What is a pivot strategy in a startup?

A pivot strategy is a deliberate change to the fundamental bet your company is making: what you build, who you sell it to, or both. It is distinct from iteration, which improves a bet you are keeping, and from quitting, which abandons the bet without replacing it. In practice a pivot strategy has three parts. First, an explicit statement of the hypothesis you are abandoning and why the evidence says it is wrong. Second, a new hypothesis that inherits something real from the old one: a customer relationship, a technical capability, a market insight you paid for. Third, a validation plan that tests the new bet before you build it. The Pylon founders' version of the third part is the tightest one we have seen stated: never build before you have very clear validation, which they learned by doing the opposite on their first idea.

What does pivot mean in simple terms?

It means keeping one foot planted and moving the other. The metaphor comes from basketball, and the important half is the foot that stays. A pivot is not starting over. You keep something you have already earned, usually the customers you have talked to, the domain knowledge you built, or the technology you wrote, and you change the direction everything else points in. Pylon kept the post-sales buyers it had spent months interviewing and changed the product from a narrow Slack ticketing integration into a full B2B support platform. A change where nothing carries over is not a pivot, it is a new company, and it should be evaluated as one.

How do you know when to quit your startup?

Separate the two questions first, because founders usually conflate them. The pivot question is whether this specific bet is wrong. The quit question is whether you still want to be doing this at all. Evidence answers the first one: you have tested the hypothesis, customers have told you no in ways you can no longer explain away, and you have exhausted the variations. A Y Combinator partner panel on startup durability names the failure mode precisely, which is that founders rarely reach that point before they act. What usually happens instead is that enthusiasm runs out while the kernel of the idea is still good, and the pivot is a mood dressed as a decision. If that is the honest description of where you are, a pivot will not fix it, because the same feeling follows you into the next idea two months later. The quit question deserves its own answer, made calmly. One of the Pylon founders describes using a written best-case and worst-case document for a different but structurally identical call, whether to leave a salaried job to start a company at all, and the same instrument works here: write both outcomes down before you decide, rather than deciding on how the week went.

Should you pivot into an idea other founders have already tried and failed at?

Yes, if you can name a specific reason it will work for you that did not hold for them. Pylon counted itself as roughly the thirteenth company to attempt B2B support over shared Slack channels, and one of the earlier attempts told them directly not to do it. They did it anyway, and several of those earlier attempts are now their customers. The reasoning is that ideas fail for many reasons that have nothing to do with the idea: the team broke up, the execution was weak, the founders ran out of money, or the market was not ready. Timing was Pylon's answer, and it was concrete rather than hopeful: business conversations had moved into shared channels after the pandemic, and language models had just made that unstructured conversational data usable. A vague belief that you will simply try harder is not a reason. A named change in the world since the last attempt is.

Startup Pivot: How to Know When to Move

A startup pivot is a deliberate change to the fundamental bet: what you build, who you sell it to, or both. It is not iteration, which improves a bet you are keeping, and it is not quitting. The decision is rarely the hard part. The hard part is the signal, which means knowing which evidence earns a pivot and which evidence is just a bad quarter with a story attached.

Pylon is a useful case for exactly that reason. Its founders pivoted twice, for two completely different reasons, and both are legible after the fact. Marty Kausas, Robert Eng and Advith Chelikani spent a long stretch in what they call pivot hell before landing on B2B customer support. Then, about a year after finding the idea, they pivoted again, away from a narrow product that was already making money. They told the story to Y Combinator managing partner Harj Taggar, and the useful part is not the outcome. It is the rules they wrote down along the way.

For scale: Pylon went through YC's Winter 2023 batch, and at the time of the interview the founders describe the company as 79 people, three years old, past a thousand customers, in eight-figure ARR, and up 5.35 times over the previous year. It has raised $51 million across a seed, a Series A, and a $31 million Series B co-led by Andreessen Horowitz and Bain Capital Ventures. Treat the growth and headcount figures as the founders' own account, since they are self-reported in the interview. The reasoning behind the pivots is the part that transfers.

Pivot hell, and the rule that ends it

The first idea, from the two of them who went to college together, was an alumni portal for their school's career center. It failed in the most common way available, which they name without flinching: "we built it before really validating with anybody."

They add a second problem with it, which is that education technology was probably a bad space to start out in. They do not elaborate, and the build-first mistake is the more portable lesson anyway, because it is the one that produced a rule. Out of that first failure they extracted the standing constraint that governed every idea afterward: "we learned to not build something before we like have very clear validation."

