You promised a date, the date is here, and the thing is not right yet. This article answers the question every founder eventually faces: is it smarter to hold the launch, or to ship what you have and fix it later?
The Night Before, and the Answer
I have shipped something before it was ready. Not once. More than once. And I still remember the specific feeling, which is not panic and is not excitement. It is a low, sick calm. The kind where you already know the answer and you are busy building a case against it.
The version I remember best went like this. We had a date. The date had been announced. People were waiting. And two days out, I knew the thing was rough in places that mattered. Not cosmetically rough. Rough in the part people would touch first. I told myself we would patch it in week two. I told myself the feedback would make it better. I told myself that momentum was worth more than polish.
Here is the honest answer to the question in the title, and I will give it to you before I earn it: if the weak part of your product is the exact part your customer buys it for, delay. If the weak part is somewhere else, ship. That is the whole test. It is not about perfectionism and it is not about courage. It is about whether the first thing a person experiences is the thing you promised them.
We shipped. And the patch in week two did not fix what the first impression broke. The people who bounced did not come back to see the improvement. They had already filed us under a category in their head, and getting out of that category cost far more time and money than the delay would have.
That is the part nobody warns you about. Shipping late costs you a date. Shipping broken costs you a reputation. Those are not the same currency, and one of them is much harder to earn back.
Key Takeaways
- The right test is not whether the product is perfect, it is whether the weakest part is the part customers actually buy it for.
- A missed date is a one time cost, while a bad first impression compounds against you for months or years.
- Research on market pioneers shows being first is not the protection most founders assume it is.
- Google delayed its flagship AI model in July 2026 and absorbed a roughly four percent stock hit rather than ship something that fell short on coding and reasoning.
- Saying “not yet” in a market where everyone is sprinting is a competitive move, not a cowardly one.
Why “Ship It” Feels Like the Only Option
Let me be fair to the pressure, because it is real and I do not want to pretend I am above it.
You announced a date because dates create accountability. Without a date, projects drift. Anyone who has run a team knows that the calendar is sometimes the only thing standing between you and eighteen months of polishing. So you set the date on purpose, for good reasons.
Then there is the audience. You told people. Maybe you built a waitlist. Maybe partners rearranged their plans around yours. Moving the date feels like breaking your word, and if you are someone who cares about your word, that lands hard.
Then there is the money. Revenue is modeled against that quarter. Payroll is real. A delay is not an abstract concept when you are the one who signs the checks.
And underneath all of it is the fear that somebody else gets there first. Every founder I know carries some version of this. The market is moving fast, especially in AI, and the story in your head is that the window closes if you blink.
So the pressure is honest. I am not going to tell you it is imaginary.
But here is the reframe that changed how I make this call.
The date is not the promise. The date is the packaging of the promise. The actual promise is that when someone shows up and uses the thing, it does what you said it would do. When you keep the date and break that, you did not keep your word. You kept your calendar and broke your word. Those are very different things, and your customer only experiences one of them.
Most of us confuse speed to market with speed to trust. They are not the same. Speed to market is how fast you get the thing in front of people. Speed to trust is how fast people decide you are reliable. You can win the first and lose the second in the same afternoon.
And in a market where everyone is racing, being trusted is a scarcer position than being fast. Everybody is fast now. AI made fast cheap. What did not get cheap is the thing that makes someone say, “if they released it, it works.”
What the Evidence Actually Shows
I do not want you to take this on vibes. Here is what the record says.
Being first is not the shield you think it is. Peter Golder and Gerard Tellis published a study in the Journal of Marketing Research in 1993 that looked at roughly 500 brands across 50 product categories, including the ones that did not survive, which earlier research had quietly left out. They found the failure rate for market pioneers was about 47 percent, compared to roughly 8 percent for early followers. Surviving pioneers held a mean market share near 10 percent, while early leaders held around 28 percent. Those early leaders entered the market an average of 13 years after the pioneers. Thirteen years. The race you think you are losing may not be the race that decides the outcome.
One bad experience is often all you get. PwC’s customer experience research found that 32 percent of customers say they would stop doing business with a brand they love after a single bad experience. Not a brand they tolerate. A brand they love. And that number climbs sharply after two or three bad interactions. Your launch is, for most of your future customers, their first and only audition.
Trust breaks faster than it rebuilds. This is well documented in the organizational trust literature. After a violation, trust does not simply return to where it was when the problem is fixed. It frequently falls below its original baseline, and the effort required to restore it is substantially greater than the effort that built it in the first place. Negative information carries more weight in how we judge people and companies than positive information does. That asymmetry is why the week two patch does not undo the launch day impression.
