A holding's crisis week: two products together, notifications redesigned

Case · Crisis week

The situation

GRIE Listen and GRIE Watch were being used to monitor a holding company and its subsidiaries. Listen was used at the desk for in-depth analysis and reporting, and Watch was used on the go for instant notifications. Notifications went to the holding's own account and to the managers at the agency serving the holding.

During one crisis week, the holding moved to the center of public conversation within a few days. This page tells the story of that week: what the two products did well, where we fell short, and how we redesigned our product based on the experience.

What happened

Conversation about the holding grew more than fivefold within a few days. On X, daily conversation rose to roughly five times the level of the previous week.

Listen: in-depth analysis and reporting

Crisis alerts reached the team by email throughout the day. During the week, a period insight and post report was prepared for the agency, and a separate summary was produced during the hours when the conversation peaked.

Watch: instant notifications

Priority records reached phones and the Notifications screen every day; on the busiest day, the number of priority records was more than four times that of the first day. The team opened and reviewed these records one by one. But Watch fell short in several places that week. We are not hiding them.

• Too many notifications. The account on the holding side was set to the "All" level and received a notification for every new batch of posts. The day after the busiest day, the user muted the brand for a week. At the time, muting also stopped priority notifications.
• The daily limit filled up early. Agency managers had a daily cap on notifications. On the busiest day, the cap was reached before noon; more than half of that day's priority records ended up in the evening summary instead of reaching phones.
• Priority was defined too broadly. Any mention from a large account counted as priority on its own. More than a quarter of priority records were neutral in tone, and about one in ten had been given priority before its tone was even determined.
• Watch's morning report did not arrive. Because of a bug, Watch's daily report was not generated for most of the week. On those days, reporting continued from Listen.
• We could not see delivery. We could not say for certain which notification had reached whose phone; we had to infer the answer from indirect traces.

What we did

The same week

• We made the notification level adjustable per person. Agency teams now receive only what is important by default.
• We reorganized the Notifications screen: the day's priority records are gathered in a single "Look now" card, and the rest of the feed is condensed into a one-line summary.
• A timed mute no longer stops important posts.
• We fixed the report bug, then went back and filled in the missing daily reports.

The following week: we redesigned notifications from the ground up

• Not every new post goes to your phone. At the "All" level, you receive records for the "Look now" tab, highlights and your own alert rules; the rest of the feed waits for you on screen.
• Quiet hours. Between 11 pm and 8 am, only "Look now" records and volume spike alerts come through; everything else is gathered into a single summary notification in the morning. This is on by default, and you can turn it off in settings.
• Volume spike alert. It lets you know when conversation reaches three times its usual level. When we applied this alert retroactively to that week, it fires on the days the spike happened.
• More accurate priority. We recalibrated priority. With the new settings, four out of five priority records from that same week are filtered out, leaving the most important ones.
• The daily limit is gone. What matters reaches you regardless of how many notifications you have already had that day.
• Delivery log. Whether each notification reached you is now on record.
• Morning briefing for agencies. Every morning at 8:30 am, agency teams receive the status of their brands in a single notification, even on quiet days.
• Reason for muting. When you mute a brand, you can pick one of these options: "too many coming in", "unimportant ones coming in" or "can't look right now".

Outcome

We saw the crisis every day in both products. Listen carried analysis and reporting at the desk, and Watch brought priority records to the team's pockets. But that week, Watch's notifications wore users out, some important records were held until the evening, and Watch's morning report did not arrive.

Today's GRIE Watch would meet the same week differently: fewer but more accurate priorities, a separate alert when volume multiplies, a night when only what matters comes through, and a briefing every morning that gives the status of the brands in a single notification.

What we learned

• The value of a notification lies in its accuracy, not in its number. If users are muting notifications, the problem is not the user, it is the system.
• On a crisis day you need priority, not a cap. Instead of cutting off what matters with a counter, we chose to get better at telling what matters.
• The desk and the pocket complement each other. In-depth analysis and reporting at the desk, instant news in the pocket. When one stumbles, the other carries the work.
• You cannot fix what you cannot measure. The delivery log now answers the question "why didn't I get a notification?" at a glance.
• A product matures on real crisis data. We made these changes based not on desk assumptions but on data from a week that actually happened.