Number one, and one minute

I kept the report. For a few years I treated it as an answer.

People kept asking whether a free site, built by two engineering students, could matter next to companies raising money at a scale we never touched. The report looked like proof that it could.

It answered a smaller question than I thought.

In March 2019, comScore measured India's Education category. AglaSem came first, with 20.7 million unique visitors and 7.1% reach across the country's digital population. BYJU'S, the company that defined Indian edtech to everyone outside India, sat seventh, at 8.1 million.

What the number was

Unique visitors is a reach metric. It counts how many different people arrived. It says nothing about what happened once they did.

We were very good at the arriving part. We had spent ten years answering the questions an Indian student types into Google at eleven at night. Previous year papers. Admit card dates. Cut-offs. Results. Do that for long enough and a very large number of people arrive.

The same report, a few lines down, said the average visit lasted one minute. Total minutes were down 8% on the month before. Total views were up 15%. More people, more pages, less time.

That is not a failure, and I am not going to pretend to regret it. A student who finds a cut-off in forty seconds and closes the tab has been served well. The one-minute average was the correct output of the thing we meant to build.

The problem is not that the number was low. The problem is everything downstream of it.

What it cost

You cannot charge for forty seconds. Nobody pays a subscription for a page they were relieved to leave.

So the business could only be advertising. Advertising is priced per thousand impressions, and impressions from Indian students are among the cheapest on the internet. That put a ceiling on us that more reach could not lift. Doubling the traffic doubled the revenue and doubled the cost. There was no point in the model where scale started paying for itself.

We were the largest. Largest was worth less than it sounded.

I did not see this clearly in 2019. I saw a report that said number one.

What the company below us was doing

BYJU'S had a third of our visitors and a business I had not seriously considered building. It was not trying to answer a question in forty seconds. It was trying to own a student's evenings for three years, at a price that made one customer worth more than a hundred thousand of our page views.

On the metric comScore published, we beat them comfortably. On the metric their investors were underwriting, we were not in the same category.

What happened next is well documented, and I did not predict it. BYJU'S raised an extraordinary amount of money, became the most valuable education business in the country, and then came apart in public. The depth I envied had been bought at a price that turned out to be the real problem.

Reach without depth has a ceiling. Depth bought at any cost has a cliff. Neither of us had found the thing in between. Being able to say that now is worth more to me than the ranking was.

The same mistake is available again

I work now on how companies appear inside AI assistants. Does ChatGPT name you when someone asks a question you should own? It is early, it is measurable, and it is attracting the same energy that search rankings attracted fifteen years ago.

So the trap is already forming. A citation is a reach metric. It counts how often you were mentioned. It says nothing about who was reading, what they wanted, or whether being named changed anything.

Here is the second number I write next to it now: how many of those answers described the company correctly.

A mention that gets your business wrong is not a win to be scaled. It is a correction to be made. An assistant that names you as the wrong kind of company, in the wrong market, at the wrong price, has just told thousands of people something false with your name attached — and it will keep saying it until the sources it reads say otherwise.

Reach told me how many people arrived. It never told me what they left believing. That is the number I did not have in 2019, and it is the one I would build the dashboard around today.

What I would tell myself in 2019

Keep the report. It was real, it was independently measured, and reach at that scale is genuinely hard to build. I would not take it back.

But write the second number next to the first one. Every time.

The metric you rank on describes your business model. It is not a verdict on your quality. Forget that, and you will start building for the measurement instead of for the person on the other end of it.

One minute was the honest number. It was just not printed as large.