The Seat vs Usage based pricing conundrum

There has been a lot of debate recently about whether seat-based pricing will get *ahem* upseated by usage-based pricing when it comes to cloud software. The recent decline in earnings guidance across software companies en masse sparked the debate around whether “software is dead”, but once you get past the hyperbole, the debate is turning into “is seat-based pricing slowly dying”.

There have been three key arguments to support this notion:

Argument 1: tightening purse-strings

As large contracts are coming under a lot of scrutiny, it is becoming difficult to justify and sell big-ticket multi-year contracts. On the other hand, usage-based adoption has a relatively easier pass, as the magnitude of billing is related only to the value actually gained. One can argue that “I will need to pay more for Snowflake only if I am collecting more data from more transactions, and hence more revenues”. (Of course, the horror stories of $65m Datadog bills should be a cautionary tale that usage-based billing could quickly go out of hand). This is not a recent phenomenon though — there have been several reports as early as 2021 that point to the value of such a model.

Argument 2: AI-led obsolescence

Another strong viewpoint is that if AI can enable organizations to do more with less people, seat-based pricing companies should see a natural decline in revenues. Also, a small team of 2 could have built a product which handles more traffic than a product built by a team of 50. Seat-based pricing would be very askew with the reality of consumption (and often providers might have to take big losses on certain small-team accounts). Especially in GenAI as compute/inference is so costly, this problem gets laid to bare very quickly. Jamin Ball has a pretty nice post on this where he takes it to an interesting side-conclusion — database companies have a lot to gain from this movement.

Argument 3: AI-led complexity

A more nuanced (and lesser discussed) argument, is that while most software so far have been function-specific (for a Dev/ SRE/ SecOps) and top-down (specific roadmap built based on the needs of the many); AI-native software could flip this upside down and build tools that are more horizontal, but able to meet the needs of the few as well as the many. I would struggle to even set seat-based pricing in such a context.

But what does data say?

While all the above arguments make sense and are believable, a more nuanced question is “how quickly is this shift happening?”. One way to answer this is “are companies that offer primarily usage-based pricing, not seeing as drastic a growth-decline as subscription/seat-based companies?”. To dig into it a little further, my team categorized the top 75 cloud software companies into seat vs usage, and analyzed some growth metrics.

While both categories of companies saw median revenue growth decline QoQ, usage-based pricing companies saw a smaller decline (~50%) compared to subscription-based companies (~75%). However, this difference is not that much!

A big reason for this variation being small could be that even though usage-based bucket of companies did succeed in signing on more customers, their usage (and consequently revenue) naturally takes time to ramp-up. So I have also plotted the % of net new customers acquired during the quarter. Here again, the difference is not that stark as of now.

I believe these two metrics are a good indicator of the relative success of both models, and I intend to track this every quarter to see if any clear patterns emerge. If you have further thoughts/data to share on this topic, do reach out to me at @sahilpatwa