You need to talk about real people, not statistics. What’s 20%? Who gives a shit. More suffering is more suffering, no matter what the percentage is.

We track the change in the number of people living in poverty to the total pop via these statistics. For example if last decade we had 20% of people living in poverty and this decade we have 10% of people living in poverty, that tells us relative to the total population there are less people living in poverty. In other words previously if we had randomly sampled 100 people we would have expected to find approx 20 living in poverty vs now we would expect to only find approx 10 if we randomly sample 100 people.

Bringing poverty down from one percentage to a smaller one as described above describes a success in the sense that poverty is more uncommon compared to the total population.

If P is the total number of people living in poverty, T is the total population and R is the ratio of people living in poverty to the total population then we have R=P/T, in other words P=TR.

Your issue is just that the number of people living in poverty P is too large. But if that's your concern then we either need to decrease T (the total population) or decrease R (the ratio of people living in poverty to total population) or decrease both T and R.

You're arguing that our efforts to decrease R aren't working (or aren't working well enough). So, then what should we do? If we do nothing, R remains fixed (or even increases) and P increases due to the increasing population T, which makes your issue worse. Decreasing the total population T seems tricky too, if that's a viable solution to you, them how do you suppose we should accomplish it? As far as I can tell the only plausible solution is decreasing R, which is exactly what the person you were replying to was talking about?

Note: I'm also ignoring that the rates of change in T and R matter a lot. If you care to argue that we're not decreasing R fast enough, then what would you suggest in order for us to decrease R faster?