It kinda depends on the details of the statistics, your hypothesis on the issue and the ethical framework that you apply and your own position in that situation.
Assume you have a non-contagious disease and you're literally the only one in the entire observable universe that has that disease. Now if I conduct a study with 2 participants (you and someone else) then 50% are infected. If I ask 100 participants it's only 1% and so on. So despite the same number of people alive and the same number of people infected, depending on the sample size you'd get a different estimate of how many people are infected.
Likewise one should be careful with the actual distribution of cases, like just because something decreased moderately in one place doesn't mean it's a good thing if it massively increased in another, raising rather than lowering the sum total.
But even if we assume that this percentage is perfect information, the distribution would still matter in terms of the question how to react to something. Whether it's peaking in one place or is uniformly distributed. Because if you distribute mitigating means uniformly to a localized problem than most places will have resources they don't need while the place that needs it will end up being short on them. Or vice versa you obliterate a problem in one place but the rest is short on even the basics.
Also the sum total of cases determines as to whether it's feasible to deal with that on the level of the individual (person or case) or whether you're in need of painting with a broader brush. Like is it a single act of criminal conduct or is it some larger pattern or even an inherent problem of the system itself?
So if it is 50% of a group but only 5 people you could have a group discussion or 5 1on1 talks, if it's 50% of millions of people distributed all over the place that is way more difficult to accomplish and having other people that share your problem can itself have consequences that range from bettering to worsening the situation.
So statistics can be tremendously helpful when you're trying to get an overview and when you're free to choose the parameter that you want to look for. But they are estimations, they are often incomplete, you have to be vary careful to look at sample sizes, methodology and what they actually want to and are able to accomplish, and they can be interpreted in various ways so often enough one statistic just prompts you to conduct even more statistics to check the hypotheses that you had constructed after reading the statistic.
Now in terms of human slavery you could for example ask the question of "Is it necessary?". Meaning does any society rely on a subset of that society to do unfree and exploitative labor and is it only a matter of who ends up doing it? Like some sort of trolley problem, where you have a no win situation and someone is going to end up in a bad place. So in that regard reducing the percentage of slaves would be a positive because it means that the situation improved so much that fewer slaves are required to make a society work and that the increase just comes from the population growth.
However that comes with several caveats. First of all it would obviously only require work to be done it wouldn't require slavery. Like nothing in that scenario forces you to force people to do something without freedom, against their will and under exploitative conditions. You might as well ask politely. Like the existence of society often is a net benefit for it's members and it's not impossible for members of a society to personally cut back on luxury and instead invest in a collective improvement that pays out in the future. So even if we assume the "benevolent slave owner" who only does it for the greater good, that person would still not really be benevolent and there's a good chance he's still just doing it to enrich himself and his peer group.
So the argument that slavery is necessary doesn't really work and the decrease in percentage is less likely to be emblematic of a change in necessity and rather of a change in acceptance. Meaning the increasing absolute numbers are still more important and more concerning.
The other question is "did the other parameters remain the same?". Like how was that reduction of slavery in percentages accomplished? Did the conditions for the slaves improve, stay the same or became worse?
So is it more people but less brutal or is it fewer people and even more brutal? Or did the quality of torture remain the same? Is it better to sacrifice the few to save the many and is what they are saved from a minor inconvenience or an existential threat? Or should you place yourself in the role of that individual (categorical imperative) and argue that if that is unbearable for a human there should never be a universal law that mandates or even allows for that? Then the percentage just gives you a scope of the problem but every single one is one too many and a preventable problem. So an increase in absolute numbers is worse and more important than a decrease in percentage.
And there are probably a lot more perspective that you can take on that and ways how you can read such a statistic and what about it matters or matters more given a specific context.