Your best growth opportunity is the one you can't prove yet

The growth opportunity that pays best is usually the one you cannot yet justify. It is visible in your own data, it has no competitor precedent, and that second fact is exactly why it loses every internal argument to a safer proposal. We call the revenue those opportunities create Return on Bravery™—and the reason most companies never collect it has nothing to do with courage.

I defined the term in a Tech:NYC feature earlier this year. What follows is the part that did not fit in a Q&A: why promising bets stall before they are tested, how to size one so it stops needing to feel brave, and what the experiment must teach whether it works or not.

Why promising growth ideas lose before they are tested

Because the approval process is calibrated to certainty, and certainty requires precedent.

This is the mechanism, and it is worth being precise about, because it is usually mistaken for a personality problem. Teams are not timid. They are responding rationally to a system that asks, “How do we know this will work?”—a question that can only be answered by pointing at someone for whom it already worked.

A proposal that copies a competitor comes with proof attached. A proposal drawn from an anomaly in your own data comes with a hypothesis. In a room that rewards proof, the second one loses every time, and it loses on process rather than on merit.

As I put it in the Tech:NYC piece:

Attribution tools tend to prioritize what is easiest to measure rather than what is most important to understand. As a result, brands repeatedly optimize the same audiences, channels, and creative patterns while overlooking entirely new sources of demand.

I saw this firsthand as CMO of Greater Than. Breastfeeding mothers buying a sports hydration drink were already visible in the company’s customer comments, but they did not resemble the market the product had been built to serve. What looked like a roughly $300,000-a-year anomaly became a $10 million business once the team treated it as an opportunity rather than noise. I wrote separately about how that overlooked customer segment surfaced.

The opportunity was valuable because the data supported it and the category had not priced it in yet. Those are the same conditions that made it difficult to defend before the company had evidence.

So the fix is not to argue harder for the brave bet. It is to make the bet small enough that it no longer requires certainty to proceed.

How do you test a growth idea without betting the business?

Write down the cost of being wrong—as a specific number and a specific duration—before you write down the upside. Then keep shrinking the test until it fits inside a pre-approved learning budget and can run without disrupting the operating plan.

Consider an illustrative example. Suppose an unexplained customer segment keeps appearing in your reviews, and you want to know whether the demand is real. There are two ways to fund that question.

The defensible version is research, positioning work, and a quarter of build: perhaps $60,000 and twelve weeks before any customer behavior comes back. It survives scrutiny because it looks serious. It also means the company gets to ask the question once that year.

The bounded version is a landing page and a small paid test using adapted creative: perhaps $2,000 and ten days. Those figures are illustrative, not a benchmark. The test is not a substitute for the research or strategy that would follow. It is designed to answer the first question gating all that downstream investment: whether the customer behavior is strong enough to justify the next test.

	A bet you have to defend	A bet you can afford to lose
Cost of being wrong	A quarter and a headcount	Ten days and low four figures
What clears it	Proof that it will work	Proof that it can be measured
Who approves it	A committee	A pre-authorized experiment owner
How many you get	One a year	Several a quarter
What a failure teaches	Whose fault it was	Whether the demand was real

The practical test is simple: if a failure cannot be reported in a sentence, the bet is too big or too vague to run repeatedly. A small experiment does not prove the market. It earns—or fails to earn—the next experiment.

Being wrong four times at that size can still leave a company ahead of the team that spent a quarter becoming confident about one option. One brave bet is a story. A cadence of cheap ones is a growth system.

What does every bounded growth experiment need?

A bounded experiment is not merely a smaller campaign. It is a decision designed backward from what the company needs to learn.

The point is not to tolerate more risk. It is to buy more useful evidence with less of it. That requires the hypothesis, limits, and decision rules to exist before the results arrive.

Start with a signal, not just an idea

An interesting idea can come from anywhere. A testable opportunity needs a reason to exist: an unexplained customer behavior, an unexpected use case, a creative pattern, a gap in the purchase journey, or a group finding the product without being invited.