And the receipts are everywhere. In December 2020, CD Projekt shipped Cyberpunk 2077 on a date it had already delayed several times, in a state that did not hold up on older consoles. Sony pulled the game from the PlayStation Store and offered full refunds, which was close to unprecedented. The stock fell around 16 percent on that news and was down roughly 43 percent from its pre release high. Samsung’s Galaxy Note 7 recall cost the company about 5.3 billion dollars and crushed a quarter’s operating profit. And Google itself learned this in February 2023, when a factual error in the very first public Bard demo helped wipe roughly 100 billion dollars off Alphabet’s market value in a single day.
Which makes what happened this month interesting. In July 2026, Bloomberg reported that Google had delayed Gemini 3.5 Pro, its flagship model, because the rebuilt version fell short of internal expectations on coding and complex reasoning. Alphabet shares dropped about four percent, erasing roughly 200 billion dollars in market value. Google published no model card, no pricing, no benchmarks. It just held. That is a company that has already paid the other price once, choosing to pay this one instead.
Meanwhile, over the same stretch, Anthropic did almost nothing flashy. Quiet enterprise work. Then it confidentially filed a draft S-1 with the SEC, with reported annualized revenue around 47 billion dollars and a valuation discussed above a trillion dollars. Note the difference in posture. One company had a launch and chose not to have it. The other had no launch spectacle at all and got rewarded anyway.
The market is not paying for speed right now. It is paying for reliability.
How I Decide Now
I would love to tell you I have a scoring rubric. I do not. What I have is a set of questions I ask myself out loud, usually to somebody else in the room, because saying it out loud is how I catch myself rationalizing.
The first question is the one I gave you at the top. What is the weak part, and is it the part they are buying? If the onboarding is ugly but the core engine is solid, ship it. Ugly is survivable. If the core engine is unreliable and the onboarding is beautiful, you have built a very attractive front door onto a house with no floor. Do not open it.
The second question is, who is going to see this first? Not who eventually. Who first. Early users are almost never a random sample. They are your most enthusiastic people, and they are also the ones with the loudest voices and the longest memories. If the people most likely to advocate for you are the people most likely to hit the broken part, that is not a launch, that is a controlled demolition of your best relationships.
The third question is, what does the delay actually cost, in specifics? I make myself write it down. Not “we lose momentum.” Actual items. Two weeks of revenue. A partner conversation I have to reopen. An email I have to send that I will not enjoy sending. Nine times out of ten, when I write the delay cost down in plain language, it is smaller than the story I was telling myself. Fear inflates. Paper deflates.
The fourth question is the one I learned the hard way. If I ship this and it goes badly, what is the recovery plan, and how long does it take? If I cannot describe the recovery in concrete terms, I have not made a decision. I have made a wish.
And then there is a fifth thing, which is less a question and more a gut check. My wife will tell you I am not a naturally patient man. I like motion. But I have learned that the desire to ship on the date is very often about me and not about the customer. It is about not wanting to make the phone call. It is about how it will look. When I notice that the loudest argument for shipping is about my own discomfort, that is usually the moment I know the answer.
Holding a launch is not a failure of nerve. It is a bill you choose to pay in the currency that is cheaper for you. A date is cheap. Trust is not.
One more thing I have come to believe. When you delay well, and you tell people why in plain language, you often gain something. People remember that you told them the truth when it cost you. That is a deposit, not a withdrawal.
Practical Steps for Making the Call
Name the weakest part out loud, in one sentence. No hedging, no engineering vocabulary that hides the problem. If you cannot say plainly what is broken, you do not understand it well enough to decide.
Map that weakness against the buying reason. Write down the single thing the customer is paying you for. If the weakness sits on that line, that is a hold. If it sits somewhere else, that is a ship. This one test resolves most of these decisions in about five minutes.
Write the cost of delay as a list of items, not a feeling. Dollars, conversations, dates, names. Then do the same for the cost of a bad first impression, honestly, including how long you think recovery would take. Compare the two lists side by side.
Ask who touches it first and what they touch. Walk the actual path a new user takes in the first ten minutes. Not the demo path. The real one, including the parts you usually skip because you know the workaround.
Consider narrowing instead of delaying. Very often the right answer is not later, it is smaller. Ship the part that works to the people it works for, and hold the rest. A narrower launch that is fully true beats a broad launch that is partly true.
