Track test speed, win rate and average win, and count inconclusive tests as failures

David BarnardA few more things on growth. One of the things you shared with me in prepping for this is a kind of equation that you think about. The mechanics of velocity plus win rate delivers compounded growth.

Giordano ContestabileSo we don't think of experimentation as a series of experiments, but we think as a portfolio and a learning portfolio if you want. And very simply, there are three levers that you can pull in terms of effectiveness of the program. Velocity, how many experiments do you ship? We rate what percentage of those are successful, but there is a caveat there that we go through now. And the average win, right?

Giordano ContestabileBasically what we're trying to do now is increase velocity mostly through AI while maintaining win rate and average win value, roughly stable. That's the kind of first step for us in going through this AI transformation.

David BarnardI love that you framed learnings per quarter as a core product metric versus wins per quarter or growth per quarter. It sounds like that was a very intentional framing. And do you actually, I mean, are teams judged on that? Is that like a KPI? Is that a metric that is measured across teams inside Life360? Or is that just kind of like more an operating principle?

Giordano ContestabileIt's very appreciable. It's something that we measure it. I would like to tell you that is the core KPI, but no, revenue is the core KPI that the teams are judged on. But it definitely we... So what we do is we measure the amount of inconclusive tests and we think about how we can reduce that? And we make sure that when there is a lose, a loss, we do have the learnings and we do have time to think about it and act on them.

David BarnardI loved your framing earlier too, that even if an experiment fails, if you learn it can be really powerful. And the thing you mentioned is maybe it just works for pet owners in suburban, small town or whatever. One, what's the system you created to be able to dig that deep into segmentation? And then two, do you then take some of those even tiny segments and do some customized personalized things for those segments?