And part of what we're thinking Greg as we go through this, I'll give you an example and this is a generalization, not a specific item, but a rule of thumb might have been that if somebody gets more than a 10% rate increase, they're likely to shop and that's likely it will start to impact your retention or persistency. In a normal environment, there is - this is a supply issue Matt talking about, there is generally broad availability and there's some place to go. In a more restricted environment where folks are either tightening, underwriting - competitors are tightening, underwriting are also raising rates, they might not have an option that's more competitive. So that 20- or 30-point rate change, they might take. In a very low unemployment environment, those folks are needing to have the car to work. They've got incoming cash, so there - that likely has a very important value to them. They need the insurance, they need the car, they need to work that. So that's triggering. And part of what we've been highlighting is sort of a normal model. Hypothetically, if we took - pick your number 25 points of rate, we'd expect a certain persistency drop, we saw persistency go up. If that in fact occurs, one of the questions that comes out is, will that result in at buying decision change a year from now or will people just get used to the new rate and we reset later. When we talk about test and learn from a gradual expansion, what we're actually looking for is trying to measure each of those behavior changes. Right now in this strong employment, low unemployment environment, we're seeing behaviors that just sort of make sense to us, but it's different than historical patterns, we're asking ourselves and are aware and that's why we have a wider margin of safety going forward to say what might happen in the future and we're very much focused on increasing our business intelligence, increasing our predictive analytics, increasing our very scientific method of testing and watching to see whether those behavior changes will move going forward. Does that help?