an unexpected bill, or a hard limit.

The product has turned success into anxiety.

That does not mean increased usage should always be free. A business still needs to capture value.

But the monetization moment should remind the customer what they achieved.

“You generated 1,000 qualified leads this month” is a much stronger message than “You have used 87% of your credits.”

Both describe the same underlying activity. Only one reinforces the value of the product.

Before asking customers to pay more, show them why paying more makes sense.

Outcome-based pricing is attractive but hard

Intercom’s Fin is one of the clearest examples of outcome-based pricing.

Instead of charging only for seats or access, Fin charges based on automated support resolutions.

That creates a straightforward value story. The customer gets a resolved support conversation. Intercom gets paid for delivering it.

Andrew described Fin as one of the strongest examples of a company successfully moving toward outcome-based pricing. He also pointed out that competitors have found the model difficult to reproduce. 

That is because outcome-based pricing introduces difficult questions.

What counts as a successful outcome? Who determines whether the outcome was achieved? What happens when the customer disagrees? How much control does the product have over the final result?

Outcome-based pricing works best when the outcome is easy to understand, objectively measurable, and strongly connected to the product.

That combination is rarer than it looks.

Pricing experiments should reduce risk

Pricing changes affect far more than conversion.

They can change positioning, customer behaviour, retention, sales incentives, revenue, and even the type of customer the company attracts.

That makes teams nervous about experimenting.

But refusing to experiment does not remove risk. It simply allows untested assumptions to become permanent.

Andrew pointed to two ways companies can reduce that risk.

The first is launching in a limited market.

HubSpot introduced parts of its move toward credits in smaller regions before expanding more broadly. A regional rollout gives the team a chance to observe customer reactions, sales friction, billing problems, and adoption patterns before committing the entire business.

The second is allowing users to experience value before introducing the meter.

When Notion launches certain AI capabilities, it may initially make them free. That gives customers time to understand the product before credits or pricing become part of the experience.

The company takes on more cost in the short term, but it learns which users adopt the feature, what they use it for, and where the strongest value appears. 

The point is not to give the product away indefinitely.

It is to avoid finalising the pricing model before you understand the customer behaviour.

Credits create new retention levers

Traditional subscription businesses have a familiar set of retention tactics.

They can offer a discount, pause the subscription, extend the trial, move the customer to a cheaper plan, or improve payment recovery.

Credits introduce another set of options.

A company can add credits when a customer is about to cancel, extend the expiry date, allow balances to roll over, or reward customers when they return.

These tactics can preserve access to the product without permanently lowering its price.

They can also make the cancellation decision feel more concrete. Customers are not simply losing access. They may be giving up a balance they have already accumulated.

In enterprise products, usage-based pricing creates a different retention opportunity.

A company can start with a relatively small commitment, allow usage to expand, and communicate value continuously as the customer grows.

That last part matters.

Value should not be explained only during a renewal meeting. The product should regularly show customers what they are achieving. 

Monetization is moving into the product and engineering organisation

Pricing used to sit between finance, sales operations, and the executive team.

At many AI companies, it is becoming a platform capability.

Monetization teams increasingly work across identity, entitlements, metering, billing, accounts, packaging, usage data, and pricing configuration.

This is happening because pricing models are becoming more dynamic.

A company may need to support subscriptions, usage, credits, outcomes, enterprise commitments, and custom agreements at the same time.

Andrew described a shift away from the traditional build-versus-buy decision.

Companies increasingly want to “build with” a platform. They want strong infrastructure, but they also want the ability to extend it, combine it with internal systems, and adapt as the pricing model changes. 

That means pricing cannot be a once-a-year project.

It requires ongoing work across product, growth, engineering, finance, data, and go-to-market.

The market will get more complicated before it gets simpler

AI companies are currently experimenting with tokens, credits, actions, outcomes, seats, usage, agents, and hybrid subscriptions.

It feels messy because the value created by AI is highly variable.

An agent can perform far more work than a traditional software user. That breaks the assumption that each seat creates roughly similar value and cost.

Andrew expects usage-based pricing to become a capability every software company needs, even when usage is not the main pricing model.

He also expects subscriptions and credits to work together for the next several years. Subscriptions provide predictability. Credits allow companies to account for changing levels of usage.

Eventually, the market may simplify again.

Mobile phone pricing followed a similar path. Customers moved from complicated metered plans toward simpler bundles once the market matured.

Software may end up in a similar place, but companies need flexibility while the model is still evolving. 

The founder mistake: pricing too low

When I asked Andrew for the most common pricing mistake founders make, his answer was immediate.

They price too low.

Early founders are often unsure of the value they provide. They discount aggressively, leave early customers on design-partner pricing for too long, and avoid monetization because they worry it will slow adoption.

But cheap or free usage can hide weak demand.

A customer willing to try the product is not the same as a customer willing to pay for the value it creates.

This is especially important in AI, where new companies can reach meaningful revenue much faster than previous generations of software businesses.

Andrew’s advice was to focus on monetization earlier. A working pricing model helps prove that the business is real before the company spends years building around weak signals. 

Pricing should make the product better

The old pricing question was:

How much can we charge without losing the customer?

A better question is:

How can pricing help the customer experience more value?

When pricing works as part of the product, free usage can build habits, credits can guide behaviour, upgrades can reflect progress, and usage can feel like success rather than punishment.

The best pricing model is not simply the one that produces the highest conversion rate this quarter.

It is the one that aligns your growth with the customer’s growth.

That is how you build a product people do not merely tolerate paying for.

It is how you build a product people love to pay for.