You're Almost Never Too Late to a Market

A crowded market hands you problems you can solve; a premature one asks you to wait for the world to change.

You found the idea. You spent a week falling in love with it. Then you searched, and there they were: a dozen competitors, three of them funded, one of them good.

If that search made you close the tab and quietly bury the idea, this is for you.

Most advice on market timing for startups treats too early and too late as the same mistake pointing in opposite directions. Pick the wrong moment, lose either way. That framing hides the most useful fact about timing: the two mistakes are different in kind, and only one of them should scare you.

Here is the short version. Visible competition almost always proves that a market exists and pays. True lateness is rare and structural: it takes locked distribution, network effects, entrenching regulation, or economics that punish every entrant. A premature market has the opposite problem. The demand, behavior, or infrastructure you need has not arrived yet.

Crowded Is Not Complete

Competitors cluster. They chase the same customers, copy the same pricing, repeat the same positioning, and fight over the same channels. Watching each other is the cheapest market research there is, so they converge.

Convergence leaves edges. A neglected segment. A different business model. A distribution channel nobody bothered to learn. A technology shift the incumbents are too invested to adopt.

In 1998, search looked finished. AltaVista, Lycos, Excite, and Yahoo were household names, and every one of them was racing to become a portal, stacking email, news, chat, and shopping on top of the search box.¹ The whole market had clustered around one belief: search is a commodity feature. Google entered late and took the one position the crowd had abandoned, search itself.

Zoom did the same in 2011. Video conferencing had WebEx, Skype, GoToMeeting, and Google, and Skype had just sold to Microsoft for $8.5 billion.

Eric Yuan had spent years inside WebEx, embarrassed by the customer feedback, asking Cisco for permission to rebuild the product and getting refused.² He called the industry extremely crowded, then entered it anyway. From the outside, crowded and complete look identical. From inside the customer calls, the market was wide open.

I've made the longer case for this in Go Ahead, Build What Already Exists: many competitors prove demand, and entering takes a genuine opinion about how things should work rather than a slightly better feature list. Take that as read. The more interesting question is what happens on the other side of the timeline.

When Late Is Actually Late

First, honesty about the exception. Some markets really are closed, and it helps to know what closed looks like.

A market is closed when structure locks you out, and the locks are specific. Network effects, where the product's value is the other users already on it. Distribution that is contractually sewn up: shelf space, default placements, enterprise deals with years left to run. Switching costs so high that even a clearly better product cannot pay them off. Regulation written around the incumbents. Or unit economics where everyone loses money, including the leader.

Notice what is missing from that list: the number of competitors. Crowding shows up in one search. Closure takes real checking, and checking is not mysterious. Ask a would-be customer what leaving their current tool would actually cost them. Read the incumbent's contract terms and default placements. Look up who the regulation was written around. An afternoon of that answers more than a month of staring at the competitor grid. Count the locks. Ignore the logos.

Even the locks get picked, at a price. Facebook launched into network effects that supposedly made Friendster and MySpace untouchable, and rebuilt density from zero, one campus at a time.³ Uber drove straight into an industry locked by licensing regulation. Postgres seeped past entrenched commercial databases by deleting the license fee.

Each aimed a deliberate move at its specific lock, and a better product alone would have bounced off every one of them. So read a lock as the price of the door. Welded shut is rarer than it looks.

Early Means Waiting for the World

Now flip it. You found no competitors. It feels like a gift, the whole map unclaimed.

Except an empty market is ambiguous evidence. Either nobody thought of it, or plenty of people thought of it and the world declined. Missing customer demand. Behavior people have never practiced. Enabling infrastructure that does not exist. Willingness to pay that stays at zero no matter how much heads nod.

One caution before you diagnose: new and early are different claims. A blue ocean, in the strategy-book sense, has no competitors because you redrew who the buyer is and what they are buying. The demand is already there, walking around in other markets. Cirque du Soleil had no rival circuses to fight because it sold to adults who were already buying theater tickets.⁴

An early market has no competitors because the demand itself has not shown up. One is a positioning achievement. The other is a waiting room, and it loves to flatter itself as a blue ocean.

Webvan is the cleanest case I know. Grocery delivery in 1996: the right idea, hundreds of millions behind it, and a $1 billion contract with Bechtel to build the warehouses the world did not have.⁵ Every missing piece of the market, Webvan paid to construct itself. It lost roughly $20 on every order and was dead 20 months after its IPO.

Instacart launched the same idea in 2012 and it worked, because by then the world had built the missing pieces on its own: smartphones in every pocket, gig labor, grocery stores that could serve as free warehouses.⁶ The idea didn't change fundamentally. The world did.

Sixteen years late looks like a fluke until you see the distribution. One study followed roughly 500 brands across 50 product categories and found that market pioneers failed 47 percent of the time and still led only 11 percent of their categories. The companies that ended up owning those categories entered, on average, thirteen years after the pioneer, and took three times the pioneer's market share.⁷ Sixteen years is not a miracle of patience. It is roughly the standard biography of a market leader.

Marc Andreessen, who lived a version of this with Loudcloud, put it bluntly: "being too early is the same as being wrong."⁸

You can hear the tell in your own conversations. If every pitch begins with convincing the buyer that the problem exists, there is no demand to capture yet. You are funding a category education campaign, and paying tuition for whoever arrives after you. People calling your idea clever while nobody feels urgency and nobody reaches for a wallet: that is the sound of a market that is not ready.

