In this post
In 2022, Clara and I were running outbound for a mid-size logistics platform, and every Monday morning looked the same: a spreadsheet with around four hundred signups from the previous week, and a decision about which ten were worth a call before lunch.
The Monday spreadsheet#
The spreadsheet had every field you'd expect — company size, job title, source, signup date — and none of it told us who was actually close to buying. Between the two of us we could work maybe fifteen accounts a day properly, out of four hundred candidates. We were guessing, and most mornings the guess was wrong.
The CRM had a built-in lead score, but it was mostly firmographic: big company plus senior title equals hot lead. In practice the VP at a 5,000-person company had usually signed up to look around, while the operations manager at a 60-person company who'd been back four times that week was the one ready to buy.
We had four hundred names and no idea which fifteen mattered.
A script written on a Sunday#
The idea that changed things was almost embarrassingly simple. Instead of guessing from who someone said they were, we looked at what they did on the site after signing up — which pages, how often and how recently.
Clara wrote a small script one Sunday that pulled page views from our analytics, scored every account by that behavior and produced a ranked list. On Monday we worked the top of that list instead of the spreadsheet.
Six weeks later, our connect rate on the accounts we called first had roughly doubled. Deals that used to appear “out of nowhere” started showing up on the list days before they asked for a demo. The rest of the sales floor asked for access within a month.
Turning it into a product#
The script had problems a product couldn't have. It ran once a night, so leads were a day old by the time we saw them. It lived on Clara's laptop. And nobody except us could change the scoring rules.
So we started again with three requirements, which still shape everything Aurora does:
- Real time. A lead is most valuable in the hour it shows intent. Aurora now recalculates a score in under 400 ms after every event.
- Explainable. Every score has to come with a one-sentence reason a rep can repeat to their manager.
- Owned by the sales team. The people working the queue should be able to change the rules without filing a ticket.
We launched publicly in 2023. The first hundred teams signed up in eight weeks, almost entirely because RevOps leads told each other about it.
Why we stayed small#
We reached profitability in 2024 with twelve people and chose to grow from revenue rather than raise a large round. That decision is why we can price per seat instead of per lead or per visitor — a model that would be hard to defend to investors expecting usage-based expansion, but one our customers consistently tell us they prefer. (Tomás wrote about the reasoning in Why the free plan has no visitor cap.)
Today we're eighteen people across four time zones, and most of us have carried a quota, run a RevOps function or supported a sales team. That's deliberate.
What we still believe#
- Behavior beats demographics as a predictor of intent.
- A smaller, better queue is worth more than a bigger one.
- Sales tools should earn attention, not demand it — which is why Aurora alerts once per lead, not once per page view.
The spreadsheet is long gone. The Monday-morning question — who should we talk to first? — is still the only one we're trying to answer. If you'd like to help, we're hiring.