Every generation inherits infrastructure. The important question isn’t whether it still works. It’s whether anyone remembers why it exists.

My three year old has entered the stage where every answer is followed by another question.

Why?

He’s not trying to be difficult. He’s trying to understand how the world works. Every answer is simply an invitation to ask another question until eventually you discover whether someone actually understands the reason behind something or has simply accepted it. Sometimes I know the answer. Sometimes, like explaining why thunder can rumble for thirty seconds one day and crack instantly the next, I realize I only understand it well enough to get by.

I’ve been thinking about that a lot lately because I don’t think our industry asks “why” nearly enough anymore.

Advertising has always been built around a remarkably simple objective: help businesses find customers.

Digital advertising didn’t replace that objective. It expanded it. We promised advertisers they could reach the right audience with increasing precision, spend money more efficiently, and understand whether it actually worked. We promised publishers they could monetize the audiences they worked so hard to build without giving up control of those relationships in the process.

Everything we’ve spent the last twenty-five years building has been an attempt to close the gap between those promises and reality.

Ad servers. Exchanges. DSPs. SSPs. Identity graphs. Header bidding. Supply path optimization. Clean rooms. Retail media. AI.

None of those technologies were built because the industry wanted more technology. Each one emerged because the existing approach had reached its limit and someone found a better way to deliver on the underlying promise. Each represented the industry’s best attempt at solving a problem that existed at a particular moment in time.

The mistake is assuming that because a particular implementation solved yesterday’s problem, it deserves to become tomorrow’s starting point.

Every generation inherits the infrastructure built by the generation before it. Some of it remains foundational. Some of it quietly becomes technical debt. The hard part is telling the difference.

We’re even seeing it play out in real time. Just as the industry forgot about Prebid’s rash decision to support the manipulation of the Transaction ID on display inventory, we’re now debating whether Transaction IDs belong in CTV. I’m not picking on Transaction IDs specifically. Trey Titone recently wrote a thoughtful piece arguing how the fastest growing channel in advertising should handle bid duplication, and it’s worth reading because it’s a well-reasoned argument. What made me smile wasn’t the proposal itself. It was the realization that no matter what decade we’re in, adtech’s default instinct is to extend the current abstraction into the next problem. Sometimes that’s exactly the right answer. Sometimes the more interesting question is whether we’ve started with the right abstraction in the first place.

Most institutional inertia doesn’t begin with a bad idea. It begins with someone solving a legitimate problem. The solution works, so other teams build on top of it. Before long there are APIs, reporting, documentation, dependencies, revenue, and customers that all assume this is simply how the system works. Nobody wakes up trying to create institutional inertia. It emerges because every new decision is made on top of the last one.
Eventually, complex industries become experts at preserving answers instead of revisiting questions. Innovation gives us new answers.

Institutional inertia quietly makes the old ones increasingly difficult to question.

Advertising technology is hardly unique here. Every mature industry carries around a collection of inherited assumptions that once made perfect sense. Some of them still do. Others simply remain because nobody remembers the conditions that originally justified them.

You can see it in product roadmaps that optimize around existing infrastructure instead of customer outcomes, integrations that survive long after the technical limitation they solved has disappeared, and organizations that continue measuring success with metrics designed around limitations that no longer exist. None of those decisions were irrational. Most were exactly right when they were made.

Spend enough time in any industry and you start noticing that certain assumptions become strangely difficult to challenge. Nobody can point to the meeting where everyone agreed, but eventually the assumption stops sounding like an opinion and starts sounding like physics. People stop asking whether it’s true because they can no longer remember when it became accepted.

One of the healthiest things a company can do is create an environment where someone can ask an uncomfortable question without everyone assuming the answer already exists. The person asking doesn’t need to be smarter than everyone else. They just need to be carrying a different set of assumptions into the room.

In my experience, the best startups rarely begin by inventing something entirely new. They begin by questioning an assumption that everyone else quietly accepted. Once you stop accepting the premise, the product almost designs itself.

Lately I’ve found myself asking those questions more often. Why does this signal actually improve an advertising decision? Why does this workflow still exist? Which constraints are still real, and which ones have simply become part of the architecture because they’ve been there for a decade?

Sometimes there is a clear answer rooted in technical tradeoffs, customer needs, or market realities. Sometimes the explanation is simply that we’ve inherited a way of working that nobody has felt compelled to revisit.

The underlying objectives of this industry haven’t changed very much.

Advertisers are still trying to find customers more efficiently.

Publishers are still trying to build sustainable businesses without giving away the relationships that make those businesses valuable.

Everything we’ve built in service of those objectives deserves to be questioned periodically, no matter how established it feels.

AI makes this even more important. Large language models are exceptionally good at extending existing patterns. If the assumptions built into our businesses are outdated, AI will often make us more efficient at executing them. Before we ask AI to optimize a workflow, we should probably ask whether the workflow still deserves to exist.

I have very little confidence that the architecture we’re using ten years from now will look much like today’s. Identity has already been reinvented multiple times. Measurement continues to evolve. AI is accelerating the pace of change even further. If history is any guide, the next decade will replace assumptions we currently consider foundational.

At some point Asher will stop asking “why” every five minutes. That’s part of growing up. Eventually we all learn that repeatedly questioning something can sound like disagreement instead of curiosity.

Advertisers will still be trying to find customers.

Publishers will still be trying to build valuable businesses.

Those goals aren’t going anywhere.

The infrastructure we build to get there shouldn’t be immune from scrutiny. If anything, the longer something has existed, the more willing we should be to ask whether the reason still holds.