How to run a startup accelerator
My recent obituary for Techstars generated a surprising amount of attention. One of the questions I heard most often from commenters was: “if Techstars is an example of a failed accelerator, what does a good one look like?” In that prior post, I had offered Y Combinator as a positive example, but I didn’t break down exactly why. This post is an effort to unpack what’s required of a startup accelerator to truly serve the needs of high-performing founders.
Step 0: Understand the goal
Before you attempt to design a system to do a specific type of work, it helps to articulate the beliefs and assumptions underlying your design.
Every time I talk to someone about the work we do at Founders’ Co-op, I have to first explain that we don’t really care about technology or business ideas per se. As our name suggests, we believe that founders – small groups of unreasonable humans obsessed with solving a specific problem – are the engine of most positive change in the world, and that’s what we’re solving for as investors.
We have lots of opinions about how those groups can best use the tools available – technology, capital, ideas – to achieve their goals. But if you ever forget that supporting extraordinary founders is the whole game, and that your only job as an early stage investor is to identify them and help them progress, you have irretrievably lost the plot.
Step 1: Find the best founders
An uncomfortable truth about this work is that not all founders are created equal. Being an effective founder of a high-growth company requires a bundle of technical, analytical, leadership and communication skills that don’t often come packaged in a single individual. Certain types of people are uniquely suited to the role, and sifting the world of ambitious humans to discover and help those who have the potential to become great founders is the hardest and most important step.
One failure mode for less effective accelerators, and even more so for the related “studio” or “incubator” model, is to pretend that people who otherwise would not be founders can be coached into the role with sufficient time and scaffolding. This has not been our experience. Teaching entrepreneurial skills is a worthy endeavor and likely offers career benefits to ambitious employees and their employers, but it is not the purpose of a startup accelerator. A good accelerator concentrates and focuses the energy of people who “can’t not” be founders. Anything else is well-intentioned startup theater at best, cynical value extraction at worst.
The essential question at this stage is: “how do you know if a founder is any good?” The internet is awash in takes on that topic, and it’s well beyond the scope of this post, but one reliable way to find the answer is to (a) meet a lot of founders, (b) pick a broad and diverse set you think are the most promising, (c) work with them closely to see how things play out, and (d) repeat, refining your approach through each successive cycle of learning.
Aviel and I know something about this. Between our 15+ years at Founders’ Co-op and another 10+ running the Seattle-area Techstars programs, we’ve screened literally thousands of founding teams and invested in over 250 of the ones we thought were most promising. About half of those then worked in our offices for at least three months, allowing us to interact with the founders and their early team members through every possible flavor of startup crisis: wrong customer, wrong solution, bad hires, bad investors, bad communication, founder conflict, financial mismanagement, latent sociopathy – the ways startups fail are almost too many to count, and we’ve experienced a rich and diverse set of those modes.
We’ve also seen a large enough set go on to varying degrees of success – an IPO, a multibillion dollar acquisition, many eight- and nine-figure outcomes and over a dozen companies still valued at a billion dollars plus – to have an informed view on how that comes to pass. And while the paths to failure are many and various, the modes of success are fewer, and all of them hinge on the unique capabilities of the founding team.
Over time, we’ve developed a strong conviction that certain types of people, and certain combinations of those types, are many times more likely to productively persist in the way that eventually produces startup success. Building a system that selects for these types of humans and invites them to join a community of similarly high-performing peers is the only way we know to set in motion a process that can eventually produce a world class startup accelerator.
Back to the Y Combinator / Techstars comparison, over the past 20 years, each of those systems had the same opportunity to screen tens of thousands of teams, select a subset of them to try to help, and then see how those experiments played out over time. Despite some drift on batch sizes during the peak ZIRP/Covid years, only Y Combinator was able to retain an institutional commitment to attracting and selecting the most promising founding teams in the world as their prime directive. That was and is the foundation of their success, but it’s also just the first step.
Step 2: Get them together in batches
The most effective founders often begin life as outsiders. They are too impatient, too persistent, too curious, too obsessed to fit comfortably into traditional social or professional roles. The early spark for high-performing teams is often based purely on the magic that occurs when a set of humans who have been lonely in their separate obsessions find a set of “similarly different” people and choose to make common cause.
But building a company from scratch is a very specific kind of work. It has many parts, and very few who haven’t done it before even know what all the parts are, much less how they all fit together and in what order. No matter how capable the group of individuals is on a founding team, when faced with the work of creating a new high-growth company, most will freely admit that they have no idea what they’re doing most of the time.
Rather than requiring each team to separately discover or invent the many skills needed to build, grow and finance a new company, accelerators work in batches.
Imagine the joy of finding a small group of people you trust and care for enough to start a company with them. Take that energy and multiply it by a Dunbar’s number of equally skilled and enthusiastic founders, each pursuing a different opportunity, but all at roughly the same stage of development, and applying a broadly overlapping set of technical approaches. Put them all together in a shared workspace with some light programming that encourages random interaction and candid discussion and see what happens.
Accelerators make a big show of their access to experts – the successful founders and big name investors who come to visit and share their experience – and we’ll get to that in a minute, but a surprising amount of of the power in the model comes from the curated peering that a highly selective, cohort-based, in-person program provides. Batches of teams that trust and communicate openly with one another form a learning network that radically accelerates feedback cycles and reduces unnecessary errors for all teams in the batch.
