Fourteen years is long enough for an industry to decide what worked and what did not. Since ICANN opened the last round of new gTLD applications in 2012, more than a thousand new extensions have made their way onto the internet. Some reached millions of registrations, some became recognizable names in their own right, and others never quite established why people should care.

Naturally, we built scorecards. Registrations. Renewals. Revenue. Market share. Those numbers became shorthand for deciding which TLDs had succeeded.

Those numbers matter. I have spent more than 40 years in sales and more than a decade in the domain industry, so I am certainly not going to argue that commercial performance is irrelevant. A registry needs a sustainable business behind it. But I do think we learned too much from what was easiest to measure. We became very good at counting how many domains an extension sold, while spending less time asking whether the namespace itself developed a convincing reason to exist.

Availability can put an extension in front of people. Pricing can encourage registrations. Marketing can generate awareness. A registry can build distribution and make the product easy to access. What none of those things can guarantee is relevance.

And relevance is much harder to put on a spreadsheet.

Some TLDs found a natural place in the way people communicate, build and identify themselves online. Others generated considerable registration volume without developing the same cultural or practical meaning. That does not make one commercially successful and the other a failure. It means we probably need a more nuanced definition of what success looks like.

Operating .link through Nova Registry has made that distinction particularly clear to me. Running a TLD gives you a very practical education in everything that sits behind the extension once the application process is over: registrar relationships, pricing, renewals, policy, distribution and the day-to-day work required to keep a namespace healthy. All of those things matter, but ultimately the market decides whether the namespace is useful enough to keep choosing it.

Looking back, some extensions from the last round also needed time before the world caught up with them. Technology changed, online behavior shifted and words took on new meanings. Other categories that looked compelling in 2012 became less important as the internet moved elsewhere. Applicants in the previous round were making long-term decisions for an online world that looked very different from the one we have today, so inevitably some of those bets aged better than others.

This matters because the temptation in every expansion round is to think of new TLDs primarily as new inventory. Which strings will sell, which categories will be popular and how many registrations can they generate are all reasonable commercial questions.

But I think the more important one comes first:

What is this namespace actually for?

That sounds simple. It isn’t.

Does the word have a coherent meaning? Can it support legitimate use beyond one narrow interpretation? Is there a reason people might still choose it after the novelty of the launch has worn off? Most importantly, can that reasoning survive changes in technology, culture and language that none of us can accurately predict today?

We should be careful not to overlearn the lessons of 2012. A smaller namespace with a clear role may ultimately prove more durable than one that produces impressive early numbers without developing much meaning beyond availability. Equally, a TLD that looks unusual today may make considerably more sense once the internet around it changes.

Some will be broad. Some will be niche. Some may become valuable for reasons nobody applying today has anticipated.

That uncertainty is part of what makes this round interesting. Applicants are choosing words today that may not reach the root for years and could then exist for decades. We are making decisions on a timeline that stretches well beyond the current technology cycle or market trend.

So perhaps the hardest lesson from the last round is that we should be a little less confident about declaring winners and losers too early. A better test is whether a namespace becomes a place people deliberately choose to be and whether it still has a reason to exist years after it is delegated.

Which raises a much harder question for anyone who applied: how do you pick a TLD for an internet that doesn’t exist yet?