Skip to main content
bradley
Expert ⭐️
August 18, 2026
New Idea

Staircase: Add Severity Indicator on Chrun Risk Signal Insight

Related products:Staircase AI
  • August 18, 2026
  • 3 replies
  • 28 views

Today when you get an alert for a churn risk signal from Staircase, there’s no real gradient on how serious the risk is. A small but potential risk would show up the same as a very high risk (up until the point where it becomes a churn indicator at least). It’s effectively binary.

The AI Risk analyst for the account has a risk level - but this is an account level metric and not a per signal score. And, even though looking into the Risk Reasons surfaced by the “agent”, each individual one has a “Severity High/Medium/Low” that is not translated to the Churn Risk Insights.

General proposal:

  • Include a *reportable* severity number/label for each Churn Risk signal
  • Number vs label likely isn’t terribly important, but thinking about ST>CS workflows, there is already a High/Medium/Low/Critical framework for CTAs, so following that could be a good start.

It’s critical this is reportable, as the use cases aren’t purely informative to CSMs/Leaders, but operational:

  • I may have a Risk Alerts channel but I only want to fire the highest band of alerts there. This is only possible if this is an attribute of the signal.
  • I may want to create a CTA, Playbook, Report, or some other automation off of this data, and report by severity and/or only for certain bands. This would again only be possible if this is an attribute of the signal rather than only available in the UI.
  • Having a threshold could even let organizations suppress the lowest grade in platform in general, though this wouldn’t always be advised.
  • This could also open the possibility of re-surfacing a slightly older risk as increased in severity, rather than a whole new signal. New evidence, same signal? Higher severity, now it gets attention.

In general, I think this would go a long way towards users - especially executives who typically are the deciding voice in signing up/renewing contracts - trust of the system. Right now, all risks are equal, so Staircase (possibly rightfully so) hedging on making a risk notification on a low grade risk looks the same as an all hands on deck issue. When those are compared side by side, the low grade risks look more like mises.

cc ​@bradybluhm 

I have a particular example in mind, but would redact it to the point of uselessness posting here, but have provided the feedback to our account team.

    3 replies

    mobrien14
    Helper ⭐️
    August 18, 2026

    Could see this being a huge value for notification purposes. While our CSMs may need notifications for things, leadership may only want a notification if it’s a higher severity or lower plus paired with a renewal in a set number of days.

    Either way, this type of customization would be big win.

    bradybluhm
    Gainsight Employee ⭐️⭐️
    September 8, 2026

    I like this a lot and like the framing of High/Medium/Low/Critical.

    I think this would go a long way for both churn risk and what's currently named “extremely negative message.” We could change that name to “negative message” and use this same high, medium, low, and critical framing to help differentiate these. Those are the two signals and events that cause the most friction when teammates see them and feel like they are low-urgency events, not truly extremely negative. 

    This would be a big value add for notifications, helping to diversify the delivery and noise levels of notifications for different teams and team members. I like it a lot. I'm going to tee it up for engineering and see when we might be able to squeeze this in. 

    AI on my mind!
    bradley
    bradleyAuthor
    Expert ⭐️
    September 8, 2026

    Love the idea of adding it to the negative messages as well. This is a bit cheeky but maybe we also add it to the positive ones? I have had some (not much) feedback that occasionally it’s a bit over-enthusiastic on what constitutes a positive message. Like someone going “Thanks!” probably doesn’t warrant a notification :)