Done right, the intensity of these relationships carries far beyond the duration of the program. Peers in a batch will become their own first-order professional network for the rest of their careers, sharing new learnings; referring customers, employees and investors; providing candid feedback; commiserating when things go wrong, and celebrating wins with a true understanding of what it took to achieve them.
Step 3: Give good advice
Most startup advice is bad. Not that many people have actually built a successful high-growth company, and not everyone who has is able to see clearly how their success came to pass, and what parts of their experience might be useful to others. In other cases, a putative expert may have relevant experience to share, but also has incentives that cause them to shade their advice in ways that serve their own interests, reducing its usefulness or making it actively harmful to the founders being advised.
But while every company’s journey is unique, there is a huge amount of repeatability in the methods and practices required to build and scale a high-growth company. One way that a startup accelerator can bring leverage to its admitted founding teams is to offer access to the best of these ideas.
Sometimes that takes the form of an inspiring speaker who began where the current cohort is now, but has progressed some number of steps beyond and can crumbtrail their experience and lessons learned to help the current group make better decisions. Another, more interactive method is to pair founding teams with individual coaches, also drawn from the set of experienced founders, who can listen more deeply to the current challenges a team is facing and offer specific advice. Still another approach is to gather information and experiences from a broader set of relevant peer companies at different stages and distill that into a set of best practices for specific disciplines like hiring, sales, fundraising or communication.
A good accelerator will use all of these approaches to help the teams in their care avoid obvious mistakes and discover potentially effective ways of solving specific problems without having to learn it all by trial and error. But no accelerator worth participating in will view the advice component of their offering as anything but a hit-and-miss attempt to reduce the inevitable frictions of building a company. The real work is done, and the hardest lessons are learned, by individual founding teams taking actions and seeing what happens, as fast as they can on as many fronts as they can, over and over.
It is worth mentioning that a bad accelerator will not only misattribute the importance of their advice, they will also invite in and tolerate providers of bad advice, including those who don’t know what they’re talking about, as well as those who do, but whose incentives are misaligned with those of the founders they are supposed to be helping. This is a primary failure mode for poorly-run programs.
Step 4: Help them raise money
Almost by definition, high-growth companies consume capital. As with advice, there are many sources of capital available to early-stage founders and many of them come wrapped in harmful packages. An important role for anyone who wants to help founders is to connect them with sources of capital that are most aligned with their goals, and least likely to derail them for reasons other than bad execution.
There is lots of information online about how to raise money for startups and it’s not really the focus of this post, other than to say that startup accelerators must exert the highest standard of care when matching founders with investors.
This process is thankfully much less opaque than it used to be, and the balance of power between founders and funders has improved somewhat, but eliminating bad investors and inappropriate terms are two of the most solvable parts of the startup success puzzle and it is something that good startup accelerators must excel at.
A well-run accelerator is a great equalizer for startups raising capital. By curating batches of high-potential teams, and reducing the defect rate through effective peering and sound advice, good accelerators significantly increase the odds of investors actually making money. And by controlling access to their cohorts and sharing information about investor behavior with founders in their programs, accelerators can weed out bad actors and reward those that most effectively balance their own needs with those of the founding teams.
Step 5: Take the long view
Well-designed and well-maintained systems get better over time; the opposite is also true. Done right, every company in every batch is better off for having participated in a well-run accelerator program. But the real power lies in the accretive benefit of a program that is founded and run by people who continuously make choices necessary for their system to become the best in the world at helping unreasonable people achieve impossible goals.
Every action in a program designed this way reinforces every other:
1. Great founders are attracted to a program that clearly focuses on their needs and delivers results.
2. Thoughtful selection creates exceptional batches that help every team accomplish more and aim higher than they otherwise might.
3. Exceptional founders who are further along in their careers are more willing to share their time and experience with teams so clearly hungry for the advice and able to put it into practice.
4. Excellent investors are more likely to see the quality of companies passing through such an effective process, and to offer terms that reflect the higher expected returns and game theory implications of interacting with the platform and not just the individual companies.
5. Having experienced the power of being selected, working alongside high-performing peers, receiving caring advice from more experienced practitioners and benefiting from a more level investor playing field, program alumni become its strongest advocates and referrers.
6. Return to Step 1
This cycle is continuously compounding – the longer it can be sustained without compromise, the more powerful and effective the platform becomes. But it is also vulnerable and subject to decay at every step. Losing sight of the prime directive, lowering standards to grow more quickly, scaling batch size past the point of human trust and authentic connection, or tolerating bad actors in any part of the system can cause it to trend toward collapse. As with the startups it aims to help, the outcome is entirely in the hands of the founders.
A final note on economics
You may have noticed that this essay did not make any specific assertions about what an offering like this should be worth to the founders who participate, or the terms under which that bargain should be struck.
It could be argued – and some have tried – that the role of the accelerator is so valuable that its owners should be entitled to a founder’s stake in each participating company. I have seen accelerators (and more often their nominal cousins, the “incubators” and “studios”) demand founder stakes of 10, 20, even 50% for their three-to-six month contribution to the actual founders’ decade-plus journey.
I don’t have an answer to this question, but it’s instructive to note that Y Combinator, without question the closest real-world example of an accelerator run by founders, for founders, has settled on a rake of around 7% for their services. Maybe that’s the right number, maybe it’s not, but it’s hard to see how any other system can defend anything more, and it may be that the most farsighted accelerator founders would find a way to ask for less.