TL;DR
Billy Beane did not win with data. He won because he noticed that the metric everyone trusted, batting average, was the wrong metric. Companies have the identical problem. The annual review and the engagement survey are the batting average of business: familiar, comfortable, and close to useless for predicting who is actually winning. Only 2% of Fortune 500 CHROs strongly agree their performance management system inspires employees to improve. Moneyball for companies means replacing opinion-based scoring with behavioral evidence drawn from the work itself.
Key insights
- The 2002 Oakland A's opened with a $40.0M payroll against the Yankees' $125.9M, about a third of the money, and finished 103-59 with a 20-game win streak.
- Only 2% of Fortune 500 CHROs strongly agree their performance system inspires improvement, and only 20% of employees say reviews are fair or transparent (Gallup, 2024).
- 61% of managers and 72% of workers do not trust their performance management process. Just 6% of organizations use data well enough to capture what worker performance is actually worth (Deloitte, 2025).
- Global engagement fell to 20% in 2025, the lowest since 2020. The entire decline came from managers, who dropped from 31% to 22% (Gallup, 2026).
- The fix is not a better survey. It is reading the digital exhaust of work itself: who ships, who unblocks others, who is quietly drifting toward the door.
Beane's real insight was not "use data"
Every business writer who invokes Moneyball flattens it into "be data-driven." That is not the story.
Baseball in 2001 had more data than almost any industry on earth. Scouts had a century of statistics. The problem was that the statistic everyone optimized for, batting average, did not correlate well with scoring runs. On-base percentage did. A walk is as good as a hit if the point is to put a man on base, but a walk does nothing for your batting average, so walk-heavy players were systematically underpriced.
Beane did not out-spend anyone. He could not. Oakland lost Jason Giambi, the reigning AL MVP, to New York that offseason, plus Johnny Damon and Jason Isringhausen. He replaced them with Scott Hatteberg and David Justice and won 103 games on a third of the Yankees' payroll. Miguel Tejada took MVP. Barry Zito took the Cy Young at 23-5.
The honest caveat: Oakland lost the ALDS to Minnesota in five games. Measuring better does not guarantee a title. It guarantees you stop paying a premium for the wrong thing.
Your batting average is the performance review
The corporate equivalent of batting average is the annual review score. Everybody has one. Everybody optimizes for it. Almost nobody believes it predicts anything.
Gallup put the number on it in 2024: among Fortune 500 CHROs, the people who own these systems, exactly 2% strongly agree the system inspires employees to improve. Only 20% of employees say their reviews are transparent, fair, or motivating. Deloitte's 2025 study of roughly 10,000 leaders across 93 countries found 61% of managers and 72% of workers do not trust the process at all, and that only 6% of organizations use evidence effectively to capture the value of worker performance.
That is not a tooling problem. It is a measurement problem. A review score is one manager's recollection of nine months of work, compressed into a number, filtered through recency and relationship. It is a scouting report, not a statistic.
The second-order damage is worse. Gartner found in 2024 that only 8% of organizations have reliable data on the skills their workforce actually has, and that when talent is not ready for what the business needs, employee performance drops 26 percentage points. You cannot deploy people well if your record of what they are good at is fiction.
We have made the longer case for what replaces it in performance management without annual reviews.
The engagement survey is the second wrong metric
If the review is batting average, the engagement survey is RBIs: a lagging number shaped mostly by factors outside the individual's control.
Gallup's 2026 global data has engagement at 20%, the lowest since 2020 and the second straight annual decline. The most useful detail is not the headline. It is that individual contributors barely moved, 20% to 19%, while managers collapsed from 31% to 22%. The layer that is supposed to be your sensor network is the layer that is breaking, and the survey told you none of that until a year later.
Gallup also found that 57% of employees discuss their goals with a manager annually or less. So the instrument runs once or twice a year, the conversation runs once or twice a year, and the resulting picture is a snapshot of a system that changes weekly. By the time a score moves, the person who caused the move has usually already decided something.
That is the gap Hatchproof exists to close: signal before the exit interview, not a postmortem after it.
The on-base percentage of work
Walks were always in the box score. Nobody counted them because nobody thought they mattered.
The same is true inside your company right now. Every day your team generates digital exhaust: Slack threads, email, calendar patterns, commit and ticket cadence, who gets pulled into decisions, who answers within an hour and who has stopped answering. None of it is new data. All of it is evidence of the work itself, produced as a byproduct of doing the work, which means nobody is performing for it the way they perform for a review cycle.
