Introduction:
What is Usage-based billing and what are the benefits of usage-based billing for AI start up’s? Usage-based billing or Metered billing is the practice of charging users a variable cost depending on how much of a specific unit they use. Examples of a unit could be requests, bandwidth, storage, compute time ect. In this new paradigm of applications built on top of large language models, a new cost unit appeared, the token. Most companies have a variable COGS (cost of goods sold) with pay per 1M tokens and charge customers a flat rate per month. When you add this problem to the fact that model providers give no insight into user level costs, you have no visibility into your costs, margins, profitability or user abuse. Do you have super users who cost you more than they pay you?
In this article I will go over the benefits of switching to a usage-based billing model, the potential risks or downsides, and how you can easily set it up.
Section 1: Understanding Usage-Based Billing
Usage-based billing for AI companies
Depending on the type of AI product you are selling you can approach usage based billing in a number of ways, the most common will be either token based, request based or credit based. I will go over each of these briefly and an example of when you might want to use each of them.
Token based
This method is when you use a pricing model that mimics the pricing of your underlying LLM. This makes the most sense when your customers are already familiar with pay per token models and would likely be switching to your product or service from another common model like Open AI’s GPT or Anthropic’s Claude.
Request based
This is much more familiar as it has been used by other SaaS companies for years keeping it straightforward and reducing any education you need to provide your customer. However, there will still not be a fixed margin as each request may use a different number of tokens.
Credit Based
More recently I have seen a lot of AI companies move to a credit based payment system where their customers prepay for a number of credits which they redeem for certain AI related takes such as text or image generation.
Usage-based vs flat-rate billing
Flat-rate billing is the practice of charging a fixed fee per month for a given set of features, for example charging $20 per month for Chat GPT. Usage based billing is charging a cost for a specific unit or measurement (requests, tokens, active users ect), an example of this is the Open AI API which charges developers $10-$40 for every 1M tokens used. In a lot of cases usage based billing is more value aligned because the more a customer uses a product, the more value they are extracting and they are charged accordingly.
Section 2: Benefits of Usage-Based Billing for AI Startups
1. Flexibility for Customers: It allows customers to pay for exactly what they use, making your services more attractive to those who may not need large, flat-rate packages. This flexibility can be particularly appealing in industries where usage can significantly vary from customer to customer.
2. Cost-Effectiveness: For some customers, particularly small businesses or individuals, usage-based pricing can be more cost-effective than traditional pricing models. This can make your services more accessible and appealing to a broader range of potential clients.
3. Competitive Pricing: It enables more competitive pricing strategies. You can adjust rates based on actual usage patterns and potentially offer more aggressive pricing for lower usage tiers, undercutting competitors who only offer flat-rate pricing.
4. Increased Transparency: Usage-based billing can increase transparency with customers, as they can see exactly what they're being charged for. This transparency can build trust and lead to higher customer satisfaction and loyalty.
5. Incentivizes Efficient Use: For services where resource conservation is important, usage-based billing can incentivize customers to use resources more efficiently. This is particularly relevant in industries like cloud computing, utilities, and telecommunications.
6. Scalability for Customers: Customers can scale their usage up or down based on their needs without having to switch plans or providers. This flexibility can be a strong selling point, especially for growing businesses or those with fluctuating demand.
7. Data-Driven Insights: Implementing usage-based billing requires tracking usage data, which can provide valuable insights into how your services are used. This data can inform business decisions, product development, and personalized marketing strategies.
8. Aligns Cost with Value: Customers perceive a direct correlation between the costs they incur and the value they receive, which can enhance their perceived value of your service. This alignment can improve customer satisfaction and loyalty.
9. Potential for Higher Revenue: While some customers might pay less with usage-based billing, others who use your services more intensively may end up paying more, potentially increasing your overall revenue.
10. Differentiation: Offering a usage-based billing option can differentiate your business in markets where flat-rate billing is the norm. This differentiation can be a key factor in customer decision-making.
Section 3: Implementing Usage-Based Billing in Your AI Startup
Props AI makes implementing usage based billing effortless. It takes less than 5 minutes and only a few lines of code. We act as a Proxy between your application and the LLM and track token usage, number of requests and cost so you can bill your customers on any of these unit level metrics.
We integrated seamlessly with Stripe and send over all of the usage information about each of your customers every day. Additionally, our Stripe App allows you to see costs and request count within the Stripe platform itself.
Implementing props is done by replacing the base url of your model with our base url and adding your own customer ID to each completion. We have online tutorials on youtube or our team will happily walk you through it on a call.
Conclusion:
The blog post outlines the benefits and implementation of usage-based billing for AI startups, emphasizing its superiority over flat-rate billing, particularly for services built on large language models. Usage-based billing, also known as metered billing, charges customers based on their consumption of specific units like requests, bandwidth, storage, compute time, or tokens, aligning costs directly with usage. This model offers flexibility, cost-effectiveness, and competitive pricing, as it can adjust rates based on actual usage patterns. It also increases transparency, incentivizes efficient use, and provides scalability for customers, which can lead to higher customer satisfaction and loyalty.
The post introduces Props AI as a solution that simplifies the transition to usage-based billing for AI companies. Integration with Stripe allows for seamless billing and insights into costs and usage. Implementing Props AI is straightforward, requiring minimal code changes and offering support through tutorials and direct assistance.