The Death of Stickiness
Modern software companies thrive on the back of recurring & predictable revenue, high gross margins, and zero marginal cost distribution. But the best thing about software is its stickiness— high switching costs that keep net revenue retention at 120%+ and help ensure sky high ARR multiples in the public markets.
But high switching costs for software are trending towards zero, and it won’t be long before software isn’t sticky at all.
Consider the core sources of switching costs & stickiness today inside an enterprise software deployment:
Integrations: Most products integrate with some combination of data sources (e.g., API integrations), other products (e.g., through ETL / reverse ETL), or systems (e.g., OS-level integrations). These integrations are complex and expensive to build & maintain. This turns out to be a huge source of stickiness, so much so that companies like Mulesoft, FiveTran, Airbyte, and Merge have built entire businesses around solving it for customers.
Implementation & customization: Customizing a piece of software for your company’s specific infrastructure, employee base, IT & security landscape, user interface, etc. requires significant technical investment. Famously, implementation costs for large systems like ERPs and CRMs can 2-3x the total cost of ownership in the first year.
Data migration: Moving data from one system to another is non-trivial. For example, the exact data fields in your current CRM tool may not match the exact data fields in a new CRM you want to migrate towards. Historically, this has required a lot of human input (read: cost) to manage migrations of data from one service to the next. As a result, enterprises may stay with their existing software tools so they don’t risk losing data.
UI/UX lock-in: Users learn to love certain aspects of the applications that they use. For complex functionality, this is especially true. Employees will scream & yell at their leadership when their company considers replacing a software tool they love.
But software stickiness based on these switching costs is a thing of the past. Consider how an LLM-powered workflow that can understand data and write code could easily automate away the common sources of stickiness:
Integrations: LLM-powered code generation with access to product or API endpoint documentation could write and maintain integrations between applications. It could even build integrations with apps that don’t have APIs by using an LLM to process the DOM or app GUI.
Implementation & customization: Implementation and customization is about tweaking existing software or creating enablement software to support an existing tool. This seems like a perfect use case for LLM-powered SWEs to tackle, with a constrained domain & clear outcomes.
Data migration: LLMs are perfect tools for fuzzy matching & mapping data from one source format into another.
UI/UX lock-in: Real-time code generation models are already making UI/UX more dynamic & personalized. It won’t be long before the entire UI/UX is dynamic. Any application could render whatever UI a user desires and route functionality on the back end.
These sources of stickiness are gradually going to erode as the cost of integration, code generation, and migration goes to zero. And when that happens, many software companies may be in big trouble. Sub-par software products will see customer lifetime value (LTV) fall as customers churn to better alternatives. Driving growth may become more expensive as sales & marketing organizations are forced to rely more on new logos instead of 120%+ net revenue retention. Enterprises have always hated vendor lock-in, and now may be their opportunity for vengeance against software products they’ve always wanted to churn off.
That said, there is a huge opportunity to be had by a new wave of anti-stickiness startups. To name a few examples:
AI-native Systems Integrators & data migration tools like Doyen and Composio
Code & platform migration tools like Mechanical Orchard, Bloop, Cognition, and Grit
If you’re a startup building in any of these areas, or that has faced any of these challenges, reach out! I’d love to talk.
Chris Paik wrote an article a year ago called “The End of Software”. While I’m not sure software is ending, I do think that a lot of the structural advantages of software may be changing. The Great De-Stickification is upon us, and may the best products win.