Conversion: Attachment is the signal

Shreman Shrestha leads business at Granola, the AI notetaker that's grown into enterprise deals with $100B+ public companies — almost entirely inbound. For the first half of 2025, he was the company's only salesperson.

For Granola, free-to-paid was never the metric that told them whether they were winning. Shre explains how the signal that matters for PLG products is attachment, and every successful conversion tactic flows from reading that signal accurately: in the data, in who shows up in the inbound form, and in how fast they move once they do.

The real conversion metric is attachment

Conversion, in most early teams' heads, is free-to-paid. That framing misses the thing that actually determines whether a user will stay or close. The question you need to care about is: how do you go from a non-attached user — someone who doesn't have an opinion on your product — to an attached one?

For a long stretch at Granola, we had a paywall that in theory existed, but for some of our most active users, we didn't enforce it. What we cared about was attachment. The earliest signs of PMF looked like: a bunch of users at the same company start using us, the security team finds out, they don't know what Granola is, they shut off access — and it causes an absolute ruckus. I've seen that at tiny startups, and at public companies worth more than $100 billion.

That's the moment you're measuring. Payment is the consequence. The ruckus is the data. Enforcing free-to-paid would have obscured the most important signal, which was whether the product had become part of someone's core workflow.

Volume without intensity is noise

Once you've decided attachment is the metric, the data to watch is simple. Number of users at an account, and intensity of use. The more users inside a single company, the more interesting things get.

Then the question is whether they showed up, tried it for a day, and stopped — or whether they're using it every day. Volume without intensity is noise. Intensity without volume is a hobbyist. Both together is a deal waiting to be closed.

Pilots are the cleanest version of this. When the other company caps a pilot at 10 users and you watch that number move to 20, to 50, to 100 — that's the signal. You don't need a conversion model. You need to watch the line go up.

Five seats can beat five hundred

Quantitative tells you volume. Qualitative tells you who. Our earliest enterprise annual deals were as low as five users. But if you looked up who those five were, they were the two founders and the CPO. Five seats can be a bigger signal than five hundred, depending on which five.

Granola is 100% inbound — the other company initiates the sales conversation. So the inbound form is a pre-qualification layer. I've had founders and CEOs of top companies literally fill it out and ask me, how do I buy this.

The other thing to read is urgency. Are replies coming in two weeks or in a day? Did you exchange Slack in the first meeting? There are prospects I have on WhatsApp from the first meeting, including one founder of a $100B public company. That kind of pace is a proxy for conversion likelihood that you will reliably never see in a usage dashboard.

Champion and decision-maker are not the same person

The best world is when those are the same person. Someone who cares, and who can directly push the purchasing decision through. That's the fastest deal. The second-best world is a strong champion who can thread the executive for you. Anything else is either slow or stuck.

What actually moves deals is making sure the champion has the right information, not more collateral. The common pressure points for us are predictable — people assume we record audio and video, and we don't, we do live transcription — so you anticipate those and have the answer ready. After a few revs, you know the objections before they come.

The rest is an open dialogue. The motion works because people like the product. Making it feel transactional has always felt incredibly silly. We don't have battle cards or spend hours on a deck. There are no attachments or files we give our champions to help them sell internally. I know that sounds atypical. It is, but it works because the product is doing a lot of that work already.

Urgency is your strength

For the first six months of 2025, I was the only salesperson at Granola. That's a constraint, but also a tool. I can't draw a hyper-detailed diagram for your security team. Being upfront about things like that — that it's just me, one lawyer, and reserved legal bandwidth — is often what actually moves things when you're early.

Smallness helps on the product side too. When enterprise asks for a feature list, we push: we're one engineer — if we point that engineer in one or two directions, which is it? You find out fast what's actually a must-have, and what was just on the list because every other vendor has it.

AI hasn't changed any of that, but it does give us flawless recall, so the second call starts in a better place than the first one ended. The relationship is still the most important thing.