Decide the recovery plan before you decide to ship. Who fixes it, by when, and what you say to customers in the meantime. If that plan feels vague, you have your answer.
If you hold, tell people early and tell them why. Do not go quiet. Say what happened, say what you are fixing, say when you will be back. Vagueness reads as trouble. Specificity reads as competence.
Frequently Asked Questions
Is it ever right to ship something you know is imperfect?
Yes, and most of the time. Every product ships imperfect. The distinction is between imperfection at the edges and imperfection at the core. If the flaw is in a feature few people touch first, ship and improve. If the flaw is in the reason they came, hold.
How do I tell customers I am delaying without losing credibility?
Tell them early, in plain language, with a specific reason and a specific new date. Do not over apologize and do not disappear. Most people respond well to a founder who says the honest version. The credibility damage comes from silence and vagueness, not from the delay itself.
Does not shipping fast beat shipping perfect in a fast market?
Speed matters, but speed to trust matters more than speed to market. Research on market pioneers found roughly a 47 percent failure rate, while early followers failed far less often. Being first is not protection. Being the one people rely on is.
What if a competitor launches during my delay?
Then they launch, and you find out whether their version holds up. If theirs is solid, you would not have beaten it by shipping something weaker. If theirs is rushed, their stumble becomes your opening. Either way, launching broken to preempt someone rarely improves your position.
Is this just perfectionism with a better excuse?
It can be, which is why the test is narrow and specific. It is not “is everything great.” It is “is the core promise reliable.” If you keep finding new reasons to hold and none of them touch the core promise, that is perfectionism, and you should ship today.
The Call You Will Have to Make Alone
I still think about that launch. Not constantly, but it shows up. Usually two days before something else is due, when the same low, sick calm rolls in and I start building the same case.
Here is what I want you to hear, because I wish someone had said it to me plainly. Nobody is going to remember your date. I promise you that. Nobody has ever built a relationship with a company because the release landed on a Tuesday in March. But they absolutely remember the first time your thing did not work when they needed it to. They remember it years later. They mention it to other people. It becomes the sentence that follows your name around.
Shipping broken is not bold. It is borrowing against a reputation you have not finished building, at an interest rate nobody quotes you up front. And you pay it back in a currency you cannot print more of.
Google just took a nine or ten figure hit on paper to avoid shipping a model that was not ready. Whatever else you think about that company, notice that a business with every incentive in the world to hit its date decided the date was the cheaper thing to lose. If they can hold, you can hold.
And when you do hold, be honest about why. Not corporate honest. Actually honest. The people who stick with you through a delay you explained well are the same people who will still be there in three years, and those relationships are worth more than any launch week you will ever have.
I write about this stuff because I keep getting it wrong and slowly getting it less wrong, and because I think more of us should say that part out loud. If that is your kind of thing, come find me. I am easy to reach and I like the conversation more than the applause. Tell me what you are wrestling with, or tell me about the launch you should have held. I have got one of those too.
The date is a promise about when. Everything else is a promise about what. Only one of those is worth your name.
About the author
Jonathan Mast is the founder of White Beard Strategies. He is an entrepreneur and speaker who writes and talks about AI, business, and the honest parts of building. He is also a husband and a father, which has taught him more about patience than any launch calendar ever did. He writes regularly about what is working, what is not, and what he got wrong.
Sources
- Alphabet shares fall on report its most powerful AI model Gemini 3.5 Pro is delayed (CNBC)
- Google Gemini Launch Delayed as Tech Falls Short of Internal Goals (Bloomberg)
- Anthropic Files Confidential S-1: Joins $3 Trillion AI IPO Race (Yahoo Finance)
- Anthropic tops OpenAI as most valuable AI startup, nears $1 trillion valuation (CNBC)
- Pioneer Advantage: Marketing Logic or Marketing Legend? Golder and Tellis, Journal of Marketing Research (SAGE)
- First-Mover Disadvantage, Tellis and Golder (Harvard Business Review)
- Experience is everything: Get it right (PwC Customer Experience Study)
- Employee trust repair: A systematic review of 20 years of empirical research (ScienceDirect)
- Verbal or Written? The Impact of Apology on the Repair of Trust (Frontiers in Psychology, PMC)
- Sony pulls Cyberpunk 2077 from PlayStation store after backlash; developer’s shares tumble (CNBC)
- Samsung says Note 7 recalls will cost at least $5.3B (CBC News)
- Google’s AI chatbot, Bard, sparks a $100 billion loss in Alphabet shares (NPR)
