The Explore/Exploit Framework for Startups

How to balance action and uncertainty in the early stages of building a company.

The hardest decisions in life — picking a partner, a job, or a city to live in — all start the same way: with uncertainty and too many options.

While some people stumble into great outcomes by luck, most need a thoughtful process: testing possibilities, gathering feedback, and then committing.

What is explore/exploit?

Explore/exploit is a computer science strategy for balancing two competing objectives:

• Exploration: you want to test new options to gather information.

• Exploitation: you want to commit to the best known option and maximize your expected value.

If you’re building a business, you’ll face an endless stream of explore/exploit scenarios:

• Which market should we target initially?

• Which candidate should we hire as our first engineer?

• Which marketing channel should we focus on?

Because each of these questions is very open-ended, finding the best solution is hard. What’s worse, too little exploration can lead to failure (RIP Quibi), as can too much (e.g. stealth mode startups that never ship). So how do you make a great decision quickly?

How to apply explore/exploit

The explore/exploit approach is to spend some time methodically exploring your options, and then at some point cut off the exploration phase and commit to the best option that you have.

Let’s look at how one might use explore/exploit in three common startup scenarios: choosing a market, hiring, and marketing.

• “Which market should we target initially?” → Make a list of 10 promising markets, cold email until you’ve spoken to at least three people in each market, then focus on the market whose response was the strongest.

• “Which candidate should we hire as our first engineer?” → Solicit resumes until you’ve seen 100 candidates, interview the 10 most promising candidates onsite, extend an offer to the best interviewee.

• “What marketing channel should we focus on?” → Make a list of 10 promising marketing channels, test each of them with a $500 budget, then focus on the channel that performed the best.

Explore/exploit is a powerful framework for navigating decisions with uncertainty, but deciding how long to spend exploring feels more like art than science. Should you try 10 marketing channels, or 3, or 50? Should you spend $500 on each marketing test, or $50, or $50,000? These are important questions, and finding the right balance is especially important for startups, where runway is limited and companies that explore for too long without shipping anything die.

But how do you decide how much time to spend exploring? If you test too few options, you risk making a suboptimal choice. If you test too many, you waste time and lose momentum.

How much should you explore?

If you don’t explore enough, you might be going all in on a bad plan. Perhaps you loved the first engineering candidate and thought they were an A+ so you made an offer, but if you had talked to a few more engineers it would’ve been clear that the candidate was actually a B+ at best.

If you explore too much, you waste time instead of moving forward. If you’ve interviewed 300 engineering candidates and haven’t made an offer, you’re doing something wrong and squandering your time and everyone else’s.

Deciding how much time to invest in exploration depends on several attributes:

• ⏳ Is there a time constraint? If you absolutely need a new engineer within 60 days, then you need to be ready to make offers within a month.

• 🏔️ How high are the stakes for the decision? If you’re betting your company on it, err on the side of more exploration. If it’s relatively minor, go all in sooner and move on to something more important.

• ↕️ Is it more important to capture upside or avoid downside? If you’re hiring a VP of sales, optimizing for the best option matters. If you’re choosing a blog template, good enough is fine.

• 🔄 Is the decision easily reversible? Hard-to-reverse decisions, like hiring or picking a company direction, require more upfront exploration. Easily reversible decisions like buying a monitor do not.

• 🎲 How much variety is there between options? If you’re picking an engineer from a group of 0.5x and 3x and 10x engineers, you’ll need to spend more time exploring than if you were picking from a set of 9x and 10x engineers.

• 🤔 How do you feel about the options so far? If you’ve seen a lot of options and have at least one you’re happy with, move forward. If you don’t like any options so far and you’d regret not exploring more, then explore more.

How to explore efficiently

Set clear boundaries for exploration

1. Establish a clear goal. Without it, exploration is aimless. Do you care more about the cost per user for an acquisition channel, or how many users that channel can reach? If it’s cost per user, is that per user who visits your website, or registers, or makes a purchase? The best marketing channel for maximizing registered users might be very different from the one that lets you acquire purchasing users as cheaply as possible.

