Writings · Customer Decisions
Moneyball for Startups
Your funnel shows where growth is lost. The customer decision behind each rate shows what to change. Testing the change shows whether you were right.
You're buying the wrong thing
In Moneyball, Billy Beane has lost his three best players and has a third of the Yankees' payroll. His scouts spend the meeting arguing over who has a good face and whose girlfriend is ugly. Then a young economist named Peter Brand, played by Jonah Hill, tells him the problem in one line: baseball thinks it's buying players, and it should be buying wins.
To buy wins, you need runs. To get runs, you need people on base. The sport was paying for batting average, speed and swagger. It wasn't paying for the thing that actually produced runs, so that thing was cheap.
Your goal shouldn't be to buy players. Your goal should be to buy wins.
Most startup GTM models have a version of the same problem. Founders track pipeline, demos booked, conversion, activation and retention, and those are the right rates to watch. What's missing is the explanation. Every one of those rates is the result of a customer decision, and there are four of them.
The batting averages of SaaS
Batting average wasn't useless. It just measured something close to what mattered instead of the thing itself, and the gap was where the money was lost.
Startups have their own batting averages:
- Top-of-funnel volume is the stat founders reach for first. It's also the one most easily faked by a good campaign aimed at the wrong people.
- Pipeline coverage tells you how many opportunities exist. It doesn't tell you whether any buyer has decided to change anything.
- Demos booked measure curiosity, which is different from commitment.
- Closed-won is a contract. A logo that never activates is a churn you haven't seen yet.
Each of these numbers is real. The trouble is that each one counts activity without saying which customer decision produced it, or whether that decision can change. More dashboard resolution won't answer that. So the model adds up, the board deck looks fine, and revenue still misses.
On-base percentage: the four customer decisions
Brand's insight was that on-base percentage was the stat that actually turned into runs. In B2B, revenue that sticks comes from customers making four decisions in order. Each decision already has a metric your team tracks, and an explanation your team is usually guessing at.
-
01 · Act
Act or remain with the status quo
Qualified buyers do nothing.
Qualified buyers decide whether the problem is worth solving now. Your no-decision rate counts how often they don't. It can't tell you why urgency failed to become commitment.
Explains your no-decision rate -
02 · Buy
Buy from us or from an alternative
Wins are difficult to reproduce.
Buyers who have decided to act choose you, a competitor or another approach. Your win rate counts who converted. It can't tell you which situation made those buyers ready to switch.
Explains your win rate -
03 · Adopt
Adopt or retain the old behavior
Activation does not become real use.
The contract says the team will use the product, but the team decides whether real work moves into it. Your activation and early retention rates count who finished onboarding and who kept using it. It can't tell you whether meaningful work changed hands.
Explains your activation and retention rates -
04 · Renew
Renew or leave
Customers reduce use or leave.
The customer decides whether the value was real and worth paying for again. Your renewal rate records who paid again. It can't tell you where the decision to leave began.
Explains your renewal rate
Each decision has a different decision-maker, a different fear and a different definition of value. A champion who says yes at Act may have no influence at Renew. That's why one funnel stat can't stand in for all four.
Work backward from the wins
In the film, Brand works backward. The A's need about 99 wins to make the playoffs. That means scoring roughly 814 runs and allowing no more than about 645. Once you know the target, every roster move is a question of how many runs it adds.
Startups can do the same, but only once the four decisions are the units of the model:
It's about getting things down to one number.
The math only holds if each rate is measured against the people who made it through the decision before it:
- Decision rate
- Qualified opportunities where the buyer decided to act. It's the flip side of your no-decision rate: if 60% of qualified buyers stall, your decision rate is 40%.
- Win rate
- Wins divided by buyers who decided to act. If your CRM counts stalled deals as losses, take them out first, or the model counts that loss twice.
- Adoption rate
- New customers who moved real work into the product.
- Renewal rate
- Adopters who renewed at their first renewal, here at 12 months. The model assumes customers who never adopted don't renew.
The result is the expected ARR from one cohort that survives its first renewal, at a fixed ARR per customer, with no expansion or contraction. It's an estimate, not a guarantee.
Take a fictional company that needs 50 retained new customers from one cohort at $40K ARR each, or $2M in new ARR that sticks. At its current rates, it needs 654 qualified opportunities, and here's how they thin out:
Act loses the most people. Buy has the lowest rate, which is why five points there buys the most when the other rates hold steady.
Here's what five points at each decision would do, one at a time:
| Change | Act decision rate |
Buy win rate |
Adopt adoption |
Renew renewal rate |
Opportunities needed | Saved vs. today |
|---|---|---|---|---|---|---|
| Today | 40% | 30% | 75% | 85% | 654 | n/a |
| +5 pts at Buy+17% relative lift | 40% | 35% | 75% | 85% | 561 | 93 |
| +5 pts at Act+13% relative lift | 45% | 30% | 75% | 85% | 581 | 73 |
| +5 pts at Adopt+7% relative lift | 40% | 30% | 80% | 85% | 613 | 41 |
| +5 pts at Renew+6% relative lift | 40% | 30% | 75% | 90% | 618 | 36 |
Rows are sorted by impact, not by decision order. The relative lift shows how big the same five points is for each starting rate.
Moving Buy from 30% to 35% means 93 fewer qualified opportunities to source, qualify and work for the same retained revenue.
The founder's instinct is to get more pipeline. But the four rates multiply, so with the other rates held constant, a five-point lift at the lowest rate, here Buy at 30%, produces the largest increase in retained customers. That's arithmetic, and it has a catch: five points isn't equally hard to earn everywhere. Moving 30% to 35% is a 17% improvement. Moving 85% to 90% removes a third of everyone you lose at renewal. Improve any rate by the same percentage and all four decisions pay back the same. Renew also matters more than this table shows, because the model follows one new cohort and churn across your existing base needs a model of its own.
The arithmetic tells you where to look, not what's wrong. Why is Buy weak? Is the company losing to a competitor, to an internal build, or to a cheaper way of getting the job done? You can't move the win rate until you know which. The formula is easy. The explanation behind each rate is the hard part.
- 01Understand the decision
- 02Identify a change
- 03Test it
- 04Measure the rate
Illustrative examples. The research tells you which one is true for your customers.
The table shows the biggest mathematical opportunity. The real question is which rate you can move enough, at an acceptable cost, to matter.
Recreate it in the aggregate
After the A's lose Jason Giambi, the scouts want to find another Giambi. Brand says they can't, and they don't need to. What they need is to “recreate him in the aggregate,” rebuilding his on-base percentage from players everyone else has undervalued.
Founders make the same mistake when growth stalls. They look for one big fix: a star sales hire, a new channel, a rebrand. The better question is which of the four decisions is leaking, and what's cheap to fix in it. Sometimes it's a sharper cost-of-waiting story at Act. Sometimes it's an onboarding step that turns a purchase into a habit at Adopt.
You can't answer that from the dashboard. The dashboard shows where the rate dropped, not why. The why comes from the people who made each decision: the buyer who decided to act, the one who picked a competitor, the user who never logged in twice, the customer who renewed without a second thought.
Adapt or die.
Beane's scouts had decades of experience and still got the math wrong, because they couldn't see what actually produced runs. Before you build your next model, pick one number your team can measure but can't explain, and find the customer decision behind it.
You aren't buying leads. You're buying four decisions, made by real people, in order.