Hatch, our AI agent, reads that exhaust and turns it into three things a founder can act on:
- Performance signal. Who is actually moving the work, including the people whose contribution is structural and invisible in a demo, like the engineer who unblocks four others every week.
- Flight risk. Behavioral drift, narrowing of the internal network, response latency, and withdrawal from decisions, surfaced as an alert while you can still do something about it.
- Hiring intelligence. A record of which behavioral profiles have actually succeeded in your company, so the next hire is a bet informed by your own history rather than a resume.
We go deeper on the underlying signals in employee performance signals and on the category itself in continuous performance management.
This edge is bigger at 30 people than at 30,000
Moneyball is a story about a resource-constrained underdog beating a richer opponent by measuring better. That is a founder's story, not a CHRO's.
The asymmetry is real and it is currently unexploited. SHRM's 2026 research found only 39% of organizations have implemented AI in HR at all, with adoption around 60% at companies over 5,000 employees and materially lower below that. Enterprises are busy integrating 40 systems. A 40-person company has one Slack workspace, one calendar, one repo, and a founder who can act on a signal the same afternoon it appears.
The stakes per decision are also higher. SHRM's 2025 benchmarking puts cost per hire at $5,475 for non-executive roles and $35,879 for executives, up 21% in three years. At 40 people, every hire bends the trajectory by more than two percentage points of your company. And the Work Institute's 2026 report, built on more than 120,000 exit interviews, found 75% of 2025 departures were preventable. Preventable means somebody could have seen it. Nobody was looking at the right thing.
How to run the play
- Name your batting average. Write down the metric your company currently uses to decide who is doing well. Then ask what evidence you have that it predicts anything.
- Find the walks. Identify the contributions your current system does not count: unblocking, reviewing, mentoring, absorbing ambiguity. These are usually concentrated in a few people you would be devastated to lose.
- Instrument the work, not the worker. Read patterns in systems people already use. Do not add another survey, and do not install a keystroke logger. The distinction is the whole ballgame, and we set out exactly where our lines are in our trust center.
- Make the leaderboard real-time. A quarterly score changes nothing. A weekly signal changes a conversation.
- Close the loop on hiring. Track which profiles succeed twelve months in, not which ones interviewed well. This is the mechanism that gets you to half the mis-hires.
FAQ
What does Moneyball for companies actually mean?
It means the same thing it meant in baseball: the metric your industry trusts is not the metric that predicts winning, and there is an edge available to whoever notices first. In business, the trusted metrics are the annual review score and the engagement survey. Both are opinion-based, both lag by months, and both are broadly distrusted by the people who run them. Moneyball for companies means scoring people on behavioral evidence produced by the work itself, continuously, rather than on a manager's recollection once a year.
Is this just people analytics with a sports metaphor?
No. Most people analytics runs on HR records: headcount, tenure, comp bands, survey results, review scores. It analyzes what HR wrote down. Behavioral performance intelligence analyzes what actually happened, from the systems where work occurs. The difference matters because HR records are downstream of the same broken judgments you are trying to replace. See our breakdown of what a people analytics platform can and cannot tell you.
Does this work at a 20-person company?
It works better at 20 people than at 2,000, because the data is cleaner and the loop between signal and action is one person long. What you give up is statistical smoothing: with 20 people you are reading individual patterns, not population trends. That is usually what a founder wants anyway. Pricing for lean teams is on our pricing page.
How do you do this without turning into surveillance?
By reading patterns rather than content, by making what is collected visible to the person it is about, and by refusing categories of data that cannot be collected responsibly. Gartner's 2025 research found 87% of employees think an algorithm could give them fairer feedback than their manager does today, so the objection is rarely to measurement itself. It is to measurement done to people instead of with them. Our full position, including what we will not collect, is in the trust center, and larger teams with works council or procurement requirements should start at enterprise.
What is the first thing to change on Monday?
Stop asking managers to rate people and start asking what evidence they are rating from. If the answer is recollection, you have found your batting average. That single question tends to do more damage to a bad performance system than any redesign.
The bottom line
The Oakland A's did not have a data advantage. They had a definition advantage. They knew what a run was made of and everyone else was still paying for hits.
Your competitors are still paying for review scores and survey averages. The work itself is sitting there, generating evidence every hour, unread. That is the arbitrage, and it closes as soon as this becomes normal.
If you want to see what your own company's on-base percentage looks like, book a Hatchproof demo.