What Changed While You Weren't Looking
When nobody is paid to watch competitors and regulators, changes that matter arrive late and second-hand. Monitoring that sets trivial changes aside before anything expensive runs makes watching cheap.
Say a competitor changed its prices last Tuesday and a regulator published new guidance three weeks ago. When watching those pages is nobody’s job, you find out late, and usually from someone else.
Watching them is cheap once software sets the trivial changes aside first, so that only the few that could change a decision reach anyone.
The Monitoring Gap
Competitors' pricing pages (five competitors, three product lines each). Regulators' announcements in two jurisdictions. Job boards, for new roles and pay trends. Review sites, yours and your competitors'. News about key clients: acquisitions, leadership changes.
Whatever appears in your LinkedIn feed. A search for a competitor when a client mentions one. Regulatory changes, when your accountant or lawyer mentions them, often after the deadline.
The gap between the two is where the surprises come from.
Why Keyword Alerts Fail
Google Alerts and tools like it match keywords. A competitor’s name triggers on every blog post, press release and social mention, and the few that matter, such as a price change or a new service, are buried in the rest.
The alerts then either flood you until you ignore them, or get switched off, and checking goes back to being done by hand.
Most Changes Do Not Matter
Picture a month of watching:
Illustrative: a month of page changes. The shares show the argument; they are not a measurement.
WatchBase, the monitoring engine we are building, is designed on the assumption that about nine changes in ten are trivial. It scores every change locally, with no paid model call, and a change that scores below its threshold is counted and left alone.
How Filtered Monitoring Works
Scheduled Check
Each page fetched on its own schedule
Find the Change
Compared with the last copy, in structure and text
Score It Locally
A price change or a typo fix? Trivial ones stop here
Explain It
A language model writes one sentence per significant change
Deliver It
Slack, email, webhook or in-app, by significance
The first three steps involve no paid model, so the model runs only on changes that pass the filter. On paper, watching 30 pages daily costs about fifty cents a month in model calls (the arithmetic, at published model prices, not a measured bill). In WatchBase, every step except the sentence in step four runs and is tested; that sentence is the part in build.
What Should Reach You
What reaches you should be short: what changed, why it matters, and a suggested action. WatchBase sends a high-scoring change at once, to Slack, email, webhook and in-app. For the rest, a weekly briefing is the right shape. In WatchBase the briefing comes from the same narration step that is in build, so the sample below was written by hand.
The work that matters is choosing what to watch: the pages where a change would alter a decision, such as competitors’ prices, a regulator’s notices and a key client’s news. The risk is a threshold set too high, which sets aside a change that mattered, so for the first few weeks someone reviews what the filter discarded.
What changed on your competitors’ pages this week?
Book Free Assessment