The Case for Fewer Alerts

Hospitals have a name for what happens when monitors beep too often: alarm fatigue. Staff surrounded by constant alerts, most of which call for no action, slowly stop hearing them. The system built to catch danger ends up training people to tune it out.

Now look at your phone. Price alerts on a dozen tokens. Whale movements. Funding spikes. Group chats. Each was set up with good intentions, and together they have taught you to swipe away the thing that was supposed to protect you.

The answer is not a better notification. It is far fewer of them.

The only test that matters

Before you create an alert, finish this sentence: “When this fires, I will…”

If you can name a specific action, keep it. If the honest ending is “…look at the chart and see how I feel”, delete it. An alert that does not lead to a decision is entertainment, and it spends the attention you will need on the day something truly goes wrong.

A second test helps with the ones that survive. Would you want this one to wake you at 3am? Markets run all night, and your alert system should know which things can wait for morning.

Three kinds worth keeping

Good alerts are about your book, not about the market. They tend to fall into three families.

Limits you set in advance. These fire when your own rules are broken: the drawdown from your high-water mark passes a line, a single position grows past its maximum share, total exposure climbs above what you allowed. They are valuable because they speak for your calm self to your busy self.

Structural hazards. These are the things that can hurt you regardless of your view: distance to liquidation shrinking, margin ratio falling, a stablecoin moving off its peg, a venue halting withdrawals. They are rare and urgent, and they are the ones that deserve to wake you.

Broken theses. Each position should have a condition that would prove the idea wrong, written down at entry. The alert sits at that level. When it fires, the question is simple: does the reason I own this still hold?

Notice what is missing. “BTC crossed $80,000” is not on the list. A round number is a fact about the market, and it says nothing about your book. If you have a plan that depends on that level, attach the alert to the plan.

Make them expensive

Treat each alert as if it costs something, because it does. Give yourself a budget, perhaps ten active alerts across the whole book. Then a new one must displace a weaker one.

Set thresholds from your own data, not round numbers. If a token routinely swings 6% in a day, an alert at 5% will simply fire every day. Tune it until the alert means that something unusual has happened.

The monthly purge

Once a month, look through the alerts that fired. For each one, ask whether you acted, and whether acting helped. Anything you ignored three times in a row gets deleted or rebuilt.

The goal is a phone that stays silent for days. Then, when it finally speaks, you pay attention, because in your whole system of alarms this is the one you have taught yourself to trust.