That sounds obvious written down. It is not how most early teams behave, because building is the pleasant part and validation is a week of cold outreach with a response rate that will embarrass you. The Pylon founders ran their discovery on The Mom Test, the standard reference for interviewing people about their problems without inviting them to be polite about your solution. Their own summary of it: if you ask your mother whether she would buy your thing, of course she says yes, so you have to ask questions around the thing to find out whether she actually would.

Each successive idea refined the process rather than restarting it. That is the part founders in pivot hell tend to miss. The ideas were disposable. The validation method compounded.

The market you pick caps the company you can build

The move from some idea to this idea came from an exercise that took a couple of days of work across the candidate ideas and is straightforwardly repeatable.

When the three of them joined forces, they arrived carrying different ideas: one in fintech, one in logistics. They argued it out, which they now describe as the valuable part of merging, and then settled it with a spreadsheet and a bottoms-up market size on each. Tab one, the fintech idea: if they crushed it, high tens of millions in revenue. Tab two, the logistics idea: if they crushed it, hundreds of millions, and that was the entire market, not their share of it. Then they added a third tab for a company they were not working on, Rippling, as a model for horizontal B2B software, and sized its market the same way. Everyone who gets paid, times a subscription.

The third tab was orders of magnitude bigger than the first two, and the conclusion landed hard enough that one of them states it flatly in the interview: you are "capped by the market you choose initially."

This is the least sentimental thing in the interview and the most useful. Effort does not lift a ceiling set on day one. If the honest bottoms-up number for your entire category is a few hundred million in revenue, then no amount of execution produces a multibillion-dollar outcome, and you should either accept that or change the category now, while changing it is cheap.

They paired that with a second filter they called the why-now: a large established category plus an emerging trend inside it that you can attach yourself to and grow with. Travel was already enormous before Airbnb; the shift to renting from individuals was the trend. Their own why-now turned out to be that B2B companies had started talking to their customers in shared Slack channels, and none of the three realized at the time that a second why-now, conversational AI, was about to arrive underneath the first one.

The horizontal choice bought them something else, and it is the argument for staying broad early. As they put it, in a large enough space "you don't have to know the end state before you start," because you can keep making small corrections toward whatever the big thing in that space turns out to be. That is the mirror image of the vertical SaaS bet, which trades that optionality for depth in one industry and, in the AI era, often wins for a different set of reasons. Both are defensible. Drifting into one without having priced the other is not.

The signal that says validate, not build

Their discovery loop was unglamorous and worth stealing in full.

Every morning, each founder sent up to 40 personalized LinkedIn connection requests, which was the platform's daily cap, to people in newer post-sales roles: solutions engineers, customer success managers. The pitch was honest about the situation, roughly that they were founders pivoting around, trying to learn about the space, could you spare 15 minutes. The response rate, by their own account, ran somewhere between 1 and 5 percent.

Run the arithmetic they ran and the abysmal number stops mattering. Three founders times 40 messages is 120 a day. Five percent of that is six conversations, which fills next week's calendar. Volume, not conversion, is the lever.

On the calls they asked about the job rather than the product: what do you do, what do you care about, what tools do you use, what do you hate about your job, what does your boss care about, what do you measure. Their structural rule was to spend roughly 80 percent of the call digging into a hypothesis they already had, and to leave the remaining 20 percent open for the other person to say something unprompted that seeds the next set of questions. Most founders run a call entirely as one or entirely as the other, and get either a leading interview or an unusable ramble.

The change that mattered most was not a technique. Two of them had been running this around full-time jobs, and when the last job went, the thing that improved was cycle time. Working nights and weekends, they changed the questions they asked about once a week. Full-time, they changed them daily. Their framing: "changing your questions every day just means like you're actually iterating every day," and "that cycle time affects the speed you get to your idea a lot," more than any of them expected. If you are searching for an idea part-time, that is the specific cost you are paying, and it is worth knowing what it is rather than assuming the only cost is fewer hours.

The signal they were listening for was not enthusiasm. When a friend described losing track of customer conversations spread across shared Slack channels, they went and mom-tested it at three companies, and what came back was a checklist rather than a compliment: the companies were actively looking for a solution, "they tried to build it in house but couldn't do it," and they had evaluated outside vendors and found them wanting. Those three facts together are demand. Someone saying your idea sounds great is not. The gap between the two is what we covered in latent demand.

