George Bonaci

Direct mail was the bet nobody wanted. Junk mail to people's homes, and the standing verdict on it was that it absolutely would not work; after a few rounds of iteration it became one of the biggest acquisition channels the company had.¹ Bonaci credits that bet as the cleanest instance of his method, and the part he credits is not the channel. It is the disbelief.

Bonaci runs growth at Ramp, the corporate card and spend-management company Eric Glyman and Karim Atiyeh founded, and he led growth at Samsara before that. He is the archive's primary voice on growth as a function rather than as something a founder does on the side, and the tension worth holding him to is that he runs two incompatible epistemologies at once. On channels he is a purist about first principles: no imported playbooks, blank slate, hypothesis, measurement designed before the spend. On people he is the opposite, and imports almost everything, a book program from a previous employer, an onboarding schedule written to the minute, a management framework from a business novel published in 1984. He does not treat that as a contradiction. Whether it is one is the question this file leaves open.

Growth is science

He states the claim without softening it: "growth is just science, and most marketers are bad at science."¹ The honest answer to how a given business grows is that nobody knows, because every business is different, so the posture he prescribes is the scientist's rather than the practitioner's. Come in with a blank slate. Form a hypothesis. Design the measurement before running the thing. Then run enough experiments that one of them works.

He is a chemist by training, and he attributes part of the difference to profile rather than to effort: the chemist or engineer scopes an experiment in a way the writer or communications mind does not, because scoping is the actual skill and it happens upstream of any spend. The failure mode he names is the operator asking what they already know and how to apply it here, instead of what their hypothesis is and how it will be measured. The reader is directed to the file on growth as experimentation, which carries the floor he puts under the velocity argument: throughput beats polish only until the tests get sloppy, at which point volume teaches nothing at all.

Where the alpha is

The word he borrows for the edge is "alpha," taken from investing, and the definition is deliberately negative: whatever is not saturated and other people are not doing. He names three places it hides, and they are ranked. The unknown is a channel too new for anyone to have crowded it. The disbelieved is the direct-mail case, where the shared conviction that something will not work is itself the moat, because a tactic that looked obviously good would already be gone. The cross-niche is the richest, borrowing from another vertical or another geography, because peers in your own space already know the local playbook and the only knowledge worth having is real but untransferred.¹ The file on seeking alpha in growth maps those three onto three modes of learning, and rates learning from peers the weakest for exactly the reason the framework predicts.

None of that works as a single bet, which is why he manages growth the way an investor manages capital rather than the way a marketer manages a plan. Big swings and reliable small wins are allocated across time horizons on purpose, with the long-horizon slice ring-fenced and everyone told in advance that results will not arrive soon; the reader is directed to the file on the growth portfolio of bets for why concentration in a winning channel is a sign the method is working and only becomes a failure if it lasts. He changed his own mind on brand inside this frame, crediting the enterprise-software company Gong, which funded work he describes as completely unmeasurable and then watched it surface in the size of their inbound channel. He does not pretend that is proof. It is an association, arriving with the same measurement problem he started from.

People, not channels

The surprising thing about a growth leader this quantitative is that his strongest views are about hiring and management. For the first growth hire at an early company he says he would always skew more junior, because hiring for potential matters more than buying experience, and what he wants is a smart generalist who can think in first principles rather than a specialist carrying a playbook from a company an order of magnitude larger.¹ The reader is directed to the file on hiring junior generalists for potential for the anti-profile and for his own concession that the bet raises the manager's burden considerably.

His management rule is the one that gets quoted back at him: a good leader needs to know how to do everyone on their team's job, but poorly, and the poorly is load-bearing. Enough fluency to step in and to ask the right question, not enough to compete, because a leader who does the job better than the person hired to do it has described a hiring failure rather than a strength.¹ The file on why a leader should know every job poorly works through the case the rule cannot reach, the deep-craft function a leader could never do even badly.

