Qualtrics, the object of a 2023 LBO by Silver Lake and the Canadian Pension Plan Investment Board, has seen its term loan maturing in 2030 decline to 86 cents on the dollar from just above par several weeks ago. This is the worst price for a debt instrument because the risk is asymmetric: it is either worth 30 cents or par. This price doesn’t reflect true credit worries as some would suggest, merely a widening of risk premiums. What’s the difference? One is perceived and one is tangible.

The widening of risk premiums reflects the market’s belief that because anyone can now build software with AI agents that customers will increasingly choose to build versus buy. This will put pressure on software margins and shorten the duration of cash flows through higher churn. This ultimately puts downward pressure on the lifetime value (LTV) assumptions that investors have made. Whether a lower LTV is realized is yet to be seen but today it is reflected via wider risk premiums.

All software companies face a trade-off between delivering unique, flexible solutions at scale versus overly generalized, rigid, and monolithic solutions. Anyone who has purchased a SaaS solution knows how rare an easy to implement, flexible, and effective solution that meets your specific requirements is. This demonstrates the real value of software which is in the design. Historically, software was designed to solve a problem or automate a workflow and the assumption was that the problem or workflow was the same across all customers and a one-size-fits-all approach was applicable. That assumption did not fit with reality and in recent years the focus has shifted to different design principles that can be both effective and universally applied to a broad range of use-cases.

There are several core principles behind good software design:

1. Separation of Concerns: divide the program into distinct sections where each section addresses a separate concern
2. Single Responsibility: every module, class, or service should have exactly one reason to change and one primary responsibility
3. Encapsulation: hiding the internal workings of a component allows developers to change the component without breaking dependencies
4. Dependency Inversion: dependency is not hierarchical; modules should depend abstractions (like interfaces) of each other

These principles help guide the implicit trade-off between simplicity and scalability in software design. AI can certainly help with better design and expanded capabilities, but as anyone who has spent time interacting with LLMs has realized the output is only as good as the prompt. Despite the amazing advances we have achieved with LLMs we haven’t moved away from the “garbage in, garbage out” principle. This is why we are seeing so much emphasis on prompt engineering, agentic workflows, and markdown-based instruction files like SKILLS.md, CLAUDE.md, and so on. All of these are different ways of making the input and question better to get a better answer. As LLMs can increasingly do the repetitive, highly structured work we are seeing increasing importance being placed on how to design and orchestrate the LLMs via agents and software design.

Therefore, we can infer that Qualtrics’ credit worthiness ultimately depends on the quality of their software design and the pros and cons of whatever approaches they have taken.

Qualtrics is the global leader in Experience Management (XM), providing a cloud platform that helps organizations measure, analyze, and improve customer, employee, product, and brand experiences. From this we can gather that they probably have an “event-driven” architecture where services communicate through asynchronous events or triggers and the system is loosely coupled and highly responsive. The drawback of this architecture is that it can be challenging to debug and maintain data consistency. Therefore, we can infer that Qualtrics’ core value proposition is in its ability to distill varied and unstructured data into actionable insights for its customers. To do this well they need good data architecture that relies on good ontology (the what), morphology (the how), and typology (the which); the form and arrangement of things and how they are related.

Unfortunately, we don’t have visibility into their code base, so we are left with looking at their consumer adoption and retention rates. A good software solution will draw customers in and retain them over the long-haul. Unfortunately, the short feedback loops in the SaaS market do not provide reasonably long observation windows to adequately determine product quality. The name of the game for software companies is to jump on a wave and ride it for as long as you can.

However, prior to the recent developments in the SaaS market investors were willing to reward companies that heavily invested in future growth through things like acquisitions and R&D spending with high valuations despite low margins and/or lack of profitability. This made sense in a world where you could grow into profitability by scaling highly profitable unit-level economics regardless of their duration. Qualtrics historically operated with high-70s to low-80% subscription gross margins with average subscription terms of 1-3 years. However, Qualtrics was operating deeply in the red when Silver Lake and CPP paid $12.5 billion for the company in 2023 on the thesis that the short-term losses were investment into a future where Qualtrics was the leader in the Experience Management category and had strong AI-driven expansion potential. Qualtrics was highly valued because their high revenue and strong gross margins led to a high lifetime value.

Today, the market has repriced its churn rate assumption for the SaaS marketplace which has had a dramatic impact on the LTV of these companies, especially the highly valued and unprofitable companies because all their value was in their assumed LTV. Profitable companies like Salesforce that have already achieved profitability at scale have also felt this impact but because they have distributable cash flow from their current contracts their share prices are not as sensitive to the higher churn rate. However, the response to this repricing from all SasS companies is important to watch. If the big companies revert to the old playbook of acquiring companies that have experienced a large drop in valuations to achieve higher growth and margins in the future it is essentially a doubling down on the old churn rate assumptions. If the market is right in its new churn rate assumption the old playbook of growing into profitability will be far more difficult because the LTV is much lower.

Alternatively, if they respond by extending contract durations and increasing switching costs, maybe by offering better AI-driven solutions and integrating more deeply with their customers, they can offset the negative impacts being felt by the AI-driven repricing. The question the market has put to Qualtrics is whether has developed a unique language of customer experiences that different companies can map their businesses onto and provide unique business insights. If the answer is yes, they can retain and extend customers, if the answer is no, they will struggle to grow into profitability and earn a return on the heavy investments they have already made.

What do we mean by a “unique language of customer experiences”? Well, the are in the business of user experiences and users are similar in that they are all human and the way human beings experience things is universal. That doesn’t mean that we all experience something the same way, it only means that we take in the external world and process it in the same way. The difference is the outputs (logical, emotional, etc.) which are unique to each person. Therefore, for Qualtrics to be able to deliver valuable insights to their customers they must have designed a system that can take in and interpret highly varied data, find commonalities and relationships in that data, and infer actionable and valuable insights from that data.

Therefore, the question becomes whether Qualtrics has designed an “ontology” for human experience that yields unique insights for clients. Ontology has become a popular word in tech circles as companies like Palantir have demonstrated its use case. This is quickly becoming the new frontier for AI-driven advancements as LLMs have reached an inflection point in their capabilities. Design philosophies like ontologies, organizational frameworks, and architectures are the critical path as AI capabilities increasingly depend on systems speaking the same language and having internal logical consistency. The power of LLMs is in their ability to interpret large amounts of information but the quality of their outputs is based on customized instructions, the ontologies, which provide the framework for the customized insights.

AI helps scale, but if you’re scaling a poor logic, you’ll only compound losses.

The inherent contradiction of the “SaaSpocalypse” is that software companies are the best positioned to leverage AI to make their products better and the question is which companies have actually built a universal architecture that users can map their unique business onto and generate unique and valuable insights. The market has opened the bidding by increasing the churn rate assumption for the sector. Now we wait to see how the companies respond. Do they use the same old playbook of buying growth through heavy spending and acquisitions, or do they use this opportunity to fundamentally change the nature of their products and how they work with their customers?