A Crowd Can Be Wrong Together

One complication before this settles into a rule you can lean on. It cuts against the claim this essay opened with, so it gets its own section.

Webvan had company. Kozmo was couriering DVDs and snacks across Manhattan in under an hour. Urbanfetch was racing them block by block. Streamline installed refrigerated boxes in suburban garages. HomeGrocer ran its own trucks down the West Coast.⁹

By any scan you'd run today, delivery in 1999 was a crowded market. Every one of those companies was gone by the end of 2001.

So the opening rule has a blind spot. A market can be crowded and premature at the same time, because crowding measures how many founders showed up, and market readiness measures whether the world did. That crowd was proving that investors believed in 1999 delivery. Nobody was proving that customers would pay for it.

A crowd validates a market only when the crowd lives on revenue. A crowd living on runway is one premature bet placed a dozen times, and each new entrant adds evidence about the fashion, none about the market.

You can usually spot the runway crowd by its era, because it gathers wherever capital chases a thesis. Consumer VR in 2016 was one: Facebook's $2 billion Oculus deal set off the stampede, and funded headset makers and content studios multiplied around an audience that never arrived.¹⁰

Magic Leap alone raised about $2.6 billion before shipping anything, then reportedly sold around 6,000 headsets in six months. You can nominate your own corner of AI for the current round.

So the earlier diagnostic gets one more line. Count the locks. Ignore the logos. Then ask what the logos live on. Run that last question over any AI category today and watch the crowd thin. The logos collecting more from customers than they burn serving them are a fraction of the market map.

The Difference Is What You Can Change

Here is the asymmetry underneath all of this. Go back over the two lists in this essay.

Everything on the crowded-market list is a decision: which segment, which model, which price, which channel. Any of them could be your move this quarter. Everything on the premature-market list is someone else's decision: customers changing habits, infrastructure getting built, a technology getting cheap. You can nudge those at the margin. You cannot make them.

Competition is an obstacle. Absent readiness is a condition. You work around obstacles. Conditions, you mostly wait out.

So run the Whose-Move Test:

1. Write down everything standing between you and your first hundred paying customers.
2. Mark each item my move (something you could act on this quarter) or world's move (someone else has to change first).
3. Read the list. If the blockers are mostly your moves, you are late in the workable sense, and a crowd that pays its own bills is just proof of demand. If the critical blockers are the world's moves, you are early, whatever the calendar says.

Run it on the essay's own case before you run it on yours. Grocery delivery, 1996: pick a launch city, partner with existing stores, recruit shoppers. Every one of those moves was as available to Webvan as it was to Instacart sixteen years later, and executing them perfectly would have changed nothing. The blockers that mattered sat on the world's side: customers who had never bought anything online, no device in their pockets to order from, no labor you could summon by app. By 2012 the world had crossed off its side of the list, and everything left was a founder's move. That is what the test reads: not how long your list is, but whose side the critical items sit on.

One warning before you trust your own answers. When you are in love with an idea, world's moves disguise themselves as your moves.

"Customers don't know they have this problem yet" files itself under content marketing. Your move. "The infrastructure doesn't exist" files itself under we'll build it ourselves. Your move again, says the part of you that already named the product.

Webvan shows the disguise at full price. That Bechtel contract was a world's move reclassified as a company to-do: the warehouses the world had not built, moved onto Webvan's own list for a billion dollars. On the pitch deck it read as ambition. On the timeline it was waiting for the world, at the world's prices.

So mark honestly. If a blocker is your move only because you plan to spend money doing the world's job, it is still the world's move. Buying it just burns money while you wait.

Early does not force you to abandon the idea. Narrow to the sliver of the market that is ready today, reshape the model to live on the demand that already exists, or park the idea with a tripwire: the specific outside change that calls you back.

Make the tripwire one concrete line, not a vibe. A founder burying grocery delivery in 2001 could have written: come back when everyone carries an internet-connected computer that knows where it is. That line trips around 2010. Instacart showed up two years later.

Even Webvan's founder heard the wire trip. By 2014, Louis Borders was building grocery delivery again.¹¹ His second act never went far, but read the signal in the return: the man who lost the most on the premature version never decided the idea was wrong. He decided the date was.

All three are conscious responses to a diagnosed condition. They are nothing like the reflex that started this essay, where a page of competitors reads as a closed door.

Here's the shorter way to hold all of this. A crowded market, the kind living on revenue, is a market where the world has already moved. The demand arrived, the habits formed, the infrastructure got built, and the crowd is the receipt.

An empty market makes no such promise. And the crowded one you almost buried your idea over? Somebody else paid its tuition years ago.

The competitors were never the door. Stop asking whether you are too late. Ask whose move is missing.

Rabbit Hole

More on this subject

• Critical Demand Signals Nobody Talks About: Use costly workarounds to verify that a crowded market contains urgent unmet demand.
• The Idea Is the Product Now: See why proven readiness still leaves idea quality as the higher-leverage decision.

See this concept in another subject

• Not All Uncertainty Deserves Respect: Apply the same early-small-or-commitment uncertainty test to important decisions beyond market timing.