.png)
Moneyball for companies means replacing reputation and instinct with measured output and cost when you decide who to hire, fund, keep or retire. It applies the logic Oakland's front office used in baseball: find the contributions that actually produce results, price them honestly, and stop paying for things that only look like value.
For a founder, CFO or COO, the translation is direct. Every team produces something and costs something. The discipline is to measure both from evidence, compare them across teams of people and AI agents, and move headcount and budget toward the work that returns the most.
What are the key principles of Moneyball for business?
The key principles are to measure what produces results, ignore traits that only look like talent, price every contribution, and make decisions from evidence rather than reputation. Four ideas carry over from baseball almost unchanged.
- Measure outcomes, not proxies. Baseball scouts valued batting average and appearance. The front office valued on-base percentage because it correlated with scoring. In a company, hours logged, meetings attended and tool adoption are the batting averages.
- Price contribution against cost. A player's value is wins above what a replacement would deliver, divided by salary. A team's value is output relative to the full cost of producing it.
- Distrust the narrative. Confident managers and polished self-assessments are the scouting reports of the corporate world. They are often wrong in the same direction.
- Find undervalued capacity. Oakland won by buying what the market underpriced. Inside a company, the equivalent is capacity that already exists and is not being used well.
The common thread is that the unit of analysis is the work, not the opinion of the work.
How much of Moneyball is real?
The core of Moneyball is real: the Oakland Athletics did use on-base percentage and other undervalued statistics to compete against far richer teams. The film and the book compress timelines and dramatize personalities, and critics have noted that the team's pitching and a few star players mattered as much as the analytics.
That distinction is useful for executives. The lesson is not that data beats experience. It is that when an industry prices something with the wrong metric, there is an arbitrage for whoever measures the right one. Most companies are still pricing talent and capacity with metrics that were convenient to collect: headcount, titles, tenure, survey scores and manager ratings.
Why most companies measure the wrong things
Most companies measure activity, sentiment and spend separately, so they cannot connect any of them to what the workforce produced. The result is a set of numbers that are easy to report and hard to act on.
Consider engagement. Gallup reported that U.S. employee engagement fell to 31% in 2024, its lowest level in a decade. Globally, Gallup found engagement dropped from 23% to 21%, with associated lost productivity estimated at $438 billion for the world economy. These are serious figures, and engagement does correlate with results: in a Gallup meta-analysis of 183,806 work units across 347 organizations, top-quartile engagement units posted higher sales productivity than bottom-quartile units.
But a survey score is still a sentiment reading, collected periodically and filtered through what people choose to say. It tells you about conditions. It does not tell you what a given team shipped last quarter or what that output cost. A CFO asked to defend the workforce budget cannot present an engagement percentage as a return.
The same gap shows up in annual reviews. Ratings built on manager recall compress the year into a few memorable moments and are exposed to the same biases that Oakland's scouts carried.
The metrics that matter: output, cost and capacity
Moneyball for companies rests on three measurements: what a team produced, what it cost, and how much capacity it truly has left. Each needs a definition that finance and operations can both accept.
| Measure | What it answers | Common mistake |
|---|---|---|
| Output per person | What did each contributor and team produce? | Counting hours or messages instead of completed work |
| Fully loaded cost | What did that output cost, including agent spend? | Leaving AI and tooling costs outside the team's number |
| Real capacity | Where is there room before a new req is justified? | Assuming a full calendar means a full workload |
| Agent return | What did the AI layer deliver for its spend? | Reporting adoption rates as if they were returns |
Gallup's explanation of productivity frames it as output relative to input, with output per hour and revenue per person as common measures. That framing is the right starting point. The work is in making "output" concrete for functions that do not ship units.
Define output by function
A support team's output might be resolved conversations at a target quality level. An engineering team's might be shipped changes that survive review. A sales team's might be qualified pipeline advanced. The point is not a universal metric. It is a defensible one per function, agreed in advance, that reflects work completed rather than effort displayed.
Count the whole cost
Fully loaded cost includes salary, benefits and tooling, plus the model usage and agent platform fees that now sit in many budgets under separate lines. When agent spend lives in a different cost center than the team it supports, every team looks cheaper than it is.
Moneyball for companies in the age of AI agents
In a workforce that includes AI agents, the Moneyball question becomes what the agent layer returned, not how widely it was adopted. Adoption is the new batting average: it is visible, easy to report and loosely related to results.
A dashboard showing that 80% of staff touched an AI tool says nothing about whether cycle times fell, whether error rates held, or whether the spend paid for itself. The cleaner test mirrors how you would evaluate any hire. Define the work the agent is responsible for, measure completed output against a baseline, and compare it to the full cost, including usage fees and the human review time it still requires. Hatchproof's piece on measuring AI agent performance walks through how to set that baseline.
There is a second effect that rarely appears in budgets. Agents change what people do, not only how much. A team that adopts agents may produce the same output with fewer hours on routine tasks, or more output with the same hours. Without a before-and-after read of actual work, you cannot tell which, and you cannot decide whether to hold headcount, redeploy it or stop adding spend.
