How it works

You give probr a domain. Four things happen after that, and it is worth knowing what they are — most of the questions people ask us are really questions about one of these steps.

You add a competitor
  └─ Profiling            once, takes a few minutes

Every day at 06:00 UTC
  └─ Discovery            find what has been published
  └─ Filtering            throw away what changed nothing
  └─ Classification       what kind of event, how much it matters

Every Monday at 06:00 UTC
  └─ Brief                the week, written

Profiling#

The first thing probr does with a new competitor is read them properly: what the company sells, which market it is in, how it positions itself, who it competes with. This produces a profile that everything afterwards is judged against.

That profile is the reason probr can tell a meaningful change from a busy one. A pricing page edit at a company that sells one product for one price is a big deal. The same edit at a company with forty SKUs usually is not. Without knowing which kind of company this is, every change looks the same size.

Profiling takes a few minutes. While it runs, the competitor shows as analyzing and has no signals yet. This is normal and it happens once — after that the competitor is ready and stays ready.

If profiling fails, the competitor shows as failed rather than sitting silent. Usually that means the site could not be read at all: it is behind a login, behind aggressive bot protection, or the domain was mistyped.

Discovery#

Once a day, probr goes looking for anything new about each competitor.

Two kinds of place. First, the channels the company publishes on itself — its blog, its changelog or release notes, its newsroom and press releases, its RSS feed if it has one, its LinkedIn page. These are found during profiling, which is why profiling has to happen before watching can. They are authoritative: a company announcing its own price change is the primary source for it.

Second, what other people write about it — trade press, industry outlets, and general coverage. This catches the things a company does not publish about itself, which is most of the interesting ones: funding, acquisitions, lawsuits, executives leaving.

Everything found is kept as an article, with the original URL. Nothing is paraphrased at this stage and nothing is thrown away yet. This matters because every signal you eventually read traces back to one of these, and the link is always there for you to check.

More channels are being added over time. When one arrives it applies to every competitor you already watch — there is nothing to reconfigure.

The sweep runs at 06:00 UTC daily — early afternoon in Vietnam, overnight in Europe and the Americas. It is not real-time. Something published at noon is usually in your feed the following morning.

Filtering#

This is the step that makes probr worth having, and it is where the AI does its work: reading everything discovery found and deciding what any of it means.

It is mostly a matter of throwing things away.

Companies publish constantly and almost none of it is a change. A reworded hero section. A partner logo added. A webinar. A "we're hiring" post. All of it is real activity and none of it should reach you, because a feed that surfaces everything is the same as no feed at all — you stop reading it in a week.

So probr asks one question of everything it found: did anything about this company actually become different? Most of the time the answer is no, and that article is dropped. What survives becomes a signal.

The consequence you will notice: quiet weeks are real. A competitor that published eleven things and changed nothing produces no signals. That is the system working, not the system broken.

Classification#

Each surviving signal is then described rather than just labelled.

It gets a specific event type — a product launch, a price change, a funding round, an executive departure, a lawsuit, a partnership. It gets an assessment of how much the event actually changes, from minimal to transformative. It gets a direction: whether this is good or bad for the competitor. And it is flagged if it is a rumour rather than something confirmed.

The importance you see in the app is derived from that impact assessment:

Importance sorts and filters the feed. It does not change where a signal goes: everything appears in the feed regardless. Individual signals are not delivered on their own: email and Slack carry the weekly brief, and a day that turns up critical signals produces one inbox notification for the lot rather than one each.

Concepts has the full list of event types and what each field means.

The brief#

Every Monday at 06:00 UTC, the week's signals are turned into something written.

A brief is not a list of the signals you already saw. It has a headline for the week, the one signal that mattered most and why, and a few short sections that connect things — a competitor's launch read against another's price cut, that kind of thing. It is meant to be read in two minutes by someone who has not opened the feed all week.

If there were no signals, there is no brief. We do not send an email that says nothing happened. Some weeks that is the correct answer, and an empty brief every Monday trains you to ignore the real ones.

A brief can be exported as a self-contained page with its own link, which is how most people send one to a colleague. That link needs no login — see Concepts before you share one.

What this means in practice#

Three things follow from the design, and they are the three things people are surprised by:

• A new competitor is quiet for a few minutes, then quiet again until it does something. The first days can look empty.
• probr is a day behind, not a minute behind. It is built for weekly decisions, not for watching a launch live.
• Nothing in your feed is there because a keyword matched. Everything was read and judged, which is why there is so little of it — and also why it is sometimes wrong. That is worth understanding too.