There is a detail here about speed that is easy to skim past. After the first call surfaced the problem, one of the founders walked upstairs in the hacker house they were living in, asked a founder of Hightouch whether that company had this problem, and was told they were looking for a solution at that moment. Their estimate is that finding prospect number one took about 30 seconds. That is not luck so much as the compounding return on months of conversations: by then they knew exactly which question to ask and exactly who to ask it of.

Someone already tried your idea, and that is not the signal you think

The same Hightouch founder also told them not to do it. They had tried the idea themselves and it had not worked.

Pylon counted roughly a dozen prior attempts at the same thing, describing themselves as "the 13th company to try this exact idea." They built it anyway, and their read now is that most of those attempts died for reasons unrelated to the idea: the team broke up, execution was weak, they gave up. The founders point out that their own YC batch was the first after ChatGPT launched, that AI for customer support was the single most common idea in it, and that almost all of those companies died while the category itself went on to produce real businesses.

The test they state is a good one, and it is a high bar rather than a permission slip: "you just have to have a reason that it'll work for you that it didn't work for other people and that is enough." Their reason was timing, and it was specific. Post-pandemic, business-to-business conversation had genuinely moved into shared channels, across Slack, Microsoft Teams, Discord, WhatsApp and Telegram, and conversational AI had arrived to make that unstructured data usable. Two independent shifts, both nameable, both recent.

A vague conviction that you will simply try harder is not a reason. Neither is the prior failures list on its own. The knowledge that someone already tried this should change which questions you ask, not whether you proceed. Several of those earlier attempts, incidentally, pivoted away and became Pylon customers.

The second pivot is the one founders miss

The first product was not a platform. It was a narrow integration that pulled conversations from shared Slack channels into an existing ticketing system like Zendesk or Intercom. They built it in about a week, and it carried the company to roughly $400K in ARR over the first year.

That is the situation almost no advice covers: the thing works, revenue is real, and it is still the wrong shape. Three signals told them so, and none of them was a feeling.

Customers asked for adjacent features that were not theirs to build. Requests kept arriving for functionality that belonged to the core ticketing system rather than to the integration layer they had scoped.

Prospects tried to skip the prerequisite. Startups told them they did not have Zendesk and just wanted to use Pylon directly, and Pylon had to tell them to go buy the incumbent first. A product whose sales motion requires the customer to purchase a competitor before they can buy you has a structural problem, not a positioning problem.

The incumbent stopped moving. Zendesk went private in a $10.2 billion buyout led by Hellman & Friedman and Permira that closed in November 2022, right as Pylon was starting, and over the following year their customers started describing the product as one that had stopped evolving. An incumbent that is not shipping is an opening with a clock on it.

Their conclusion was blunt: "we definitely aren't going to build a really big company off of just this integration." So they rebuilt as a full B2B support platform, which meant restarting at the bottom of the market with small companies and working back up, because you cannot ship the surface area a large customer needs on day one of a new product.

A Y Combinator partner panel on what makes startups durable describes this category of move directly, and separates it from the panicked kind: "There is another good version of pivoting, which is that through working on something that turns out to be really hard, you are exposed to an even better idea." Pylon did not find the platform idea by brainstorming. They found it by spending a year inside the narrow version and watching where customers kept pushing.

One thing to price honestly if you make this move. Going from an integration to a system of record changes what you are custodian of. Pylon went from passing messages between two systems its customers already trusted to holding the conversational record itself, which means retention policy, access control, audit trails, and a much larger blast radius if something goes wrong. That is a real cost of the pivot, and it belongs in the decision rather than in the post-mortem.

Pivot on evidence, not on how you feel

Pylon's second pivot reads as obvious now because the three signals were all external. The failure mode is the version where they are not.

The same YC durability panel names it exactly. The condition for a real pivot is having genuine evidence that the fundamental thing you were trying to build was wrong, having exhausted the variations. Then comes the honest bit: "Normally that's not what happens. What happens is you run out of enthusiasm." The kernel of the idea is still good, you are a long way from product-market fit, you are bad at sales, you keep getting rejected, and you get sad. The panel's warning about what follows is the most useful sentence in it: "The grass will not be greener over there, particularly if you're pivoting from something that you are deep in into something that you are super shallow in." You trade problems you understand for problems you have not met yet.

