Notifications people read, instead of ignore

A team receiving a notification for every change learns quickly that none of them require reading.

The expensive part is not the interruption. It is that the important one arrives in the same stream as the noise and gets treated the same way.

Notify the person who acts

A notification going to everybody is a notification belonging to nobody. Each person assumes somebody else will handle it.

One named recipient per event changes that entirely, and the volume drops without losing anything.

The test is whether it needs an action

Something the reader should do today is a notification. Something they might like to know is a report they open when they choose.

Most notification fatigue comes from the second category arriving like the first.

Bulk actions need grouping

An import creating twenty records should not send twenty messages. That single behaviour destroys more trust in a notification system than any other.

One message covering the run says the same thing and stays readable, which is the whole difference.

Say enough that nobody has to open the app

A notification reading that a record changed forces the reader into the system to find out whether it mattered.

Naming what changed, on which record, and by whom lets somebody decide without moving. That is the entire job of the message.

Nobody needs telling about their own action

A notification confirming what the reader just did teaches them the stream is not worth watching.

Excluding the actor is the single cheapest improvement available in most systems.

Separate the urgent from the daily

A payment failure and a completed job do not deserve the same channel. Push, email and an in app list each suit different urgency.

Matching the channel to the consequence makes the loud one meaningful, because it is rare.

Review them after a month of real use

The settings chosen at setup reflect what somebody imagined the work would be. A month of the real thing shows which events actually mattered.

Asking the team which messages they now ignore is faster than any analysis, and the answers are usually unanimous.

Volume is worth counting as well as judging. A person receiving forty messages a day is not reading them, whatever they say in a meeting.

Silence has a cost too

Turning everything off solves the noise and loses the point. An overdue invoice nobody hears about stays overdue.

The aim is a small number of messages people read every time, which is a different target from either extreme.

Let people tune their own

Somebody who can adjust what reaches them stops ignoring the rest, because the stream becomes theirs.

A blanket setting decided centrally ignores that different people act on different things, which is the whole reason the stream got noisy.

Check that the important ones arrive

A message caught by a spam filter or sent to a former address fails silently, which is the worst failure available.

Testing the handful that genuinely matter, once, is worth more than tuning the rest.

Escalation matters for the few nobody can afford to miss. A message nobody opens within a day should reach somebody else rather than sitting unread.

That applies to very few events, which is precisely why it works when it does. A payment failure or a safety report qualifies; a completed job does not.

What to change first

• Each event names one recipient who can act on it
• Anything without an action becomes a report instead
• A bulk action sends one message, not one per record
• Each message says what changed, on what, and who did it
• Nobody hears about their own action
• The team names the messages they ignore, after a month