The signal does not have to prove the opportunity. It has to make the opportunity more than a guess.

Choose one gating customer behavior

Do not ask the first experiment to validate the audience, message, channel, unit economics, and total addressable market at once. Identify the earliest observable behavior that would justify learning more.

That might be a qualified visit, a signup, a sample request, an add-to-cart, or a purchase. The right behavior depends on the business. What matters is that the result changes the next decision.

Fix the downside before modeling the upside

Set the spending limit and timebox before discussing how large the opportunity could become. Otherwise the size of the imagined prize quietly expands the test until it is no longer an experiment at all.

The limit should be small enough that a negative result leaves the operating plan intact, but large enough to observe something the company is prepared to believe.

Write the success, stop, and next-test rules in advance

Decide what result earns a larger test, what result ends the hypothesis, and what ambiguous result would justify refinement. If those rules are written after the data arrives, every outcome can be made to look encouraging and nothing has actually been tested.

A useful brief fits in one paragraph:

We believe [audience] will [behavior] because [observed signal]. We will spend no more than [amount] over [duration] to test it. If [threshold] occurs, we will [next test or investment]. If it does not, we will [stop or refine].

That paragraph is the boundary. Everything outside it may matter later, but it does not get to inflate the first decision.

What does Return on Bravery look like in practice?

Best Day Brewing had a high-NPS product, growing retail distribution, and no mature demand engine keeping pace with either. The obvious move was to compete harder for people already shopping for non-alcoholic beer. Every other brand in the category could see—and bid for—the same customer.

The more valuable opportunity lived outside that funnel. MOST helped connect the product to a larger human tension the category had not claimed, then turn that tension into a measurable creative experiment.

The “Sober Sex” campaign launched in 21 days with an initial creative investment of $15,000. It was provocative, but it was not provocation without a hypothesis: the campaign tested whether non-alcoholic beer could become relevant to millions of people through an occasion and motivation bigger than the category itself.

The result:

• More than $3 million in trackable revenue
• An estimated 10 million additional potential customers
• A Webby nomination
• Evidence that the brand could create demand outside the existing non-alcoholic-beer funnel

The full Best Day Brewing case study carries the campaign, measurement, and commercial results in detail.

The first experiment did not prove every future audience, message, or channel for the brand. It earned the right to keep building against a larger market—and it did so without requiring the company to bet the entire plan before the first evidence arrived.

What should a failed growth experiment teach you?

A failed experiment should eliminate an assumption, sharpen an audience, or change the next test. If the only conclusion is “it didn’t work,” the experiment was too vague.

Useful failure is specific. It distinguishes between a customer who does not care, a message that did not make the value clear, a channel that could not reach the group efficiently, and a test that never produced enough signal to answer the question. Those are different outcomes and they lead to different decisions.

The report should be as bounded as the test:

We believed X because we observed Y. We spent Z over N days. The target behavior did—or did not—cross the threshold. We will now stop, refine, or scale.

If the result cannot be written that way, the team has probably measured activity instead of a decision.

This is why a portfolio of small bets compounds while one heroic bet does not. The failed tests leave evidence behind. The heroic failure leaves a crater and a debate about whose fault it was.

What changes in a team that can test ideas cheaply?

Confidence, mostly, and it compounds in an unglamorous way: the arguments get shorter. When being wrong has a known limit, nobody needs to win the meeting before the company can learn. The debate moves from whose opinion is better to what the test said.

Teams that reach that point stop relitigating the same playbook and start accumulating evidence nobody else in the category has. Each test makes the next hypothesis sharper, the next experiment easier to size, and the next decision faster to make.

When the only visible output is efficiency against known demand, leadership manages marketing like a cost center—regardless of the team’s actual capability. The remedy is not a more persuasive deck. It is a repeatable way to show new demand being created.

Finding those opportunities is the first half of the problem, and it is the half a Growth Scan is built for: turning unexplained demand in your own data into a ranked set of customer-acquisition opportunities. Sizing them so being wrong is cheap is the half that decides whether any of them ever get run.