The imported structure

Here is where the blank slate stops. Bonaci's argument is that companies universally say they value learning and development and rarely fund it, because doing so takes time, resources, intentionality, and top-down insistence, and management is therefore a skill that either gets deliberately invested in or never develops. His worked example is the Leadership Principles program from Samsara: a box of roughly fifteen business books shipped to each leader's home, one a month, each tied to a peer discussion of a specific principle and a requirement to demonstrate that principle in practice.¹ The books are not the point, he insists. The structure and the accountability around them are, because without the demonstration step the program decays into reading theater.

The same intentionality shows in onboarding, which he describes as having been written out for him in excruciating detail, the first two weeks scheduled to the minute. The reader is directed to the file on management as an invested skill for the side effect he considers the real payoff, which is that identical onboarding makes new hires comparable and turns assessment into a matter of fact rather than impression.

His defense against the obvious objection, that decades-old management books cannot speak to the present, is a distinction between tactics and fundamentals: channels and tools change, business fundamentals do not change much, so the discipline is vetting the book for a transferable principle rather than reading it for instructions. His illustration is Eliyahu Goldratt's The Goal and the theory of constraints, which he re-reads and applies directly to his own job, naming the velocity of experimentation as the binding bottleneck. The file on the theory of constraints carries the diagnosis and its uncomfortable corollary, which is that constraints move, so the answer is only ever today's answer.

Where things stand

The structural claim he cares most about is that growth should be as independent as possible and report to a founder, because its mandate is broader than either marketing or product and it needs latitude to work across the whole organization. He names this as a concrete reason he values Ramp, where the growth team reports to one of the co-founders, and he pairs the structure with a cultural rule that keeps it from becoming a turf fight: growth's job is not to make anyone happy, it is to make the business successful.¹ The file on growth team independence records both halves and the friction between them.

The unresolved thing about him is a disagreement with the company he works for. His prescription of junior first-principles generalists sits in open tension with the spiky-specialist hiring model Ramp's founders describe, and neither side states a reconciliation. The usual one drawn from outside is stage and role type, a blank-slate growth problem rewarding a generalist and a deep-craft function rewarding a spike, but that is an inference rather than a position either party has taken.

No plate is entered against his name, and the reason is worth stating rather than leaving to a blank field. One systematizing instinct aimed repeatedly at whatever structural inefficiency is nearest is precisely his method, and the archive files six of his ideas under exactly that mechanism. What has never fit is the shape of the career it would have to describe: every plate in this catalog was built for a founder whose habit compounds into an organization, and Bonaci runs a function inside a company somebody else founded. His own emphasis on knowing every job poorly and hiring people who will beat him at it points somewhere else again, and the archive does not yet have the plate that would hold it.

Key facts

• Runs growth at Ramp, where the growth team reports to one of the co-founders, and led growth at Samsara before that.
• Trained as a chemist, and attributes his experiment-scoping instinct to that rather than to marketing craft.
• States the thesis as "growth is just science, and most marketers are bad at science," and prescribes a blank slate, a hypothesis, and a measurement designed before the spend.
• Names three sources of edge, the unknown, the disbelieved, and the cross-niche, and rates learning from peers the weakest form of it.
• Credits direct mail, a bet dismissed as junk mail that would obviously not work, as becoming one of the biggest channels after iteration.
• Manages growth as a portfolio allocated across time horizons rather than as a plan, on the assumption that most bets fail.
• Would always skew junior for a first growth hire, preferring a first-principles generalist to a senior specialist carrying a big-company playbook.
• Holds that a leader should know every job on the team but poorly, and treats knowing a job better than the hire as evidence of a hiring failure.
• Points to Samsara's Leadership Principles program, roughly fifteen books shipped to each leader with a peer discussion and a demonstration requirement, as the model for funding management as a skill.
• Re-reads Eliyahu Goldratt's The Goal and names the velocity of experimentation as the binding constraint on his own function.