The firm: Champaign Capital Research#
Every chapter of this book is set at one company. Read this page first: the people, the product, and the plan are the same from chapter 0 to chapter 9.
The business#
Champaign Capital Research is an independent equity research firm in Champaign, Illinois, with a second office in Chicago. It employs 38 people: 14 analysts, 6 associates, 2 compliance officers, a data team of 3, sales, and operations. About 120 institutional clients, mostly pension funds, endowments, and mid-sized asset managers, pay an annual subscription for its research on three sectors: industrials, agricultural equipment, and semiconductors.
The product is the written note. A note takes an analyst between two days and two weeks: pulling filings and vendor data, re-running a model, writing the view, getting compliance sign-off, and publishing. Clients then call with follow-up questions, and answering those is where a surprising share of the week goes.
The problem#
Three things eat the firm’s time, and none of them is the hard part of the job.
Lookups. A client asks for Deere’s revenue growth last quarter and how it compares with Caterpillar. The answer is in a filing the analyst has already read. Finding it, checking it, and writing it back takes forty minutes, thirty times a week across the firm.
Handbook checks. Can an analyst quote a vendor’s raw estimates in a note? Is a trip’s hotel within policy? The handbook is twelve pages. People ask a colleague instead of reading it, and get different answers.
Pre-clearance. Every employee must ask compliance before trading a stock in a covered sector. Elena Ruiz’s team handles about thirty requests a week by email, checking each against the restricted list, the blackout window, and the holding-period rule. It is careful work, and it is the same work every time.
Each of these is a lookup followed by a short judgment. The firm’s plan for the year is to hand the lookups to software and keep the judgments with people.
The goal#
By the end of the year, Champaign Capital Research wants its investment research workflow to run largely on its own, from pulling the data to a draft note ready for sign-off. Three pieces make that possible:
An agent that gathers figures and handbook clauses, cites where each came from, and drafts the memo or note.
Machine learning on the firm’s own data: which clients may not renew, which names deserve coverage, how a forecast compares with the consensus.
A person in the loop who approves every published number, every trade clearance and every recommendation before it leaves the firm. Software proposes; people decide.
Each chapter adds one piece, and every chapter opens with this map showing where it sits. Click a chapter to see what it automates and what stays with a person.
How a note gets made#
The plan for the year#
The firm is adopting agents and machine learning one careful step at a time, and this book follows that roadmap. Each chapter is one task that lands on the data team’s desk.
The people#
None of them are real. The problems are: the fixture data behind every cell in this book (filings figures, the policy handbook, the pre-clearance requests) is modeled on what a firm like this actually keeps.
Your first Monday#
You have just joined the data team. Your inbox has three items in it:
A portfolio manager at a pension fund client wants Deere’s revenue growth last quarter, compared with Caterpillar, by noon.
Elena has forwarded four trade pre-clearance requests from staff.
Priya asks whether the handbook lets her quote a data vendor’s raw estimates in a client note.
Chapter 1 starts with the first item. By the end of it, you will have built the thing that answers all three.