AI Broke Every SaaS LTV Forecast. What Can FP&A Teams Do About It?

At a dinner event recently, someone floated the idea that every SaaS company that forecasted customer lifetime value (LTV) before AI was wrong.

I had not made this connection before, but it makes sense that many companies overestimated LTV. It would have been impossible to predict the competitive threat of AI-native companies, both direct competitors and those building tools that let anyone create software.

If LTV was overestimated, then companies would have likely overspent to acquire customers. Similarly, investors overestimated the durability of SaaS revenue.

Forecasting is inherently limited, particularly when a typical FP&A team or investor looks several years into the future.

Effective forecasting usually involves one to five years of historical data, understanding the baseline, extrapolating trends, and then making logical guesses about what might change the trend.

Even if an FP&A team followed all these steps, they still would have gotten LTV wrong before AI.

Each of these deserves its own discussion, but the core issues are that most models (1) are not designed to predict multi-decade disruptions, (2) cannot always anticipate the timing and magnitude of known risks, (3) fail to capture second- and third-order effects, and (4) cannot predict Black Swans or unknown unknowns.

That said, long-term forecasting can be useful.

While forecasting for multiple years is challenging, an accurate 12 or 18-month forecast is more doable, and can serve as an early warning system for the long-term model. For example, if churn starts to tick up, this might be worth a deep dive into the causes.

Another way to improve long-term forecasting is to formalize how external data enters the planning process. When I ran competitive analysis at a prior employer, I distributed insights monthly to help leaders easily incorporate outside data into the planning process.

It’s also helpful to think of the long-term forecast more as an exploration than a prediction. The goal should be to better understand how the future may play out, so that you can plan for multiple scenarios.

Your long-term forecast may not be accurate, but it should help the company become more prepared.

In what other ways is the long-term planning process helpful for the business?