The 8-Step Startup Ideation Algorithm

Coming up with a good startup idea is hard.

You’re faced with a complex value function, full of constraints and variables, and you’re looking for that one idea that can optimize the expression.

When my cofounders and I were brainstorming startup ideas, we were always looking for the one solution that would change the world. As a result, we naturally considered a diverse, unrelated set of ideas. We’d pick one that got all of us excited and then tried to test the problem space with interviews and research to see whether it could be the one. If that idea didn’t pan out, it was back to the drawing board to find another.

Eventually, we found a much better process to ideate that led us to Vivid — our in-browser CSS editor that makes frontend development incredibly fast.

The “One Idea” Approach for Startup Ideation

We always thought our startup idea needed to be the “right answer.” There had to be something special about the one idea that would guarantee our success.

Typically, when trying to optimize a function, you use the first derivative to find where the slope of the function is flat, meaning you’ve reached either a minimum or a maximum. In our ideation, we were taking an all-or-nothing approach to optimization — calculating the first derivative and randomly testing ideas to see if plugging any of those in set the first derivative to zero.

In other words, we were throwing darts blindfolded, checking after each throw whether we had hit something. Then, we’d blindfold ourselves again to make the next throw.

For our team, the one idea was Backstage — a platform where fans could discover and invest in early-stage musicians. We were deep into the crypto craze and were convinced that this idea (and only this idea) was going to be a home run.

The Backstage landing page

This approach — I’ll call it “the one idea” approach — poses a major problem for founders like us. If you’re only trying to find the one idea, giving that idea up means starting from scratch.

The more we looked into the Backstage idea, the clearer it became that this was not the right startup for us. We knew next to nothing about the music industry (a notoriously complex and unwelcoming space) and worse yet, we didn’t care about it deeply enough.

The music investing space is highly fragmented, with a lot of competing crypto companies each with strong advantages (including sound.xyz, founded by the son of Sirius XM founder Scott Greenstein and royal.io launched by DJ 3LAU) and we didn’t really have a strong differentiation. Not only that, fans and musicians alike were souring on the idea as the crypto bubble began to implode, making it clear to us that music investing was NOT why people were originally interested in the idea. And yet, we kept going. For months.

As we came to these conclusions, we were also traveling to Oregon for a startup accelerator led by Neo. At this point, our conviction levels were at an all-time low. We knew we had to change our idea, but we didn’t know where we had gone wrong in our ideation the first time around.

Why the “One Idea” Approach Doesn’t Work

Early-stage startups are all about maximizing founder conviction at every step of the process. Most startups don’t die because they run out of money — they die because the founders give up. This has two implications for people trying to find the one idea from the get-go.

First, it makes it really hard to give up on an idea that clearly doesn’t have a future. When you’ve already poured your time, heart, and energy into an idea, going back to square one poses a huge mental block.

You avoid learning more about the space because you’re worried you’ll discover a new competitor. Or worse, you’ll realize you’re not even solving a real problem. Sunk cost fallacy takes over and founders lose their agility when they lock themselves into an idea hoping that something will work out.

Unfortunately for founders stuck in this loop, successful companies need to be able to pivot. Very few winning startups began with the idea they’re famous for. Instead, founders become experts in their space and keep iterating until they find the right niche.

Second, even when founders taking this approach are able to pivot, they typically pivot to a relatively unrelated idea. That means starting from scratch: your learnings from idea 1 can’t carry over to idea 2 if those ideas have nothing in common. A good founder needs domain expertise, and the one idea approach makes that difficult.

We quickly realized that Backstage suffered from both of these problems. We had waited too long to pivot because we knew we didn’t have conviction in the overall space — we just had conviction in our one idea. When we had finally exhausted every possible option trying to make Backstage work, we had to start at square 1 to find something new.

We had diagnosed the problem with our ideation approach, but still didn’t know how to do it better.

The Algorithmic Approach to Startup Ideation

Once you take an introductory machine learning class, you learn a new way to optimize functions. Instead of trying to find local minima or maxima analytically, you do it algorithmically with gradient descent.

Here’s how it works in a nutshell:

• You choose a starting point.
• Move in the direction of the steepest slope.
• Recalculate slope at this new point.
• Repeat the previous two steps until the change in slope from step to step becomes minimal. Then, you know you’ve reached an optimal point.

Gradient descent starts with a relatively arbitrary point, but keeps moving in the right direction

The best part of gradient descent is that it works step by step — each step brings you a little closer to the optimal point you’re looking for.

So what does the gradient descent algorithm look like for startup ideation? We decided to give it a shot.

We started by listing out spaces we were 100 percent convinced would produce a billion dollar company in the next 5 years. Then, we trimmed our list. Were we excited enough about the space to work in it for the next 5–10 years? Did we feel like we had some unique insight into the space that the market hadn’t yet realized?

This got us down to one industry we were convinced we wanted to work in. After that, our singular focus became narrowing that space as much as possible through research and user interviews — until we got to the smallest possible space that we were convinced could still produce a billion dollar company.

At this point, more research was barely moving the needle on our conviction levels, and that’s when we knew we were ready to commit to an idea.

After picking an initial idea, we kept talking to potential users, pivoting repeatedly in the “direction of greatest need”. Each change in the idea was smaller in scope than the last as we honed in on the true solution we could build into a successful startup.

We still don’t know for sure if this idea will achieve product-market fit. But we have a lot of conviction that we can successfully pivot to an adjacent idea in the space that builds on what we’ve learned in the process. There’s still a chance we need to shift gears — after all, it might turn out that the local maximum we’ve found simply isn’t valuable enough to create a full-fledged startup — but we know we’re moving in the right direction.

By taking the algorithmic approach, we can pivot with conviction and learn more with every step.

A Guide for Future Entrepreneurs

This algorithm has been a game-changer for us, so I thought I’d share the step-by-step guide we used in case it helps any other entrepreneurs out there.

Here’s how you can replicate the algorithm:

1. Find spaces that you’re super passionate about — enough to work in them for the next 5 years. Filter them with additional criteria that are important to you such as unique insight into industry, competitor differentiation, or large market size.
2. Eliminate spaces where you don’t have at least 90% conviction that the space could spawn a unicorn.
3. Interview potential users in these spaces. For each space, use these interviews to narrow scope. The goal is to come up with the narrowest space where your unicorn conviction is still high.
4. After progress on narrowing begins to stall, choose the space where you have the most conviction.
5. Pick an idea in that space based on the biggest problem you recognized in your user interviews.
6. Talk to more users to find out whether the idea you‘re currently investigating is addressing a real user need.
7. Take stock. If you gave users unbiased, open-ended questions that pass The Mom Test, they probably told you about problems in adjacent spaces. Find the “direction of greatest need” and pivot your idea to address that direction.
8. Go back to step 6 until you find your maxima.

This algorithmic approach to ideation kept us nimble, encouraged us to pivot, and gave us far more conviction in the direction we were moving. My cofounders (the incredible Jorge Zreik and Alberto Rigail) and I are very excited about where it’s led us and we can’t wait to share it with the world! Stay tuned for an update within the month on where this approach led us. If you hate CSS styling, you’re going to love it.

The elusive “one idea”

Huge thanks to Jesse Zhang — one of the amazing mentors our team met at Neo Accelerator — for inspiring us with the algorithmic approach. Major shoutout to Aryan Bhasin for going line-by-line through this article to help me make edits. Have to also thank Cheyenne, Sid, and my dad for other ideas on the article!