Forecasting in Fundamental L/S Equity

Multi-year stacks, defensible assumptions, revenue decomposition, alt data

It is painful to see fundamental long/short analysts with decades of experience rely on regressions or overly simplistic assumptions to make forecasts.

Even worse, many omit important fundamental factors when building their projections.

Making forward projections is one of the most important aspects of the job and is a highly analytical, deeply fundamental process.

Quants or systematic PMs often talk about using different types of regressions and more recently AI agents to iterate and identify the combination of models with the highest predictive accuracy for a given KPI. But you are not a quant. Your competitive advantage is being a fundamental investor, so be that.

While we often talk about using the model to quantify your variant view, you can also build incredible variant views just by going through the full end-to-end fundamental process of making a forecast.

Remember that an important part of driving variant views (in The Process Top L/S Analysts Use to Turn Models Into Ideas) was linking the bull and bear case to the relevant KPIs and making thoughtful forecasts.

It will also be interesting to hear what Arun has to say about forecasting in our live session on Monday (calendar invites are rolling out).

Introduction

Just to preface from the beginning, we will frame the core ideas from a specific sector angle, explaining how to deal with noisiness, particularly when there is no clear underlying trend. But almost everything we discuss here applies more broadly across any sector, company and trend scenario. The terminology might change, but the underlying logic and process remain the same. You will almost always work with growth rates when making forecasts and it is incredibly important to be fluent in them. Later in the post, we will discuss how everything generalizes and touch on the use of alternative data in forecasting (when and how it matters).

Multi-year stack vs regular growth rates

We need to quickly clarify a few definitions so we don’t get confused (feel free to skip this section if you are already familiar with growth rates, multi-year stacks and their definitions, but please make sure you know the approximation formula we derived well!).