Engineering Note
Our monitoring showed 2,000+ exceptions. Only 3 actually mattered.
One morning, Sentry showed more than 2,000 exceptions.
The first question wasn't,
"What broke?"
It was,
"Which one actually matters?"
That's a problem many growing SaaS teams face.
As the product scaled, every small issue started creating alerts.
Some were genuine production problems.
Many were duplicate exceptions, bot traffic, or temporary third party failures.
The signal was getting buried in the noise.
So instead of tracking every exception equally, we changed the approach:
* Grouped duplicate errors
* Prioritised customer impacting exceptions
* Ignored known noisy patterns
* Alerted only on high-severity issues
The impact was immediate.
Daily exception alerts dropped by more than 80%, while the team became faster at identifying real production incidents.
The interesting lesson?
More alerts don't improve reliability.
Better signals do.
Because in production, the biggest risk isn't missing an exception.
It's ignoring the important one because thousands of others looked the same.
The first question wasn't,
"What broke?"
It was,
"Which one actually matters?"
That's a problem many growing SaaS teams face.
As the product scaled, every small issue started creating alerts.
Some were genuine production problems.
Many were duplicate exceptions, bot traffic, or temporary third party failures.
The signal was getting buried in the noise.
So instead of tracking every exception equally, we changed the approach:
* Grouped duplicate errors
* Prioritised customer impacting exceptions
* Ignored known noisy patterns
* Alerted only on high-severity issues
The impact was immediate.
Daily exception alerts dropped by more than 80%, while the team became faster at identifying real production incidents.
The interesting lesson?
More alerts don't improve reliability.
Better signals do.
Because in production, the biggest risk isn't missing an exception.
It's ignoring the important one because thousands of others looked the same.