Which alerts are worth having on during a season

Alerting is a year-round topic that turns urgent for about six weeks a year. During a season, the value of an alert is set by one thing: how much you spend while it stays quiet.

The arithmetic

A store spends 6,000 SAR a day through peak. Tracking breaks one evening and nobody notices for three days.

• Spend across the period: 6,000 × 3 = 18,000 SAR
• Three days run with no purchase signal returning to the platform
• After the fix, the algorithm needs time to relearn, in the middle of your most expensive week of the year

Worth being precise: the orders aren't lost. Your store recorded every one and the money came in. What's lost is the signal going back to the platform, which is what optimisation learns from and builds its targeting on.

With an hourly check, the same failure exposes you to 6,000 ÷ 24 ≈ 250 SAR instead of 18,000. An alert doesn't prevent failures. It shortens the period you spend paying without seeing.

The three worth having on

Purchase flatline. Zero completed-order events for longer than your longest normal quiet period. In Ramadan the normal quiet period is the daytime lull, not the small hours, which is exactly why a threshold set in February misfires all month. See why fixed thresholds fail.

Store-to-platform gap. Your order count against conversions delivered per destination, diverging past a set point. That's the one that catches a single destination going quiet while the rest keep working.

Pace against target. This isn't a failure detector, it's a commercial one. If your required daily rate has drifted away from your actual rate, you want to know on day twenty, not day twenty-nine.

What to leave off

Everything else. During a season an alert you don't act on is worse than no alert, because the first unhelpful notification you tolerate leads to the fifth, and then someone mutes the channel and every detector behind it stops mattering.

Specifically leave off creative fatigue warnings, minor budget-pacing notices and small day-over-day variance alerts. All are reasonable in an ordinary month and all are noise in a season, where variance is high anyway.

If you have no alerting

The substitute is a deliberate daily check at a fixed hour, in the evening during Ramadan:

1. Conversions delivered per destination: has any one gone quiet since yesterday?
2. Store order count against the sum of delivered conversions: has the gap changed?

Two minutes, and it turns a three-day exposure into a one-day exposure.

In Flowfy, the delivered log with per-event platform responses, revenue by source and conversions delivered per destination are live today, and that's what the manual check reads. Automated detectors, thresholds you set yourself and goal pace alerts haven't shipped. Until they do, the check above is the whole alerting strategy, and it's worth putting on a named person's calendar.

Common questions

How fast would I notice a break without alerting? When someone next opens the dashboard. That's why a fixed daily time matters more than the check itself.

Does a gap mean something is broken? No. Some gap is normal: blocking browsers, excluded bots, events outside the window. What matters is the gap changing, not the gap existing.

Should I check more than once a day during the final ten? Twice a day is reasonable when daily spend is at its annual peak. Beyond that you're watching rather than working.

What do I check first when an alert fires? Credentials, then event mapping. Those two explain most silent failures, and there's detail in six checks that find a break before you do.

What to do tonight

Put the two-minute check at a fixed evening hour and turn on two detectors only: purchase flatline, and the gap between your store and your destinations. The rest can wait until the season ends.