Useful pivots tend to keep a foot planted. Framer's move from prototyping tool to website builder kept the browser canvas, the multiplayer and the version-history work, and repointed all of it at a different job. It did not keep the revenue: the rebuilt product restarted near zero, which is the honest shape of this kind of move. The foot that stays planted is capability, not customers. Pylon's second pivot kept the customers, the channel integrations and the buyer relationship, and changed the product's scope. In both cases the thing that carried over was the expensive part.

So before you move, write down two things. What specific evidence says the fundamental bet is wrong. And what you are carrying with you. If the first is thin and the second is empty, you are not pivoting, you are starting a different company while telling yourself a kinder story about it.

Pick co-founders who survive the pivot

One more decision from the interview earns its place here, because it determines whether you can pivot at all.

Before the three of them teamed up, two of the founders nearly co-founded a logistics company with two sales veterans from that industry, on the reasoning that they lacked the domain network. The equity split never came together, and looking back they are relieved: "Especially if you pivot the company and the idea changes off of that industry then those people don't make any sense anymore." The conclusion they reached, independently and from different directions: "you want to pick someone you want to work with and that you will have a good relationship with."

The sharper version of the point cuts at conventional advice. One of them had worked with a co-founder deeply embedded in the community around a Parkinson's app they were building, and notes that this bond made it harder to leave the idea when the evidence said to leave: "I actually think that's one of the reasons that passion is not like the best reason for starting a company too," because "it might blind you to building a big business if that's what you actually want to accomplish."

That is worth sitting with rather than agreeing with too fast. Passion for a space genuinely helps you survive the years. It also raises the cost of the exact decision this article is about. If you are building with someone whose commitment is to the industry rather than to the company, you should know that your pivot conversation will be harder, and you should have it early rather than at the moment you need to move.

Pivoting well is a company capability rather than a moment of courage: cheap validation, honest market math, and the organizational nerve to leave a working product for a bigger one. That is the same operating discipline our pillar on AI for startups is built around.

What to do this week

Write your hypothesis down in one sentence. Name the fundamental bet: this customer has this problem and will pay this much for this solution. You cannot have evidence that a bet is wrong until the bet is written.
Run the three-tab spreadsheet. Your current idea, your best alternative, and a horizontal comparable you admire. Bottoms-up revenue for the whole market of each, not your share. If your ceiling is smaller than your ambition, that is today's problem, not next year's.
Score yourself on the three-part validation test. For your last five customer conversations: were they actively looking, did they try to build it themselves, did they evaluate other vendors? Enthusiasm without those three is not demand.
Fix your cycle time before you add hours. Count how often you changed the questions you ask prospects. If the answer is weekly, that is your real constraint, and it is fixable without quitting anything.
List the adjacent asks you have been declining. Every request for something just outside your scope is a datapoint on where the market thinks you should be. Three from the same direction is a signal.
Name your why-now out loud. What changed in the world in the last 24 months that makes this possible now and did not before? If you cannot answer in one sentence, you are competing on execution alone.
Have the pivot conversation with your co-founders before you need it. Ask what would have to be true for each of you to leave this idea. Find out now whether someone's commitment is to the space rather than to the company.

If the honest answer to most of these is that you do not have the data, the fix is a week of conversations, not a quarter of building. Turning that habit into an operating system your team runs without you is what we teach in AI Operating System for Startups.

Sources

Rebuilding Customer Support for the AI Era (YC Root Access), the interview this article is built on. YC managing partner Harj Taggar talks to Pylon co-founders Marty Kausas, Robert Eng and Advith Chelikani about pivot hell, the market-sizing spreadsheet, their validation method, and the second pivot from integration to platform. All quotations attributed to the founders are from this conversation.
What Actually Makes A Startup Durable (Y Combinator), a partner panel, for the distinction between pivoting on evidence and pivoting because enthusiasm ran out, and for the observation that the grass is rarely greener when you move from depth to shallowness.
Pylon company and funding background: the $31 million Series B, its Andreessen Horowitz and Bain Capital Ventures co-leads, and the $51 million total are stated in Pylon's own announcement. The company's Y Combinator profile names all three co-founders and the Winter 2023 batch, and a16z's separate post covers its earlier Series A investment.
The Zendesk take-private: Zendesk's own announcement of the $10.2 billion Hellman & Friedman and Permira deal completing on 22 November 2022, corroborated by Reuters.
The Mom Test by Rob Fitzpatrick, the customer-interview method the founders name as the basis of their discovery process.