Coming next: The Second Factory →

Back to Library
Strategy & Focus⚠️ Pitfall

When Simple Would Have Worked (But We Got Fancy Instead)

The Problem

Analytics team builds a "Customer Health Score" with 23 variables, weighted by a machine learning model nobody understands. It spits out numbers from 0-100. Sales teams ignore it because they don't trust a black box. Meanwhile, the real signal was obvious: customers who don't log in for 30 days churn. That's it. One simple rule would've worked better than the fancy model. But we overcomplicated it because smart people like to build smart things. Complexity feels impressive. Simple feels too easy. But complexity kills adoption. If people don't understand it, they won't use it. It's like building a Swiss watch when all you needed was a sundial.

The Principle

Start simple. Solve the problem with the simplest solution first. If "customers who don't log in for 30 days churn," just track login recency. You don't need a PhD-level model. Add complexity only when simplicity fails. Think manufacturing: you don't need a robotic arm when a lever works. The magic happens when solutions are so intuitive that people adopt them immediately because they just make sense. Less wizardry, more clarity. Stupidly simple always wins over impressively complex.

🎯

3 Steps to Take Action

  1. 1

    Start with the Simplest Solution: Before building anything complex, ask: "What's the dumbest simple version that could work?" Try that first. Add complexity only when simplicity fails.

  2. 2

    Measure Adoption, Not Accuracy: A 75% accurate model people trust beats a 90% accurate black box nobody uses. Optimize for understanding and adoption first.

  3. 3

    Kill Complexity Creep: Every time someone suggests adding a feature, ask: "Does this make it easier to use or harder?" If harder, say no. Protect simplicity ruthlessly.

In Practice

A bank built a credit risk model with 47 features and a neural network. Took 9 months, cost $800K, and predicted default risk with 82% accuracy. Compliance team couldn't explain it to regulators. Got shelved. Intern built a simple model: 3 variables (income, debt-to-income ratio, payment history), logistic regression. Took 2 weeks. Accuracy: 79%. Close enough. But here's the kicker—compliance could explain it in 30 seconds, it was approved immediately, and it went into production. Prevented $15M in bad loans in year one. The fancy model sat unused. The simple model shipped. Sometimes 79% accuracy that people trust beats 82% accuracy nobody understands.

More on Strategy & Focus

From the book

In the book, one decision sits for six months while everyone waits for more analysis, and three teams build the same thing without knowing it.

This card is one of 57 drawn from The Auction Block by Brian Cariveau — the story of an analytics leader with 90 days to fix what the cards describe.

Each leadership pattern falls into a few pitfalls most often. Find your pattern.The Analytics Leadership Mirror: free, 5 minutes, no account