The spectrum between outcome-based and usage-based pricing metrics: Insights from Intercom’s Fin AI and more

The question of what makes a success-based metric is something I’ve been thinking on lately – especially after reading Kyle Poyar latest State of SaaS Pricing and seeing some of the unique pricing models popping up across the software industry.

Many are touting Intercom as one of the best examples of this pricing strategy, after they announced that pricing for their AI support chatbot (Fin) would be $0.99 per successful 'resolution’ – situations where the customer either confirms their satisfaction with the answer or leaves the conversation without escalating to a human.

I agree with Kyle's distinction that this pricing at Intercom is interesting, as it’s based on a specific outcome that drives value. It’s a spectrum, and many different metrics in the usage-based space operate across this. In the post below, I’ll try to distinguish some points on the line between tangible value and price.

The spectrum between “outcome metrics” and “usage metrics”

To illustrate the blurry semantic line between metrics, let’s say that there exists a spectrum of pricing metrics between outcome-based and usage-based. (Other metrics exist outside of this range, but I’ll keep the focus narrow here.)

Any of these metrics – from pure outcome to pure usage – could be linked to some form of success, so when I talk about outcome-based, I’m focusing on the metrics most closely linked to provable and discrete ROI (return on investment) from the product in question.

Since not all metrics fall explicitly into either the outcome-based or usage-based camp, what do the different parts of the spectrum actually look like?

Left: Purely outcome-based pricing

This pricing strategy is rare in most industries. Purely outcome-based pricing is based exclusively on some metric of ROI: e.g. the revenue you gain, the costs you save, etc.?

Most frequently, this can be seen in areas such as payments, where a direct percentage of the economic value generated (in this case, revenue for the customer) is levied by the company processing the transaction. Sometimes, this is accompanied by a base fee for the transaction as well.

Across other spaces, this model is somewhat more rare, but it has been seen at times in procurement management platforms or chargeback management platforms (such as Chargeflow, which charges 25% of successful chargeback revenue). In these cases it also makes sense. The savings or revenue increase is directly tied to the product, and the better the product performs, the more the customer should pay for that service.

In many other sectors, a big issue with this type of pricing is attribution – what percentage of that ROI can you say is due to the product vs. other influencing factors??

For example, let’s say Software Company A’s pricing metric is the amount of increased revenue and they keep 5% of that total. If revenue increases by $1m during the time period, but other changes were also made beyond using the software, then who is to say it's only the software causing the increase and not marketing or demand gen or sales improvements?

The secondary impact of this is that incentives can be misaligned. If there is a gray area in terms of attribution, then the buyer of the software is incentivized to make the most of this in negotiation so that they directly reduce their cost of buying the software. If the buyer knows that the software really made them $1m (and therefore they should pay $50k), but could convince the vendor that it only made $200k – leaving their price at $10k – why wouldn’t they? As such, it can create an anti-pattern where the buyer of your software is actively arguing against, rather than for, the value it generates.

Because of this, purely-outcome-based metrics work when the outcome generated is discrete and attribution to the product is clearly provable.?

Left middle: Pricing based on valuable product actions

In these cases, the metric is tracking a product action that definitively links to an outcome, but is not a direct proportion of the value of this outcome. Often this will be because it is easy (for the vendor) to quantify and/or attribute the product impact on the outcome itself, but not the actual economic value of said outcome.?

A great and storied example of this is click-through rates in online advertising, where you pay for each successful click on an advert. That’s clearly linked to value, as that click could lead a customer to purchase, subscribe or use a product or service. But the advertiser doesn’t know whether that happened or if that customer would have performed that action regardless of the advert if it did not exist. Even if they did know, it would be hard to prove beyond a reasonable doubt.

Another good example is Zapier ‘Tasks’ metric and Intercom’s Fin AI ‘Resolutions’ metric. In this case, you would assume that the action would derive value, but how much is potentially highly variable. In the former case, a task could be as simple as avoiding a copy/paste or manual input of some data from A to B, or it could be a critical part of a hugely valuable business process. In the latter, a successful resolution could save a happy customer from having to bother Google or ChatGPT, or it could prevent a valuable and irate customer from churning from your product or service. In both cases, it would be nearly impossible for the vendor to calculate and normalize the metric for this value question.?

With so many difficulties in many sectors with charging on a purely outcome basis, these type of metrics are the most realistic way for many companies to get a more direct link between product ROI and price.

Right middle: Pricing based on a value proxy

On this part of the spectrum, the pricing metric might be one that is an indirect measure of value; for example, a proxy for end customer value that is tracked or surfaced in your product. Unlike pricing based on valuable product actions, the link between usage of the product and an increase in this metric is not a direct correlation, and there may be an indirect link in terms of attribution.

A good example of this is Segment ‘Tracked User’ metric. The entire reason you would want to use Segment is to understand and gain insights on visitors to your website or mobile apps. Segment (and other data platforms) can also help you improve your operations and advertising to improve the volume of visitors to your site. But this is not a direct link; there are many other reasons why your web or app traffic might be increasing that have nothing to do with your usage of a data platform. Similarly, analytics platform Amplitude also prices based on Tracked Users.?

‘Tracked revenue’ or ‘Costs under management’ as used by some financial services, procurement and spend management platforms is also a good example of this metric; the metric isn’t specifically targeted on the revenue increase or costs saved, but the amount of money at play is assumed to be a good proxy for how much value the customer is getting overall.

Right: Purely usage-based?

With purely usage-based metrics, customers have control over their usage without a clear link between that metric and value. Instead, they are simply purchasing usage for the product that could or could not have value.?

Examples of companies using a purely usage-based metric would include Amazon Web Services (AWS) , Snowflake and m3ter customer ClickHouse , all of whom bill based on a user’s consumption of storage, compute and cloud services. While you’d hope that customers can accurately use the product in a way that ensures it drives value, the pricing model doesn’t have any direct correlation with how much value is derived, or any measure related to value.

In many cases, especially cloud infrastructure, this is the preferred model because:

• The majority of usage is directly linked to critical infrastructure, so should be high value by design
• Usage of these products should in many cases be a fast-growing metric, so this metric provides a fast route to impressive net revenue retention
• There is a direct and clear link to cost-to-serve for that product, so it protects the vendor’s margin ?

My take: The semantic line between an outcome-based metric and a usage-based is a bit blurry, and choosing the “best” metric will still be very dependent on the company, customer and use case. One of my favorite parts of the world of usage-based pricing is that it allows companies across industries flexibility in choosing the metric that works best for their business.

In a market where adoption and ROI is increasingly scrutinized by customers, I feel we could see increased adoption of metrics such as Intercom’s ‘Resolutions’, where customers are explicitly paying for the value that they derive from the product’s actions. But it definitely won’t be the norm any time soon.

?Thoughts on how to define a success-based vs. outcome-based vs. usage-based metric? Are we in for a wave of Intercom imitators in the coming year? What other companies would you like me to add to this spectrum? Please comment below!