Why your next KPI to optimize for should be an opt-out rate

While my full wrap-up of this week’s Productized Conference I wanted to share an interesting thought with you, which got presented by Atlassian’s Claire Drummond and Dragana Boras.

When their team recently rolled-out the excellent new version of Confluence, they established a <5 opt-out rate from the new design as their KPI to optimize for in the first place and to base roll-out decisions on (they even managed to lower this metric to <1,1%).
This caught my eye because it seemed unusual to me to set a ‘negative’ metric as the primary goal to aim for when rolling out an improved version of your product.

But when you look (or think) below the surface, it’s an intriguing way of looking at things.
When thinking of KPIs, I usually group them into additive or destructive categories. Additive metrics are associated with generating value – Usually meaning growth (downloads, sign-ups, contact requests, songs played, upsells, etc.)
Destructive metrics, on the other hand, reduce value for your business. Perfect examples are profile deletions, subscription cancellations, contact deletions and even opt-out rates.

A simplified funnel for that might look like this:

Typical Product Funnel KPI Optimization

Usually, we tend to push for additive metrics when setting the success criteria for a new future. When you’re looking at a self-reflected product team, you might discover a destructive metric as a pre-defined boundary for the project.
It just doesn’t seem ‘ambitious’ enough, to look at the lower part of the funnel.

But what defining a destructive metric as your primary goal shows to me instead is how much you care about the satisfaction of your existing users.
It’s not about hockey stick user growth at any cost or just squeezing some more % of WAU users out of your platform.
It’s about building something which truly resonates with your most loyal users and looking at them not invalidating your efforts by just leaving.

Especially when you’re running a subscription-based product like Atlassian is doing with Confluence, it’s highly recommendable to look at the destructive metrics of your existing user base first and to only optimize for additive metrics when adoption within the core user group has been proven.

As destructive metrics like churns or profile deletions are much harder (and costlier) to reverse then ‘just’ iterating product nuances to boost additive metrics, I think it’s a healthy approach more of us should adopt for future overhauls of their products.