Where capacity really sits before you open a req
Before approving a new role, find out whether the capacity you need already exists inside the team. A requisition is the most expensive answer to a capacity problem, and often the wrong one.
Capacity is hard to see from a calendar. Meeting load, context switching, handoff delays and rework consume hours that never appear on a headcount plan. Two teams of the same size can have very different real capacity depending on how decisions get made and how much work waits on a single person.
A practical review before any hire looks like this:
- Measure current output per person for the team, using completed work rather than activity.
- Identify where work queues: which people, approvals or handoffs hold things up.
- Check agent coverage: which tasks an existing agent could absorb and at what cost.
- Compare marginal return: estimate what an added person or added agent spend would produce against what it would cost.
- Decide, then re-measure after the change so the next decision starts from evidence.
This is the same logic Oakland applied to roster spots. You do not buy a player to fill a position by habit. You buy the contribution the team is actually missing.
How to read performance from work already happening
The most reliable measurement comes from the work itself, not from a survey layered on top of it. Communication, execution, decisions and collaboration all leave traces in the tools a company already runs.
That is the premise behind workforce performance signals: patterns in how work moves, who unblocks others, how decisions get made and how output flows. Reading those signals avoids the two weaknesses of the traditional approach. There is no administrative burden of collecting self-reports, and there is less distortion from manager recall or from people performing for a review cycle.
Hatch is built on this approach. It reads performance from communication, execution, decisions and collaboration across tools such as Slack, email, 1:1s and calendars, and it shows what the workforce produced and what that cost, agent spend included. It surfaces patterns to leaders and does not rate, rank publicly or make the decision. Individual biometric data stays visible only to the individual, never to their manager or the organization. The platform's security practices and data-handling commitments are described on its site. For leaders who want to see how this works in practice, the Performance Management product gives a live view of individual contribution, team velocity and project ROI built from real work data. You can also review how Hatchproof connects with your existing tools on the integrations page.
If your organization will not connect its work tools to any platform, this approach is not available to you, and a survey-based process may be the only option on the table. The tradeoff is that you will be managing from sentiment rather than output.
Common mistakes when applying Moneyball to a company
The most frequent failure is borrowing the vocabulary of analytics while keeping the habits of opinion. A few patterns recur.
- Choosing a single metric and optimizing it. Any one number gets gamed. Pair output with quality and cost so that gaming one shows up in another.
- Measuring individuals without context. Output per person varies with role, team design and the quality of the work handed in. Compare like with like.
- Treating the data as the decision. Metrics surface patterns. A leader still has to weigh context, and the evidence should inform judgment rather than replace it.
- Ignoring the trend line. For perspective on what healthy looks like at the macro level, the BLS productivity data is a useful benchmark. U.S. nonfarm business labor productivity rose 2.2% in 2025, with output up 2.6% on 0.4% more hours worked. A team whose output per hour is flat against that backdrop deserves a closer look.
- Keeping agent cost invisible. If AI spend is not attached to the work it supports, the return is unknowable.
Fairness also matters. Measurement that people do not trust will change behavior in ways that damage the numbers. Be explicit about what is measured, who sees it and what it will and will not be used for. If you are working to reduce bias in performance decisions, evidence drawn from the work is a stronger foundation than recollection.
A 30-day plan to start
You can begin moneyballing a company in a month by defining output for two or three functions, connecting the data that already exists, and reviewing the first read with finance and operations together. Keep the scope narrow so that the first decision is a real one.
- Week 1: Pick the teams where headcount or agent spend is under debate. Agree on what output means for each and what the fully loaded cost includes.
- Week 2: Connect the sources that hold the work, and establish a baseline for output per person and cost per unit of output.
- Week 3: Identify where capacity is queued or underused, and where agents are returning value versus simply being used.
- Week 4: Make one reallocation decision from the evidence, whether that is holding a req, redeploying capacity or cutting spend that is not returning, and set the date to re-measure.
Hatch's Launch onboarding is designed to get a team set up quickly, with a named engineer helping connect your stack and configure the signal.
FAQ
What is a famous line from Moneyball?
The film's central argument is that value can be reduced to a measurable contribution, and that teams should focus on a small set of numbers that reflect it. In business, the useful version is a small set of numbers covering output, cost and capacity rather than a single score.
Where can I stream the movie Moneyball?
Streaming availability for Moneyball changes often by region and service, so check a current listing such as the rental and subscription options in your country.
Is Moneyball only for large companies?
No. Smaller companies often benefit most because each hire is a larger share of cost and a wrong call is harder to absorb. The method scales down well since it requires a clear definition of output and an honest view of cost, not a large analytics team.
Start by picking one team where a headcount or agent spend decision is pending and write down what it produced last quarter and what that cost in full. If you cannot fill in both numbers from evidence, that gap is the first thing to close. Hatchproof reads the work already happening across your tools and shows what your people and agents produced and what it cost, so you can see output per person, where capacity really sits and what the agent layer returned before you commit to the next req. You can explore the Performance Management product, and when you are ready to see your own numbers, Book a Demo with Hatchproof and walk through your first read with the team.