2. Estimate the ROI of exploring additional options. Suppose you're evaluating options and assign them scores based on their performance. If the last five scores were 10, 15, 3, 29, and 19 (on some scale), by all means, keep exploring. Maybe the next option will be a 43. But if the last 5 options were 27, 28, 27, 26, and 29, then the variance between options is low, and you should just commit to an option and move on.

Optimize how you explore

1. Balance effort vs. insight for each option. When interviewing job candidates, you don't conduct hour-long interviews with all 100 applicants. Instead, you do 15-minute phone screens with the top 30, then hour-long phone interviews with the top 10, then you bring the top four candidates onsite, and finally you extend an offer to the best one. This incremental approach works in other areas, too. For example, you can spend $200 on each of 10 promising marketing channels, then $2000 on each of the top three channels, and then go all in on the best channel. Frequently, you’ll get enough negative data on an option with the first few minutes/dollars invested, and you can move on quickly.

2. Evaluate options in parallel when possible. For example, if you want to test ten different marketing channels with $500 each, do the tests at the same time instead of one at a time.

3. Prioritize the order of exploration. If you have a lot of options, start with the most promising ones first. For example, if you are considering 20 marketing channels and test four at random, their CACs (customer acquisition costs) might be $20, $30, $7, and $11. This is a wide range, and it suggests that more testing is needed. But if you rank the channels by their perceived potential and test the most promising four, perhaps their CACs will be $2, $4, $9, and $7. Now you might feel comfortable stopping the exploration phase and picking the channel with the $2 CAC.

Explore broadly and learn from others

1. Sample from different types of options. If you’re interviewing engineers, don’t restrict yourself to your classmates from Harvard class of ‘08. Instead, cast a wide net and consider past colleagues, cold inbound candidates, cold outbound candidates, candidates from sites like interviewing.io, bootcamp graduates, and so on. Similarly, if you’re trying to pick a marketing channel, test the channels you’re the most comfortable with, but also channels used by your competitors, channels ignored by your competitors, and channels that your audience is likely to see. The more starting points you explore, the less likely you are to climb the wrong hill and get stuck in a local maximum. The outcome of the exploration phase is only as good as the quantity and diversity of the inputs.

2. Learn from others. If you’ve never hired a salesperson before, you can spend 100 hours interviewing candidates and slowly calibrating your understanding of sales ability. Or you can ask a few of the best salespeople you know (and perhaps the best salespeople that they know) to share which attributes they consider to be important.

“Only a fool learns from his own mistakes. The wise man learns from the mistakes of others.”

- Otto von Bismarck

Common traps

The most common mistakes are not exploring enough or exploring too much.

• If you’re avoiding exploration, ask yourself if you’re picking the best option or just the easiest one.

• If you’re stuck in endless testing, consider whether new options are meaningfully better, or if you’re just afraid to commit.

Not exploring enough usually stems from avoiding discomfort. People like easy choices, and so picking the option that seems good enough or the one that you’re most comfortable with is tempting. If your decision is important, ask yourself if you’re choosing the best option or if you’re taking the easy way out.

Exploring too much usually stems from indecisiveness. If every new option you try is 10% better than the last, great, keep exploring. But if each incremental option provides very little value then perhaps you’re anxious about committing to a single choice. Ask yourself if you’re allowing the perfect to be the enemy of the good.

Putting it all together

As a founder, you’re paid to make decisions under uncertainty. Mastering the explore/exploit tradeoff isn’t just a clever mental model — it’s a survival skill. Know what you're optimizing for, follow the recipe below to explore intelligently, and then commit with confidence.

Here’s a concise guide on navigating this balance:

1. Decide on the metrics you’re optimizing for.

2. Allocate some amount of time to explore your options. If a decision is more important, less reversible, or involves wildly different options, commit to spending more time exploring.

3. Prioritize options and explore — in parallel if possible — until time runs out or you’re happy with the best option.

4. Double down on the best option.